110107074

INDEX
S. No Topic Page No.

Week 1
1 Introduction to Supply Chain Management 1
2 Evolution of Supply Chain Management 15
3 Analytics in Supply Chain Management 26
4 Supply Chain Planning 41
5 Different views of Supply Chain 54
Week 2
6 Supply Chain Strategy 69
7 Supply Chain Drivers 80
8 Developing Supply Chain Strategy 91
9 Strategic Fit in Supply Chain 104
10 Demand Forecasting in Supply Chain 117
Week 3
11 Bullwhip Effect and Time Series Analysis 130
12 Exponential Smoothing Method of Forecasting 143
13 Measures of Forecasting Errors 153
14 Tracking Signal and Seasonality Models 165
Forecasting using multiple characteristics in Demand Data and Inventory
15 Management in Supply Chain 177
Week 4
16 Inventory Management in Supply Chain 189
17 Multi echelon Inventory Management 200
18 Multi echelon Inventory Management (Continued) 212
19 Multi echelon Inventory Management for four stations 222
Multi echelon Inventory Management for four stations (Numerical
20 Example) 231
Week 5
Multi echelon Inventory Management for four stations (Numerical
21 Example continued) 241
22 Network Design in Supply Chain 251
23 Network Design of Global Supply Chain 262
24 Alternative channels of Distribution 278
25 Location Decisions in Supply Chain 289
Week 6
26 Network Optimization Models 303
27 Using Excel Solver for Network Optimization 315
28 Uncertainty in Network Design 327
29 Network Design in Uncertain Environment and Flexibility 339
30 Flexibility in Supply Chain 350
Week 7
31 Optimal Level of Product Availability in Supply chain 360
32 Time Value of money in Supply Chain 370
33 Different types of Analytics in Supply Chain 380
34 Predictive Modelling in Forecasting in Supply Chain 393
35 Representation on Uncertainty in Supply Chain 408
Week 8
36 Using Decision Tree for handling Uncertainty 418
Example of using Decision Tree incorporating Uncertainty in Single
37 Factor 430
Example of using Decision Tree incorporating Uncertainty in two Key
38 Factors 442
39 Modelling Flexibility in Supply Chain 455
40 Trends, Challenges and Future of Supply Chain 470
Supply Chain Analytics
Prof. Dr. Rajat Agrawal
Department of Management Studies
Indian Institute of Technology-Roorkee
Lecture-01
Introduction to Supply Chain Management
Good afternoon friends, we are going to start our discussion on very important topic for
managers of the day that is supply chain analytics. Supply chain analytics is the combination
of data analytics into the supply chain management. Supply chain management we all know
in one of the most important aspects of business now a days, we know some of the popular
names of fortunate five hundred companies like Walmart, Dell and similarly 7, 11 etc from
Japan Alibaba from China.
These companies are known because of their supply chain competence and nowadays it is
believed that supply chain is one such area in the organization which can provide lot of
competitive advantage to a particular organization. If you go to the website of Harvard
publishing you will find maximum cases in the area of supply chain management and I come
particularly from the domain of operation management.
And now, many of us feel that supply chain management which was once upon a time
considered to be a subset of operation management. And now-a-days the situation is almost
reversed. The supply chain has become an umbrella term and operation management has
become a subset of supply chain management. So therefore, the course of supply chain
management and using the data analytics technique into the supply chain management. So the
name has become supply chain analytics.
(Refer Slide Time: 02:15)
1
In the first, this lecture we will have some kind of introduction of supply chain. The evolution
of supply chain over last hundred years then what are different important decision phases in
supply chain. There are different ways in which we case study the supply chain, those process
views we will see in our discussion in first two lectures which are the introductory lectures of
this course.
Then we will also see what are the importance of various types of flows in the supply chain.
We will also see some very important examples which makes supply chain so popular and
then we will also see the strategic aspects of supply chain management and than we see that
how can we achieve that strategic fit field with respect to the supply chain management and
the competitive strategy of the organization.
So in first two lecture we will have a broad over view of the subject and we will see that how
supply chain management has become the use of data analytics and particularly the present
form of supply chain management where most of the decisions that we are taking with the
help of real time data. So let us start with understanding of what is supply chain.
2
Now supply chain has become such a common term that we find so much mention of this
term in media reports. We find the mention of this term in parliamentary debates, we find the
mention of this term in day to day discussion also. So many times we may wonder whether it
is an academic term or it is more term of media and reports, but now we are going through a
formal course of supply chain management.
And therefore it will be very relevant for us to understand the real meaning of supply chain
and what does it mean, and what are the meanings of supply chain management further?
Now as, we can see that supply chain involves all the stages in the fulfilment of the
requirement of the customer. And these stages may include manufactures, suppliers,
transporters, warehouses, retailers and customers himself. It is very important to understand
3
all time and I will be emphasizing althrough this course that customer is a very much integral
part of supply chain all the time.
Please we need to ensure that customer is always considered, whenever we are taking a
decision related to supply chain, whenever we are talking of supply chain so customer is
always an integral part of the supply chain. Many a time we feel that up to retailer is our
supply chain and customer is not the part of supply chain, but now onwards once we are
going through this formal course please remember that customer is always very integral part
of the supply chain.
The role of supply chain is integration of demand and supply. This is very important, demand
comes from the customers side and supply comes from all these people manufacturer,
supplier, transporters, warehouses, retailers, they all help us in fulfilling the demand of the
customer.
So proper integration of demand and supply that is very much necessary for customer
satisfaction as well as for the profitability of the service providers. So therefore the
importance of this subject is becoming more and more in the coming competitive times and
within each company the supply chain includes all functions involving product development,
marketing, operations, distribution, finance, customer service, after sale services, etc.
Many of us may feel that supply chain is a more related to operations management but if we
talk in a very holistic manner now a days it is no longer specific to operations rather it is
more holistic concept and we require equal involvement of product development, the
involvement of marketing, the involvement of distribution, the involvement of finance, and
after sale services in achieving the objectives of supply chain management.
Because, as I am saying that it is integration of demand and supply. So the demand side
information, the demand side data is normally captured by the marketing people and on the
basis of that data, that information the product development team, the operations team, they
develop new products, they ensure the supply of those products to the customers and then
further wherever the customer is ensuring the supply to that very point, to that very place is
the responsibility of distribution team.
4
So, it is very important that the scope of supply chain management to be understood in a
wider perspective and just not limited to the operation side. So therefore we will see in our
coming discussion that supply chain management has all three verticals of decision making. It
deals at the strategic level also, it is deals at the technical level also and it deals at the
execution level also because of involvement of all the functions.
It is very much similar when we talk of integrated marketing, when we talk of total quality
management or when we come to this class of supply chain management, we talk of
involvement of all functions. So this holisticity is very important in modern management
concepts and that is what we are trying to emphasize.
As I mention that customer is a very much integral part of the supply chain and the supply
chain includes movement of products from supplier, those are here in the left side of the
supply chain to the manufacture and to distributor and also includes movement of
information funds and products in both the direction.
We will see these things in a better pictorial manner in our coming slides. With increasing
competition it is more appropriate, nowadays to say that supply chain is now changing into
the form of a supply network or supply wave.
5
Because when we talk of a chain it is linear like this, but actually it is no longer a linear
system, you have many important players at each of these locations and there maybe a
crisscross at each of these stations and therefore it is more appropriate to say that supply
chain is now becoming supply network or supply wave. Let us see, these are some of the
generic type of supply chains.
Where you can see that in a very conventional type of normal supply chain which we all
know. There is a manufacturer, manufacturer is producing the goods, these goods are coming
to wholesaler, wholesaler is distributing to large number of retailers in the market, and those
products are coming to the customer. This is every conventional type of supply chain and we
all have used this type of supply chain.
But nowadays you can also see you can also understand that there are supply chains where
wholesalers are not there. You have large retail houses like big bazaar, like reliance chains,
like Walmart where probably wholesalers are not there and manufacturer directly supplying
these products to the retailer and retailer those big retailers are directly supplying these
products to the end customers.
And then there, supply chains where manufacturer is directly distributing products to the
customer. For example, BHEL take the example of BHEL. So BHEL is the manufacturer and
the customers are a state electricity board, so if BHEL is making a turbine, so there is no
wholesaler, no retailer in-between and in that supply chain BHEL is directly distributing
those products to the customers, end customers.
6
So depending upon your product to product, depending upon the nature, the unit cost and so
many other factors you may have variety of supply chain. These are some of the generic type
of supply chain and the suitability of the supply chain depends up on the type of products, the
type of customers, the type of intermediaries you require, but as I told you that these are very
linear kind of arrangements.
These three which I have shown here, these are very linear kind of arrangements. But actually
it is not so, actually the situation is like this where you have manufacturer, but many
wholesalers, many retailers are there, and these wholesalers, and retailers are making a
network kind of situation, so these are chains but what I have drawn on the board, this is
more like a network. So, this is supply chain network.
So, in the present environment we say that supply chain network is more appropriate than
simply supply chain. And one more thing which I will, like to emphasize in the beginning
that now a days the competition is not between one organization to another organization. It is
not correct to say that Maruti in India is competing with Tata Motors. Rather it is more
appropriate to say that supply chain of Maruti is competing with supply chain of Tata motors.
So, one supply chain competes with another supply chain and therefore for the success of the
organization it is very important that we develop the competence of entire supply chain. It is
not the competence of a single organization. If Apple is competing with Samsung, so it is not
the Apple and Samsung directly, but it is the result of entire supply chain of Apple which is
competing with the entire supply chain of Samsung.
Because lot of wholesalers, retailers, and there are large number of players on the left-hand
side of these manufactures also. So, in our coming slide we will go to those also and then
only this competition is taking place between one supply chain and another supply chain. So,
these are some generic supply chains but as we go ahead in this course, we will see more
specific supply chains also.
7
Now coming to a very specific example of a supply chain where some customer is going to
purchase some detergent at a big bazaar mall, supermarket and let us see how things are
happening in this real supply chain, those were the generic supply chains in the previous
slide. Now this is an example of a real supply chain. Now in this particular case you see
customer is coming to the big bazaar to purchase the detergent and big bazaar is keeping that
detergent.
Now how many different parties are involved in this entire process. Big bazaar is procuring
the detergent from the Procter and gamble. Now Procter and gamble are manufacturing the
detergent or any other manufacturer which is manufacturing the detergent, it requires plastic
producer, it requires packaging, it requires chemical manufacturer. And they also like plastic
producer is further requiring the supplier like chemical manufacturer to get the plastic
granules.
This packaging supplier this requires timber industry and paper manufacturer. Paper
manufacturer is the supplier to the packaging industry and paper industry requires the timber
industry. And then again, this chemical manufacturer which is required for making the
detergent. So now you see to get a product like detergent from the big bazaar and which is
manufactured by some manufacturer like P and G and someone like that.
You have even timber industry in that supply chain. For a normal customer it is very difficult
to realize that the detergent which I am purchasing from a big bazaar mall is actually
8
originating the supply chain from the timber manufacturer or timber industry. But when you
see in a very holistic manner you will find that this entire supply chain is there.
And now because of you can say limitation because of our interest is not that much, you can
further find that there may be few suppliers before this chemical manufacturer also, there
may be few other suppliers before this paper manufacturer also. I have only listed the timber
industry here. But there may be some other suppliers also before this paper manufacturer. So
just to give you an idea that how real supply chain rather a supply network looks like.
So you do not have only single supplier here rather you have a variety of suppliers supplying
different types of components, different types of raw material to Procter and gamble and then
it is going to make the final product and finally you get a packet of detergent at the big bazaar
mal. So, this way this looks very simple but there are so many individuals to whom we need
to manage and therefore this management can provide a very important competitive desk to
the organization.
Going further this slide is very important to understand that normally we feel or we normally
in our day to day discussion are more concern about the flow of product in the supply chain.
But actually, as I told you in the beginning the second slide that in a supply chain we are
actually having the flow of 3 important elements. And these three important elements are
information, product and funds.
9
You can see the arrows in both the direction in this slide just to tell you normally the primary
thing which flows in the supply chain that is product and product flows from the
manufacturer side to the customer side, product flows from the manufacturer side to the
customer side. But nowadays because of environmental issue and because of so many other
cause related issues lot of products also flow in the reverse direction also.
And therefore, we have arrows with respect to product in both the directions. You have
example of LPG cylinders, so filled LPG cylinders flow from the manufacturer to the
customers, but empty cylinders flow from the customer to the manufacturer for refilling. You
have glass bottles of Pepsi, coke, the filled in bottles flow from manufacturer to the customer,
but the empty bottles flow from the customer to the manufacture side.
So, and in pharma, in FMCG and in variety of these organizations you have expired products
also. So expired product flow from the customer side to the manufacture side for proper
disposal. Information is also very important. Information related to product availability flows
from manufacturer side to the customer side. But information related to what types of
products are required.
In how much quantity these products are required that type of information flows from the
customer side to manufacture side. So that accordingly new products can be develop.
Accordingly, designs can be changed, accordingly quantities can be adjusted for all those
things this information flows in both the directions. Funds that is the third important flow
which is there in the supply chain.
Normally the flow of funds, the source of that is the customer. Customer is the only positive
source of cash flow, or fund flow in this supply chain. But when we talk of reverse logistics,
when we talk of products going from the customer side to the manufacturer side, in that case
funds may flow from that side, left side to the right side also. So therefore, the arrows in all
three cases information, product and funds are in both this directions.
But one thing is very important, flow of information is independent. Flow of information is
independent, why flow of product and funds are related. The direction of flow of products
and funds is always opposite to each other. If products are flowing left to right, so in that case
10
funds will flow right to left and if products are flowing right to left in that case funds will
flow left to right.
So, these directions are restricted directions and the flow of information is independent, it can
flow independently in both the direction. So that is about 3 important flows in supply chain.
Normally, most of our cost related diseases are linked with the flow of products. That how
the flow of product is taking place in your supply chain. But at the same time when we are
talking of supply chain analytics the flow of information is equally important.
Because now a days we are moving for real time decision making of supply chain. And for
that purpose the efficient flow of information, efficient flow of data is very important in this
particular case. So that our decision making becomes more efficient, more useful helping the
supply chain or helping ultimately the objectives of the organization.
Now let us see what are the objectives, typical objectives of a modern day supply chain. So
modern day supply chain the objectives are to maximize the value which we are creating in
the supply chain. The supply chain value is actually the difference between what the final
product which we are offering to customer is worth and the customer is giving how much for
that and this value we can directly relate with the profitability of the supply chain.
11
And just to give you the idea of the process of this value in the supply chain. This slide is
very helpful for us where you see now in this case you have 3 entities to the left side of the
manufacturer and these are vendors, these are suppliers to the manufacturer, which we call
them Tier-1 supplier, Tier-2 supplier, Tier-3 suppliers. They are supplying raw materials,
they are supplying components, they are supplying parts, some assemblies etc. to the
manufacturer.
And most of the value addition from this side till the manufacturer finishes the final product
is because of processing and the manufacturing activities. You are doing lot of processing,
you are doing painting, you want fabrication, and all these things are happening. So
incremental value addition is taking place because of all these activities. Once product leaves
the manufacturer and it reaches wholesaler, retailer, and to the customer.
The value addition is taking place in these phases also. But here the value addition is because
of marketing and logistics activities. So, we need to see that finally when the product reaches
in the hand of the customers. So, what is the value the product has acquired and how much
customer is paying for that value and we want to actually maximize this difference.
So, that difference is actually the profitability of the organization. So it is very important to
understand that we need to do value addition but for that value addition how much customer
is ready to pay. If you do excess value addition and for which it is difficult to get money out
of the customers pocket then your profitability will be certainly under the question mark. So,
12
you need to see that, do not waste your resources in those things where you do not add
significant value and therefore the concepts will help us in achieving that objective.
For an example this slide gives you that if I purchase a mobile phone of micromax at the
price of Rs. 10,000 and the supply chain this is the revenue of the supply chain when I am
paying Rs. 10,000 for purchase of a micromax mobile, so if I am paying Rs. 10,000 so this
revenue of the supply chain. And supply chain incurs cost in getting the information, storage,
transportation, the components, the assembly etc.
So, all that is the cost of supply chain. Now the difference between this Rs.10,000 and sum of
all these cost is the profit of the supply chain. And throughout this course we are actually
targeting to maximize the supply chain profit. So, supply chain profitability is total profit to
be shared across all stages of this supply chain. Now it is also very important to understand
that when we are talking of supply chain.
So, we are talking of total profit and that total profit is shared among all these supply chain
partners, all these stages, so I am not talking of profit of an individual stage, or individual
member of the supply chain. So, supply chain success should be measured by total supply
chain profitability and not profit of an individual stage. And as soon as we start of individual
stage, the very objective of supply chain, the very objective of working together is defeated.
And then the purpose of the competing with supply chain, competing with supply network
will be lost. So all the time we need to keep this in mind and later on with the help of some
13
data and some kind of modelling exercises, we will see that how this supply chain profit
which we are talking in totality is always more than we talk of profit at the individual stage.
So, it is not a more like preaching statement, it can be proved with the help of some kind of
quantitative data also. Now we also need to see the evolution of supply chain over last 100
years and we will use our part 2 of the lecture to discuss the evolution of supply chain
management, so at the time we are just stopping here to discuss that what is supply chain and
how we need to do business in a supply chain environment with the help of coordination with
the help of all entities working together.
And not to think for individual supply chain, individual gains, or individual profits, it is a
business with the help of coordination, with the help of cooperation of all the members of
supply chain and therefore the success of supply chain lies in the coordination supply chain
success lies in trusting each other and these are the fundamentals of supply chain
management. Thank you very much.
14
Lecture-02
Evolution of Supply Chain Management
So in our first lecture we discuss the role of supply chain, the importance of supply chain and
how the supply chain has become such a very important function for modern day business.
Now we will start discussion of this lecture with evolution of supply chain management in
last 100 years, if we start the discussion of evolution the first very important case for the first
important revolution which is in the beginning of 20th century around 1910 and 1920 that
time and this is characterized by the ford supply chain. The Ford motor company is pioneered
in integrating the entire supply chain.
Right from birth to death of the car ford, supply chain is known to be one of the most
efficient supply chain, on one side Ford used to have its own Iron ore and on the other side
they used to distribute the cars, finished car on the market, so on one side they were having
the mining business, they were used to process the iron ore to get the steel, use that steel in
making the car and then distribute the car.
And they efficient it that system so well that it is documented, that Ford use to deliver a car
from mine to the retailer in just 81 hours and therefore ford example is one of the most
pioneered example in the case of efficient supply chain that in 81 hours ford can deliver and
15
it was a very well saying at that time that as long as a car is of black colour it can be delivered
by ford.
And as long as it is T model, it can be delivered by Ford. So, T and black these are the
symbols of efficient supply chains. Then the second important phase of the development of
supply chain is from Japan, when Toyota created a different type of model.
In the model of ford right from the beginning where you have iron ore mining to customer,
all the activities which are required in making the car are owned by ford company. But
nowadays in this Japanese revolution which came around 1950s and 1960s after Second
World War, now what they did, Toyota developed a pool of vendors and these pool of
vendors they used to supply different types of products, components, assemblies, sub-
assemblies to Toyota.
And then Toyota used to distribute these finish cars to the customers, so instead of doing all
things on its own, Toyota started developing the capabilities of the vendors and actually this
is a model which nowadays most of the companies follow, now it is very rare, though the
example of ford is always known, always popular for its very high level of efficiency.
But there are certain limitations also with this ford model that it was almost inflexible
because you are controlling everything in your supply chain, so it is very difficult to change
the product and therefore only black colour, only T model became the synonyms of ford
16
supply chain. On the other hand you have a pool of vendors with their own expertise and now
the Toyota become more responsive to understand the customer's requirement.
So this ford model which was totally inflexible and with this Toyota model which came
around 50 years after this initial development of supply chain concept, so we started moving
from this inflexible model to a bit of flexibility in the supply chain. So some flexibility stared
coming in the supply chain from ford to Toyota and now in Toyota, you have many owners
of the supply chain, so many vendors are there.
And all of them share some part of ownership in the entire supply chain and then very
recently you can say almost at the end of 20th century or around beginning of twenty first
century at that time the third revolution which is more IT driven revolution in the supply
chain. This IT driven revolution is the Dell supply chain and that is also very interesting type
of development in the supply chain.
And this development helps us in understanding the supply chain in present context that what
is happening in the present scenario in the supply chain. Now Dell powered on the advantage
of Information Technology, so Dell used to get information from the customer and Dell used
to pass this information to its various vendors, these vendors may be located at different
locations, these vendors provide different components to Dell.
So a customer normally we all know the capability of Dell, Dell was known to provide the
customised products to its customers and we were having the opportunity to design our
machine, design our laptop, design our computer on the Dell's website and Dell used to
collect this information from the customer and depending upon whatever specification,
whatever configuration a customer has ordered this information was passed to different
vendors.
Now all these vendors they supply their components to Dell and then Dell used to assemble
them in a single packet, so that packet will go to the customer. So now this is leveraging the
power of information technology for the better customer satisfaction, so that customer get the
unique product which he or she is requiring. But over a period of time Dell realised that all
the customers.
17
All the normal customers like me, like you, we are giving almost similar kind of
configuration for our requirements and then till 2006 Dell was giving only it's product
through online ordering, but when Dell realized that now customers are not so particular,
customers are so unique with respect to their requirements, so from 2006 onwards Dell
changed its model of supply chain.
And after that Dell's products are also available through retailers, through other conventional
supply chains, because earlier time we for the sake of uniqueness of our orders were ready to
wait for 10 days, 15 days from the Dell's website, but when we are not having so unique
requirements, so why will I wait and this question came in the minds of officers of Dell
company and therefore they took a very drastic decision of changing the supply chain of Dell
and 2006 onwards Dell change it supply chain.
But nevertheless Dell became a very interesting case in the modern supply chain where we
can see that how you can integrate various vendors and your customers just with the power of
your information technology. So these are three important changes which has happened in
last 100 years and the model of Dell, this model of Dell which is based on information
technology power, this is actually the starting point of supply chain analytics.
Where we are using real time data, where we started using information for the success of our
supply chain. In these two cases the role of information was there, but it was not up to the
extent at which Dell started exploiting the use of information for the success of its
organisation. So the evolution of supply chain tells us that what are the important changes
which has taken place in last 100 years and now the current model of Dell supply chain is
based on that information technology the real time data analysis and that will become the
basis of our supply chain analytics course.
18
So now once we have understood this development of supply chain in last 100 years, so let us
quickly review the objective of the supply chain and the objective of supply chain is the
management of flows between and among supply chain to maximize total supply chain
profitability. So I again and again always request you that this total word is very important
for the success of supply chain, we never talk of individual stage, we always talk of totality.
And therefore all the decisions are of totality nature. Before we go into the supply chain
analytics let us see what are the specific challenges of current supply chain which we all see
and these challenges will give us some kind of appreciation that what is the requirement of
supply chain analytics, why are we making the simple things so complicated and therefore
this particular slide will help us to understand some of the challenges.
19
This list may not be very exhausted because depending upon the type of supply chain you
encounter depending upon a particular market, depending upon a particular type of product
category, the challenges may be many more, but some generic challenges considering the
current business environment I highlighted here and these are, first is lack of synchronisation
between planning and execution.
We do not have what we plan and what we execute, so there is always a difference between
that. So we if whatever we are planning and we are not able to execute that, so the objectives
of supply chain which we have set for our self will not be achieved, so therefore this is the
first important issue related to success of supply chain that we lack the synchronisation
between planning and execution.
The second important challenge is lack of real time data visibility and the supply chain
analytics type of interventions will help us in improving this challenge of real time data
visibility with no common view across all businesses and channels, in fact different persons
in the supply chain, if we leave this, then model if you have various intermediaries in your
supply chain they all we have different sources of data collection.
And maybe there will not be consistency in the data collected by different intermediaries, so
there is no common view and when you have lack of consistency in the data or information
available with different channel partners obviously lot of problems are going to happen, then
irregular reviews of safety stock levels causing frequent stock outs and excess inventory. As a
customer when I visit a retailer I always like to have whatever product I want it should be
available readily in this stock with that retailer.
But because of poor safety stock levels it is quite possible that out of 10 times I visit a retailer
only 5 times that product is available, so very high stock out of that product and if it is very
high stock out like I explained my customer satisfaction will go down and it is also possible if
a product which is not in demand so much and you are keeping the inventory of that product.
And if it is happening that way the product is not in demand and you are keeping the
inventory, so it is unnecessarily going to increase the cost of your supply chain and that will
pull down the profitability of your supply chain. So it is very important that you should have
20
a very proper review of your safety stocks and in supply chain we talk of a term which is very
interesting and lot of our discussion in the coming classes will be based on that term.
And this is known as bullwhip effect, where because of improper reviews you keep on
collecting excess inventory at each stage of supply chain and therefore that inventory in your
entire supply chain increases and creates a total failure of your supply chain management.
Bullwhip effect is a big threat to the profitability of the supplier. So whenever we talk of
supply chain immediately we need to find a good solution and efficient solution for the
problems like bullwhip effect.
So, this is a very serious challenge and certainly the advantage of data analytics can help us in
minimising the issues of frequent stock out or excess inventory. Then other important issue
another important challenge which is there in the supply chain and that is the need of the our
lack of flexibility in the network, our supply chains are not very flexible and this case we
have already discussed just now with the example of this ford supply chain which was totally
in flexible.
Only known for two things, T car and black cars, but nowadays we all know that you cannot
fulfill the demand of a customer just by providing only one colour and one model. You need
to provide large number of colours and you need to provide large variants also, so you need
more flexibility and this flexibility is required with respect to variety, this flexibility is
required with respect to quantities and therefore we need flexible supply chains.
But right now we have very limited flexibility in our supply chain, and therefore it is the data
analytics activities which may help us in improving our flexibility aspect of the supply chain
and this will certainly be a very important area and we will like to deliberate more on related
to flexibility aspects of supply chain in our future classes, future lectures, then another
important challenge of supply chain is price, volatility, and difficulty in the de-risking.
A lot of papers you can find in the area of supply chain risk management and you have lot of
threats for the proper supply chain management and we also discuss just now that there is a
issue between planning and execution, so lot of academic discussions are going on with
respect to de-risking of supply chain that how you minimise the risk in your supply chain, but
21
because of globalisation, because of a lot of exchange issues, because of price wars between
the competitors.
And all these things are posing regular challenges to the supply chain and probably we need
more futuristic decision making where data analytics will come very handy for us that how to
make that futuristic decision and that will probably help us in de-risking our supply chain. So
this is also a very important area where lot of things are to be done. Then another challenge
which supply chains are facing currently that is related to production line imbalance and
suboptimal batch sizing which creates asset underutilization.
Because of for the sake of flexibility you can say, for the sake of doing better customer
service, for the sake of better customer satisfaction for all these things we are doing lot of
things which are resulting in underutilization of assets, so actually we need to have a very
important trade off between economies of scale, asset utilisation, line balancing and customer
satisfaction, flexibility service level.
Because these two things or some kind of trade off, if you go for one you have to sacrifice
other, if you want to achieve higher asset utilisation and at the same time you are looking for
flexibility, then you need to do some kind of trade off, it is very difficult to achieve both these
things simultaneously, but the current requirement says that you need to achieve flexibility
also, you need to have higher asset utilisation also.
This will help you in reducing the cost and this will help you in better customer satisfaction
so we need a quantifiable trade-off between these two varying aspect, so these are the
challenges which are in front of us and we will like to solve these challenges with the help of
supply chain analytics. So let us see what is this all about supply chain analytics, so supply
chain analytics as we discussed in the first lecture is all about integrating the data analytics
into the supply chain management.
22
And this supply chain analytics aims to improve operational efficiency and effectiveness by
enabling data driven decisions, at all 3 levels that are strategic level, operational level and
tactical level, so all three levels of decision making in the organisation, strategic, tactical and
operational level. We want to take data into consideration and on the basis of this data which
is available with us we will like to take efficient and effective decisions.
So now the science is coming more into play for the development of supply chain decisions.
Many a times our decisions may not be rational, many a times decisions are based on lot of
qualitative factors, but this supply chain analytics is one area which try to capture even the
qualitative data also or qualitative data also it is not only the numerical data which is
important here.
Qualitative data can also be very useful in taking the decision, but the data at all the level,
strategic, operational and tactical level so decisions if are taken with the help of proper
background data available to us, then it is expected that those decisions will help us in a
better, efficient and effective supply chain. So this is how we can understand the meaning of
supply chain analytics.
And we will see the use of certain algorithms, we will see the uses of some of the modelling
techniques for making the decisions at all these three different levels of supply chain. Now it
also encompasses the complete value chain. The complete value chain which we will see in
the next slide started from the sourcing and up to the logistics including the manufacturing,
including the distribution and all the aspects.
23
So the supply chain analytics is not only limited to a particular area of the supply chain, it
takes care of your entire value chain right from the sourcing of raw material or sourcing of
components or sub-assemblies to the logistics and distribution of products through the
customers end.
So just to give you an idea of supply chain value, so this is the value chain which I am
talking, that right from the development of new product to the services. All these functions
right of new product which is based on the information provided by the customer, so the role
of data comes into picture here that what type of information, what type of data is being
provided by the customer and whether the new product fulfills those aspects of the
requirement.
Then marketing and sales which are very important to capture that data from the customer,
because if marketing and sales people are unable to capture the right data of the customer it is
very difficult to develop new product as per their requirements, then operation people are
responsible to add value to, add the incremental value into the component, so that product is
produced as per the specification.
So they are also based on the background data available with them with distribution and
services where the customer is we need at what time customer requires that product, it is also
very important that the place and the timing, so data related to place and timing that will help
24
us to make efficient and effective decision with respect to distribution, at what stage of
product consumption the services will be required.
So data related to product uses, how the customer is using the product and it is very
interesting and we all know our self that many a times we use products in variety of
innovative ways and as a service guy I need to know that what are the different ways in
which a product can be used, interestingly when I do a class on innovations management we
normally ask that what are the different innovative uses of toothbrush.
Now toothbrush, we all know we use it for cleaning the teeth, but when we ask this question
in a class of innovation management you can find variety of innovative uses of toothbrush
may be as wide as cleaning the floor, as wide as for making the paintings. So just to give you
an example, that is, service guy must have all the data that in how many conditions in what
type of situations my product can be used and accordingly that data will help me to take
efficient and effective decision with respect to my service requirement.
So all these decisions which are there at the different stages of value chain if all these
decisions are taken with the help of data, these decisions will be very efficient and effective
and with the help of supply chain analytics we will try to take decisions at all these stages
which will make the purpose of supply chain analytics in line with the supply chain
management which is in line with the business strategy of the organisation.
And that is in line with the overall objective of the organisation, so in this lecture we
discussed the evolution of supply chain management and the key challenges which our
supply chains of the modern day are facing and then we also discuss that how data analytics
can help us in solving these challenges and what is the purpose of data analytics and what is
the meaning of supply chain analytics and objectives of supply chain analytics.
So now in next class we will discuss the various strategic aspect of the supply chain
management and which will give us food for thought for our further classes that how these
strategic issues can be handled with the help of available data. Thank you very much.
25
Lecture-03
Analytics in Supply Chain Management
Welcome back. We already have 2 lectures on this topic of supply chain analytics. The first
lecture helped us to understand the basic principles of supply chain management. We
discussed the importance of supply chain in the current business scenario. And in the second
lecture we started that how supply chain has evolved over last 100 years and then we also
discuss the use of analytics in supply chain decision making.
That what are the current challenges of supply chain management and how analytics can be
very handy for improving those challenges to solve those challenges and in todays discussion
we will further like to elaborate on those very aspects of using data, using data driven
decision making for the supply chain decisions or various levels of supply chain. Just to re-
capsulate whatever we have discussed in our earlier 2 lectures.
We discussed that supply chain starts from vendors on one side then there is a manufacturer
and then you have distributor till consumer uses those product. And we also emphasised that
now a days it is more supply network and supply wave, these type of terms are more popular,
the reason being when we talk of chain it is more like a linear phenomena, but when we see
that at each stage of chain there are so many entities which are involved.
And therefore, it is more appropriate to say that it is now supply network or supply wave.
And now moving further into the discussion that we discussed about the initial idea of supply
chain which is started from Ford motor company, which used to control the mining of iron
ore on one side and distribution of finished car on the other side. So vary integrated supply
chain was conceptualized by Ford motor company.
And because of the high level of integration the supply chain was very much efficient, but at
the same time the problem was lack of flexibility. As we discussed in our second lecture that
it is more popular with the name of supply chain sub line only black and T model of cars.
26
Then we discussed about Toyota concept, where Toyota developed a pool of vendors. And it
started incorporating flexibility into the supply chain.
And over a period of time most of other industries maybe electronics, maybe consumer
durables, maybe even FMCGs. All these types of industrial segments started adopting Toyota
model of supply chain. Late in the 20 th century or in the beginning of 21 st century we have
this IT revolution and as a result of this IT revolution we have another very important
revolution in the development of supply chain phenomena.
And this model is characterised as we discuss with the name of Dell company. The Dell used
power of information technology for delivering their products in a highly customized fashion
and this model of supply chain became very popular model because of competition, because
of increasing expectations of the customer. Now we need very specified customized products.
And through the power of IT, Dell was able to deliver a high degree of customized products
to its customer.
So that is the latest phenomena. Now in last 2 decades there are very rapid changes which are
happening in the business environment and at the same time there are rapid changes which
are happening in the technological environment also. When I am talking of technology so my
focus is more on the information technology. And now the current wave of supply chain
development is integrating both these things.
It is integrating the development of supply chain management as a concept, as a theory, as a

practice and on the other side it is taking care of development of information technology, data
management techniques and now the supply chain analytics will provide a very good
platform where we are going to use both these technologies, both these concepts for solving
some of the challenges which we discussed in our lecture number 2.
27
We will like to re-capsulate all those challenges in today's lecture also. So just to give you
these summary of whatever we discussed in lecture number 2 regarding supply chain
analytics. So supply chain analytics plays a very key role in enhancing the performance of
supply chain by improving supply chain visibility, you see these 3 important activities we
expect from supply chain analytics.
One is the visibility, now the visibility is very important because of customer experience.
There was a time when we use to send a letter in the post office and it was like a black box
that you just put a letter whether it is registered letter in the post office and you do not know
at what time your letter will be delivered to the receiver. But now a days you have a tracking
number available with you.
And continuously on internet, on the given portal you can put that tracking number to track
the movement of that letter, that packet. Large number of courier companies are putting RFID
tags in the packets, so that the online movement of those packets can be tracked and
accordingly customer is having a better experience about those products. At in variety of
critical items like if I talk of disaster supply chain, if I talk of medicines, if I talk of some
emergency equipments the supply chain visibility becomes very important.
In almost all e-commerce sites whenever we purchase a product we get a portal number, a
document number and through that document number you can always see, you can always
have this visibility component satisfied that where is the location of your order number. So
28
this supply chain visibility is a very new and interesting area to give you a better customer
experience.
The second is volatility, managing unprecedented changes in the supply chain. And for that
purpose the supply chain of previous years supply chain of forth time may not be suitable in
present context. So we need more flexible supply chains to handle issues like volatility. And
flexibility can be provided with the tools like supply chain analytics. So we will see that how
real time data availability can help us to inculcate more flexibility into the supply chain.
That is what we mean with managing these disruptions. And then reducing fluctuations in the
cost. Yesterday in our lecture number 2 we discussed that there are issues related to frequent
stock outs or issues related to excess inventories in this supply chain. All these things create
fluctuations in the cost of offering a product to the customer. If I go for the sake of flexibility,
too much changes of my production system.
All these things will lead to fluctuations in the cost. So I certainly need to be very careful that
there should not be too much fluctuations in the cost because ultimately if there is too much
fluctuations in the cost I cannot have a good idea about the profitability of my overall supply
chain. So these 3 important issues are there which we will like to develop with the help of
supply chain analytics.
So visibility of global supply chain and logistic processes that is certainly one area which we
will like to work on. The second is to manage the demand flexibility as we discussed
29
yesterday also. So we need to manage the demand with respect to varieties also, new type of
products, more designs, more variations are required by the customer and at the same time
you do not know what quantities a customer may require all of sudden.
Sometime in some products like if I talk of salt, so you have a very steady demand pattern of
salt. When Apple launches iphone7, so Apple does not have any idea that on day 1 what will
be the demand of this iphone7. So handling products like iphone7 versus handling products
like a routine salt requires 2 different types of supply chain strategies and as you move
towards new automobile products as you move towards new electronic products as you move
towards new type of consumer durables.
Because when we are discussing this supply chain analytics at the same time you all know
that there is lot of emphasis on innovation. And when innovation is coming into the new
product development this problem of demand uncertainty. This problem of demand
uncertainty further increases and for that purpose our supply chain need to be very
responsive. If our supply chain is not responsive if our supply chain cannot provide
immediate solutions.
For these changes in the demand probably we will be behind the race, we will not be able to
compete with the most competitive organization. So, managing the changes in the demand,
managing the fluctuations of the demand, that is also very important. And third issue is how
do we manage the fluctuations with respect to cost in the supply chain. There are 2 very
important types of cost which we talk in supply chain anytime.
30
And these 2 types of cost include the fixed cost and then variable cost. In a supply chain we
always try to minimize the combination of fixed and variable cost. You need to design, you
need to develop your supply chain processes in such a fashion that you actually get finally the
minimum fixed and variable cost. Fixed cost is that cost which is independent of units and the
variable cost is proportional of units.
So you need to do a proper exercise, you need to do a proper decision making, so that you
have minimum of fixed and variable cost. And for that purpose again, analytics will be a very
handy tool that will help us to optimize the cost structure of a supply chain.
So these are 3 important issues and let us see that how analytics in supply chain will help us
in addressing these 3 important issues of supply chain visibility, demand flexibility and
31
reducing the cost of fluctuations in our supply chain. So let us see that, first is we need to
move to a smarter logistics to improve supply chain visibility. We need to move into the
direction of a smarter logistics.
So far this term smarter was not there. So far in our logistics practice it is simply logistics for
supply chain. We used to say in our conventional supply chain classes that we need to have a
robust, efficient, effective logistic systems for supply chain management. But now a days
because we are talking of supply chain visibility, so we are moving towards a smarter
logistics and for this purpose of a smarter what is the meaning of this smarter.
The meaning of smarter is where my supply chain can take decision on its own. I need to
build that type of data, that type of sensors, that type of objective, that supply chain can take
decisions on its own for improving the better visibility of my supply chain objects. So that
component of developing the smarter logistics is related to date driven decision making and
analytics will be directly related to make my supply chain, a smarter supply chains.
Then second is to manage the uncertainty of demand. Now to manage the uncertainty of
demand. The only solution available to us is to have a very effective, efficient inventory
management. But if it is not effective, so inventory management or inventory particularly can
be a disaster also. In most of the supply chain, we at the end of the day find that because of
improper inventory management.
The cost of supply chain has increased tremendously and dead weight of inventory has eaten
up the entire profitability of your supply chain. So we need to find better ways of inventory
management, which are leading to customer satisfaction because for the sake of good
inventory management, if you are going for very less safety stocks then probably you can end
up with stock out situations also.
So that is also not desirable. So, for that purpose we need to have a smart inventory
management where you can take decisions on inventory management on the real time
databases. Realtime databases means normally we in a conventional system follow P or Q
type of inventory management either we have fixed period of review that after every 1 week
or every 2 week we are going to review the inventory position or we on a certain basis we
have a fixed quantity of order and that keeps on going after reaching the reorder point.
32
So whenever we reach that reorder point we give order of a fresh quantity, so that our stocks
are replenished. But in both these situations we are either facing the problem of stock outs or
we are facing the problem of excess inventory. Both are not desirable and at the same time
we are also facing the problem of this uncertainty in our demand particularly, so we need to
have more real time data analysis for our inventory management.
And again analytics will be very much useful for achieving this objective of a real time
inventory management where as soon as any item is taken away from your stocks. So that
data is captured and that data is flown in the entire supply chain. So in our technical term we
say that Point of Sales data, POS. We now need to use this POS data for the purpose of
inventory management.
Point of Sales data and this point of sales data is available simultaneously at all stages of
supply chain. The supply chain let us have a figure of that first.
So here my customer is there and some sales is occurring at the customer end. Now this POS
data is generated here and as soon as this data is generated at this customer end, the customer
is going for the purchase at the retailer, so he picks up this product, so at this retailer
inventory is subtracted by 1 unit. If he purchases 1 AC, 1 refrigerator, 1 colour television, 1
mobile.
33
So since all parties, retailer, wholesaler and manufacturer, all of them have an integrated
systems and as soon as customer purchases product from the retailer, there is no need to
communicate that information separately to wholesaler or to manufacturer. Because it is a
continuous integration, so this data is immediately available to wholesaler and manufacturer.
And on the basis of this information, basis of these types of data coming from different
retailers and different wholesalers.
Each of these different entities can plan their inventory in the real time environment. So we
want to have more you can say better inventory management, smarter inventory management
with the help of this real time data analysis. So again that is all about analytics that how do
we do this real time data analysis. Then third is this, reducing cost fluctuations, by
optimizing, sourcing and logistics activity.
Now you see there are various wholesalers, various retailers at different stages and products
are flowing from wholesalers to retailers and with each retailer there are n number of
customers also. Now we need to optimize again with the real time data that from which
wholesaler, how many products will go to a particular retailer, from which manufacturing
facility, how many products will go to a particular wholesaler.
So that data which is available to us will help us to optimize our sourcing and transportation
movement facility side related activities. So analytics will help us in reducing the cost related
to sourcing and logistic activity. Because if we do not do this data driven exercises because
34
we always need to keep this in mind that whenever we talking of analytics it is all about data
driven analysis, it is data driven decision making, it is data driven implementation.
So everything is data driven, so all these activities which are most important for the present
time, one is for the customer side supply chain visibility type of thing where customer is able
to actually continuously track the movement of the packets, the products, and on the other
side for the company to optimize the availability, to optimize the souring, to optimize the
cost, and all those aspect.
So, analytics will help us at all level of supply chain to achieve these 3 objectives. So with
this now, we are very clear I hope that what are the uses of analytics data driven decision
making in our supply chain. So once we are through with this now let us move to some more
fundamental issues related to supply chain management.
Now in supply chain as we have discussed yesterday also that data analytics, a very important
at all 3 levels. And 3 levels are you have the strategic level of supply chain where you design
the entire supply chain, then the second level is the planning level, where you plan to
implement the strategy and the third level is the operation where you actually execute those
plans. So these are the 3 phases of the supply chain.
And the data analytics play important role in the strategic and the planning and at the
operational level of the supply chain. Tactical word is the another word which we use for the
planning, so yesterday if you remember in the lecture we discussed the strategic, tactical and
35
operational level. So tactical and planning are used synonymously and we will be using these
2 terms interchangeably throughout this course. So now let us see what are the important
activities in these 3 different phases.
So in the supply chain strategy, the designing phase of the supply chain in that we take
decisions about the structure of the supply chain. The generic supply chains which we
discussed in our lecture 1, in that we have some very common type of supply chain designs,
so which particular design is suitable for our type of organization, our type of product, our
type of target market, that type of decisions are normally taken.
That type of choice of structure is done in case of supply chain strategy, most important
decision is how many intermediaries, we are going to have, whether there will be a retailer or
not, whether there will b ae wholesaler or not, these type of decisions are normally the most
important decision in case of supply chain strategy like we discussed yesterday.
The example of Dell, where Dell earlier use to distribute their products through online system
and it was Dell was directly dealing with the customer. So there was no intermediary between
Dell company and the customer, no wholesaler, no retailer, but when Dell changed its supply
chain so we use to say that Dell changed its supply chain strategy and it started using retailers
in between.
So that is a type of strategic decisions that what is the structure of the supply chain, whether I
want to have retailer, or I do not want to have retailer. These type of issues are the most
36
important issues with respect to design of a supply chain and therefore these decisions are
taken by the top of the organization. Those who are may be the board of directors, CEO, MD,
CMD, these types of people they take the decision with respect to supply chain structure.
Then another important strategic decisions are the location and capacity of the facility. I want
now I have decided that I want to have wholesalers, and retailers also in my supply chain.
Now whether I want to have a wholesaler in Maharashtra or I want to have a wholesaler in

Gujarat or I want to have a wholesaler in Goa, then another wholesaler I want in South India.
So whether to have in Tamil Nadu, whether to have it in Telangana, whether to have it in
Andra Pradesh. So issues like that, so where these wholesalers will be and what will be the
capacity of these may be all these wholesalers because of their further target market may not
be of equal capacity.
So a wholesaler in Tamil Nadu may be much bigger than a wholesaler at Goa. Then it is not
about only wholesalers, we take decisions because all these manufacturers, wholesalers,
retailers, all of them are known as facilities. So I have to take decision about manufacturers
also that where should I have my manufacturing facility, where should I have my
wholesalers, where should I have my retailers, and what will be the size as I just discuss.
So location and capacities of the facility that is one type of decisions which are the part of my
supply chain strategy. Then second important strategic decision is about the products to be
made or stored at various locations.
37
You have 3 manufacturing facility in the country, I have let say Hero Moto corp. and Hero
Moto corp. and one of my facility is at Haridwar, 1 is at Dharuhera and another is at
Gurgaon, which is Gurugram now. Now I will decide that my Splender motor cycle will be
made at my Dharuhera plant. My CD 100 motor cycle will be made at Haridwar plant, my
scooter pleasure will be made at my Gurugram plant.
So which product to be produced at which facility that is to be decided by again the strategic
team of the organization. Then another important decision is the modes of transportation.
Modes of transportation is another important decision which is again part of strategic
decision. Just taking few minutes about modes of transportation. Mode of transportation may
include lot of different types of modes are available and many a times we may take decision
about multimodal transportation also.
38
So the conventional systems are road, rail, air, water, pipe, these are our conventional
systems, conventional mode of transportation. But many times we may use a combination of
mode of transportation you must have seen many time that road and rail are used together
many times you have seen road and air are used together, many times you have seen that rail
and water has been used together.
So you see that we have combination of different modes of transportation which are used to
make the transportation system more effective and at the same time cost sensitive also. So
what type of mode of transportation you want to use that is also a strategic decision. Then
development of proper information systems for your organization. That is also a very
important strategic decision.
You need to see that the information system should be robust and this information system
should be able to handle the requirement of your supply chain system. So putting lot of
emphasis on the information system because companies like Walmart, the success of these
companies are basically attributed towards their information system. The investment, the
continuous up gradation of their IS which company like Walmart are doing.
That is one of key component in the success of these organization. So putting lot of emphasis
on using the latest information systems, latest network across the supply chain is also very
very important for the success of the supply chain and it is also a type of strategic decision.
So finally we can say that supply design must support the strategic objectives of your
39
organization. So strategic objectives of the organization can be supported only by the supply
chain strategy.
And then supply chain design decisions are the long term decision and it is very expensive, it
is very costly to change these decisions. So normally top management must be very careful in
selecting the supply chain strategies, supply chain structure, and all these things and as we
have discussed that the current scenario is full of uncertainties, it is highly volatile and
therefore it becomes very important for us to design our supply chain in such a way which
can take care of the modern requirement.
So we close this lecture at this point of supply chain strategy, the other 2 phases of supply
chain decision making we will discuss in our lecture number 4. Thank you very much.
40
Lecture-04
Supply Chain Planning
So welcome back friends. So far we have 3 lectures on supply chain analytics and we are
developing a frame work in which we can understand that how the analytics part can help us
in improving our conventional supply chain. And we are also understanding in the course that
the evolution, the challenges of supply chain and what are the important decision making
areas in the supply chain management where data driven analysis, data driven information
can help to improve our decisions.
So in our third lecture we are discussing the different decision phases of the supply chain. In
third lecture we also discussed in detail about the strategic level of decision making. We
discussed that what is the strategic decision and in what many different factors you take those
decisions. We discussed about decisions related to location of the facilities, decision related
to size of those facilities, decisions related to mode of transportation, decision related to
information system.
So we discussed all these things. We also discuss that strategic decisions are long term
decisions. And you have to be very careful in making those long term decisions. Because
changing those long term decisions involve huge amount of cost, if I develop a factory in
41
Uttarakhand, but tomorrow I feel that it is not very economical to have this factory in
Uttarakhand. Now I should have this factory in Sikkim.
So you on your own can understand that this decision is a very expensive decision. Shifting
the manufacturing base from Uttarakhand to Sikkim will involve huge amount of cost and
ultimately this cost will result into poor profitability of your supply chain. So you have to be
very careful, but at the same time we want enough flexibility in our strategic decision making
also.
And for that purpose the role of analytics is going to be very important. Now let us discuss in
this lecture another phase of the supply chain decision making and the second phase which is
also known as tactical phase, a supply chain planning.
So supply chain strategic decisions are at the top, then strategic decisions when we come to a
specific implementation part you have supply chain planning and finally when we move to
execution of these things it is the operational level decisions, so now we are at this level of
supply chain decision phase that is the planning phase. So here you develop a set of policies
that govern your short term operations.
42
As I discussed the scope of strategic decisions is long term. The scope of planning decision is
relatively short, so you develop a long term road map, you have decided that I have to build a
factory at Uttarakhand. Now in how many years you are going to build and in that 4 years of
span, 5 years of span, or 3 years of span, you develop some kind of 6 months annual targets
for yourself. That is the part of your a set of policies for short term operations.
Then this is the planning phase is largely govern by the previous phase. Previous phase is
your strategic management, so the planning phase is largely govern by the phase which is
previous to its. So previous phase that is the strategic phase gives you the broad scope and
within that broad scope you have 5 year planning and from that 5 year planning you draw the
annual budgets. So it is like that you have one umbrella study.
And within that you identify the scope for short term objectives. So the configuration of
planning is determined by the previous phases. And the starting point of the planning is very
interesting now here the role of data comes. For these 2 things I do not find much of the
application of data driven analysis, but now you see the starting point of planning is the
forecast for the coming year.
And the forecast the demand forecast is one very important part of the planning exercise and
here comes the role of data driven analysis, data driven decision making that how are you
doing the forecast. Normally we do forecast based on the previous years data, but now a days
we require more predictive modeling, we require that type of analytics tools which can help
us on the basis of some data which is not has happened so far.
43
But may come in the future and that change in planning approach is only possible with the
help of supply chain analytics. So this forecasting point of which is the starting point of your
planning exercise requires the deep knowledge of data analytics and there are large number of
forecasting methods many of you may already be aware that these forecasting methods like
the very popular is time series analysis.
The time series analysis is all based on my historical data, the previous years data, but now
we will see that how we can do predictive modeling, how can we do this analysis based on
the future data. So that is the important part where data analytics will play every handy tool
for us.
So the type of decisions which we take in the supply chain planning phase. So which market
will be supplied from which location?
44
So we had that picture where you have this kind of network, you have so many retailers and
you have different wholesalers, so you take a decisions that from which wholesaler to which
retailer, these type of decisions are taken in the case of planning. Then what are the inventory
levels. What are the inventory levels, individual facilities are going to have. What will be the
inventory level available with these wholesalers, what will be the inventory level with
individual retailers.
So far, you all must have observed that in our conventional setting retailer are independent to
take decisions about their inventory level. Wholesalers are independent to take decisions
about their inventory level, but when we are talking of supply chain we feel that the different
parties, the different entities, different players involved in supply chain they should not be
independent in these types of decision makings, these types of decision makings which are
affecting the total profitability of the supply chain.
And therefore the decisions related to inventory are not by the individuals rather these are the
planned decisions about the inventory that which stage will hold how much inventory. So that
is again a very important supply chain planning decision we take and this decision of planned
build up inventory is based on real time data which we have already discussed that what is
the meaning of real time data and how it flows into the supply chain.
So that is a planning decision. Then subcontracting, back up locations etc. Now a days we all
know that all these activities of manufacturing, transportation, warehousing, retailing etc. the
time of hold is no longer there. You require different partners in this process of supply chain.
45
From left to right you need so many different partners and when I am talking of so many
different partners you many a times subcontract your various activities.
Sometime you may subcontract your manufacturing activities, sometime you can subcontract
your transportation activities, sometime you can subcontract your warehousing activities etc.,
depending upon the nature, depending upon the abilities of your subcontractors etc. So
subcontracting and back up location, how much to subcontract, to whom to subcontract and
what type of products to be subcontract, all these things are very important and decisions
about the subcontracting activities is also very important planning activity.
Then inventory policy we have already discussed. Then timing and size of market promotion,
that is also important. You see in supply chain many of us may argue that this is more related
to marketing and less to supply chain. To some extent it is correct also, that timing and size of
market promotion is a more marketing decision, but your supply chain must ensure that yes
when there is a promotion the product is available in the market.
The point which I am trying to say that if Apple says that on 7 th September 2016 iphone7
will be launched across the globe. Iphone7 will be launched across the globe on 7 th
September, so now when this campaign is starting to popularize the launch of iphone7. My
supply chain must ensure that iphone7 is available secretly, securely to all my global retailers.
So that is a very important thing.
And like in India we have normally the movie releases on every Friday. Now every Friday
when a new movie starts or released you need to ensure a supply chain that movie must reach
to the respective cinema halls, to the respective theatres wherever it has to be screened. And
you need to ensure enough safety, security, that there should not be any kind of theft of the
CDs, or DVDs, or the prints of the movie.
And at right time it should be screened across all the theaters. So these are some of the points
which are related to timing and size of market promotion. So you need to do promotion only
in those areas where your supply chain can ensure timely availability of those products, and
what time the product will be available accordingly you start the promotion. So timing and
size of promotion should be very much in sink with your supply chain planning.
46
If supply chain planning and timing and size of promotion are not in line with each other you
see the recent case of reliance Jio also. So when Jio is launched when Jio launch was
announced so their supply chain ensures that the Jio sim card, Jio mobiles are available to all
their distributors. So it is though on prima facie timing and size of market promotion looks
more marketing area.
But when we go deep into it, when we try to understand the cracks of this statement we will
find that supply chain is very closely associated with this timing and size of market
promotion. Then at the planning stage though at the strategic level also but at the planning
level also the demand and certainty, we already discussed this in the third lecture, the
exchange rates when we are talking of global markets, the exchange rates, and therefore the
third point which we discussed in the third lecture that is the fluctuations in the cost.
So that exchange rate, changes in the exchange rate, the fluctuations in the exchange rate may
also result into the fluctuations of the cost of supply chain. Then competition over the time
horizon. The competition is also changing over a period of time. Once upon a time for
American companies, American automobile companies, American electronic companies,
Japanese companies used to be the big competitors.
But if you see today's scenario, the scenario of 21 st centaury, Chinese companies are big
competitors for American as well as to Japanese companies also. You go to Japan
everywhere you will find made in china product. You go to America, you go to Walmart
there everywhere you will find made in china products. You come to India you are very
fortunate that you still find lot of made in India products.
But in America and Japan both these countries are facing very tough competition from
Chinese companies. So competition is changing over a period of time. So your supply chain
planning should also keep an eye on the changing competition over a period of time.
Competition is more related to the strategic decision, but since strategic decision as I
discussed earlier are long term decisions and competition is changing very fast.
Competition is changing very fast and the new players, new products, product lifecycles, all
is getting reduced day by day and therefore the issue of competition, how to handle
47
completion. This type of questions are coming more in to picture in our planning phase of the
supply chain.
So strategic, you can take only once in a while decision with respect to competition, but daily
how to change your supply chain planning for that purpose we need to take care of
competition at the planning stage also. So these are the important decisions which we make in
the planning phase of the supply chain. Then we come to the third level of decision phase that
is the operational decisions which we do in the supply chain, so these operational decisions
are related to execution of the supply chain.
48
So now we are finally coming into the implementation stage of supply chain. And here as is
mentioned the time horizon is very small. So the time horizon is very small, it is long term in
case of your strategy, it is relatively small in case of planning and it is very small and it is as
small as you can see weekly or even daily in some cases also. So if you are talking of let say
food supply chain.
If you are talking of food supply chain so the time horizon may be daily. If I am talking of
apparel supply chain the time horizon may be weekly, so we can go up to that level of
execution monitoring that time horizon can be weekly or daily. You take the example of 7, 11
Japan, in that case the supply chain operation time horizon may be few hours also. Because
they replenish their inventory 3 times a day and when they are replenishing their inventory 3
times a day.
So in fact the time horizon for companies like 7, 11 Japan is less than a day, it is just 4 to 5
hours. So here actually the time horizon is really very small. Then here we take decisions
regarding individual customer orders. Now at this level we started taking care of individual
customers. That what are the requirements of that particular customer, what are the quantities
required, where it is required.
So all those things with respect to individual customer you are taking care at this phase. And
here also the past data may certainly help us to take better decision with respect to customer
satisfaction, with respect to customer experience, with respect to customer experience to our
supply chain decisions and so that you can make your individual customers more happy. Here
our supply chain configuration.
And this decision making is govern by the planning phases. So this is a kind of hierarchical
arrangement planning activities are fit by the strategic nature and operational are fixed by the
planning phase of the supply chain. The goal is to implement the operating policies as
effective as possible. So whatever you have decided, whatever quantitative as well as
qualitative goals you have for your organization.
You try to implement at the operational level with the maximum effectiveness. Then
allocation, orders to inventory, or productions at order due dates, generate pick up list from a
warehouse, allocate, order to a particular shipment, set delivery schedules, place
49
replenishment orders. All these are the various types of supply chain operational issues which
we handle in the operation.
So day to day, if you talk of a supply chain officer or a supply chain manager, so on a day to
day basis, week to week basis, hour to hour basis, these are the important jobs. These are the
important activities which he or she performs in an organization. So that raising the intents,
making the challans, making the invoices, delivering the products to a shipment etc. So all
these things are the part of our supply chain operation.
And in a good organization you have some standard operating positions for all these things
that how do we have inventory replenishment policies, how do we have the delivery schedule
challan and all these thing. And here one interesting thing is that in operation level, you can
work whether less amount of flexibility because the time horizon is very small.
Here the time horizon is very small, time horizon is few hours or day or a week, so things
may not change so fast within a day and therefore, the advantage is there that you do not
require much of flexibility and on the other way you can also write that less uncertainty is
there because of short time horizon. So that is a very positive sign for supply chain operation
and therefore the skills required, therefore the kind of exposure required at the operational
level is much less than what we require at the planning at the strategic level.
Because at the strategic and planning level you require a bigger vision and you need to handle
bigger challenges. And because of the operational requirement because of your day to day
affair involvement it is not expected that you also focus on the external issues. So the focus of
strategic and planning these 2 level is equally on internal as well as external environment of
your supply chain.
While operational people they are more concern with the day to day, with the internal
environment of the organization. So that is how you can also understand the difference in a
scope of work of different phases of supply chain.
50
So now once we are clear with the different phases of supply chain decision making. Now the
next part of our discussion which we will continue to the next 2 lectures also. That is about
different ways in which you can study a supply chain. There are 2 very popular ways which
are available in literature which can help us to study the supply chain and these are one is
cycle view and another is push pull view.
Cycle view that there are different stages in the supply chain we have discussed in our all the
classes earlier that you have wholesaler, you have distributors, you have retailers, you have
manufacturers, you have customer, so there are different stages and the processes happening
between 2 stages.
51
This is manufacturer, wholesaler, retailer, customer. So now the processes happening at any 2
stage at manufacturer or wholesaler, wholesaler or retailer, retailer or customer can be seen in
this type of cyclic view.
So you have supplier, supplying products to manufacturer, manufacturer giving products to

distributor, distributor supplying products to retailer, and at retailer finally customer
purchases the product. Now you see this cyclic view conceptualizes a cycle at each of these
interfaces. When customer is coming to retailer this cycle is customer order cycle. When
retailer is giving order to distributor or otherwise also you can say that when distributor is
supplying products to retailer, there is a replenishment cycle.
The meaning is distributor is replenishing the stocks of retailer, fulfilling the requirement of
retailer. When manufacturer is supplying products to distributor this is the manufacturing
cycle. Here manufacturer is making the products, producing the products and supplying those
products to the distributor. And when supplier is supplying raw material, components, sub
assemblies part to the manufacturer this is the procurement cycle, that all these components
etc. are procured by the manufacturer.
So this way, there are different types of cyclic activities, which are happening, now when we
go for microanalysis, when we go for microanalysis of all these cycles you will find that in
each of the cycle you have almost similar type of processes. Though here you have different
names of the cycles, customer order cycle, replenishment cycle, manufacturing cycle or
52
procurement cycle. But actually the cycles are consist of various processes also. And we will
see that all these cycles are consisting of similar type of processes. So that is very interesting.
This each cycle as we mention that each cycle occurs at the interface of 2 successive stages.
And these types of cycles we just discuss. And cycle view clearly defines processes involved
and the owners of each process. It specifies the roles and responsibilities of each member,
these are the different members, manufacturer, wholesaler, retailer, customer etc. So within
each cycle who are responsible for what type of activity.
So that is what is specified in the cyclic view and therefore cyclic view is very important for
the proper distribution, proper understanding of roles and responsibilities of different
members of this entire supply chain. The cyclic view is one way to look for the supply chain
development. But there is one more way that is push and pull view and this will help us.
This push and pull view will help us in developing the concept that what are the things in a
supply chain can be done in anticipation and what are the things can be done in reaction in a
supply chain. So this push and pull view is also very important for understanding the
philosophy of supply chain. So we will continue with this push and pull view in our next
lecture and thank you very much till then, let us see.
53
Lecture-05
Different Views of Supply Chain
So Dear friend, I welcome you to another session of supply chain analytics. So far we have
discussed the basics of supply chain, the need of supply chain management in the changing
business scenario, then we also discussed how analytics can help us in some of the challenges
which current supply chain management is facing and then in last 2 sessions we are
discussing that what are the philosophies which are governing the supply chain management.
And in that school of thought we have already discussed one school of thought which is
represented as cyclic view of supply chain where the processes happening at each stage can
be represented by the cyclic view or some kind of cycle.
Now today we will discuss, we will start the another very important concept of supply chain
which is known as push, pull view of the supply chain. Now here in the supply chain you see
some of the part is represented in green and some of the part is represented in the red. In a
complete supply chain starting from the supplier source to the customer end, you have some
of the processes which are known as push processes.
54
And some of the processes which are known as pull processes. As the name represents, pull
processes are those processes which are pulled. And customer is at this end pulling products
from the supply chain. So these are the pull processes. On the other hand in case of push
processes, the products are pushed to the supply chain. So demand is there or not there.
These products are forced to the supply chain and pull processes as a name indicates are there
because demand is there, a customer wants these products and customer wants these
products, that therefore customer is pulling these products from the supply chain. So now the
take of this push and pull for a supply chain manager is that processes which are happening
as a result of pull processes are all reactive.
Because demand is there, therefore supply chain is trying to fulfil the demand. So these
processes are reactive in nature. Push processes because we feel that there will be demand,
there will be demand, so these are in anticipation of demand. Demand has not occurred so far,
but these are in anticipation of demand and these processes are more proactive in nature.
These are proactive in nature, you feel that demand is going to be there in the future.
And therefore you are pushing products into the supply chain, that there will be a customer
who will come and he will buy these product. So therefore I should maintain the sufficient
inventory, sufficient stocks of these products. So in that anticipation push processes are
taking place. On the other hand pull processes, you go to a restaurant and when you go to a
restaurant the food will be ready when you give a order.
So, preparation of food and serving the food to a customer is an good example of pull
processes. While you go to a grocery store, you find so many chips, so many biscuits, so
many other kind of food items are readily available. So these are push processes, that shop
owner feels that customer will come to me and in anticipation of those demand, he is stocking
the products.
But normally you will find that in all supply chains, in all supply chains there are few
processes which are done in anticipation. So these are push processes and there are few
processes which are done was the order has come. So these are pull processes. Now for a
manager for a decision maker with the help of the data, with the help of our supply chain
55
analytics the very important job with respect to push and pull is to identify the correct
boundary of these processes.
Where should we keep our boundary of push processes and pull processes which is that point
in the supply chain. And probably this is one of the most crucial decisions a manager has to
make and now a days, with the help of data driven analysis with the help of data driven
decision making processes, it is very important that you should have a dynamic type of push/
pull boundary.
We in the old age system in a supply chain we always use to have a static push/pull
boundary, but now a days because we are more into the real time data driven decision
making, so it is quite possible that the boundary of push and pull processes may shift. So, we
will see with the help of an example that how this boundary of push and pull processes may
shift with the help of new data which may be available to us over a period of time.
Let us see a generic supply chain where we have manufacturer, distributor, retailer and
customer. Now the processes up to retailer are push processes. And only at the retailer end
the processes are of the pull nature. When customer reaches the retailer, so retailer will give
that product to the customer. So only this thing is happening as the pull process and rest all
these stages are bound by the push processes.
When things are happening with push processes all these places you will have very high level
of inventory. Because in anticipation you are stocking the products. It is quite possible that
56
over the period of time when you have more data available with you, you may shift the
boundary of pull process to the distributor level. And only up to the distributor level
processes may take place as push.
And when it is happening, when you are shifting the boundary of push and pull from retailer
to the distributor and it means now retailer is not keeping the inventory in anticipation.
Retailer will only be a source of contact to the customer and whenever order comes,
whenever a customer walks to the retailer's place, retailer will further pass this information to
the distributor and the distributor will supply the products to the customer.
The example of Dell, when Dell was directly distributing products to the customer before
2006, so Dell was following this type of model where the push/pull boundary was a dealers
there, distributors there, but now in the present system when Dell is following the retail
distribution in that particular case the boundary of push/pull is shifting to this case.
So you can see the free 2006 and this is present, so in the case of Dell the boundary of push
and pull has moved from the distributor to retailer from 2006 to now. So you have to be and
this is by the way based on all real time data analysis. So this is a very good example that
how your data analytics may help you to shift the boundaries of your push and pull processes
and you cannot be static all the time.
In present time scenario, where you have lot of competition, where you have so much into the
market phase that everyday things are changing.
57
So push/pull view as I told you, pull view is reactive and push view in anticipation it
speculative and the boundary of push and pull separates the push processes from the pull
processes. So we have seen the role of boundary that before this and after this the processes
are in anticipation and after this the processes are in reaction to the order of the customer.
So push/pull processes because another important part of the push and pull processes as we
have just discussed here and it is further written in these slide that it also gives you an idea of
who is going to share the maximum risk in your entire supply chain. You can understand very
well, those processes which are happening in reaction to the order. The risks in those stages
are less.
58
But those stages which are performing processes in anticipation or those stages which are
having the push phenomena, the risk is high in those stages. So now it is a very important
strategic decision in entire supply chain that which stage is going to have how much risk and
many times you will see that manufacturer needs to keep maximum risk with itself.
Sometime in some supply chains very few supply chain you will find where distributors are
keeping the maximum risk, but there will be very few supply chain, very few supply chains
like Walmart or something like that where you will find that retailer is keeping the maximum
risk. But in most of the cases it is the manufacturer which is keeping the maximum risk.
Because manufacturer has to prepare everything in anticipation of the order.
But there are some very specific examples particularly those manufacturers dealing in the
industrial products like BHEL making turbines, making transformers, these types of
organization, these types of manufactures, they even do not keep inventory of those products.
This companies is start everything in the pull fashion.
The example of BHEL, you will find that right from the beginning right from the
manufacturers end everything is happening in pull fashion. And very little is happening in
anticipation. So some type of raw material procurements, some type of spare procurements,
only those things may happen in push form. But rest of the things are happening entirely in
the pull form.
59
So those products which are industrial marketed or where the other word can be the unit
value is very high. So those products where the unit value is very high you normally find the
entire processes are happening in the pull form. And but still in this particular case also the
procurement of raw material, procurement of the spare parts and some of the components
may happen in anticipation.
So a little bit element or push will also be there here you see you cannot have the entire thing
starting when the order comes to you. You cannot start going to a mine to get the iron ore for
making a tribune only after the order comes. So you procure the steels, you procure the raw
materials, you procure different types of alloys in anticipation that orders will come. And we
need to have these materials ready with us whenever order comes.
So you need to have a fine though, these are some of the guiding principles, but still
depending up on your own data, depending upon your own information you will find a
suitable boundary of push and pull in your supply chain. So that is very important that where
is the boundary of push and pull in your supply chain, because this as I mention that relative
proportion of push and pull can have an impact on supply chain performance.
So it is very important if this boundary is not properly selected it may result in excessive
inventory or it may result in a stock outs. So our data, our analytics will help us to
appropriately select the boundary that will give us better supply chain performance.
60
Now another point which we will like to discuss in this session, that is about the competitive
and supply chain strategies. You see supply chain strategy should always support. Supply
chain strategy is a type of functional strategy and in an organization there are variety of
functional strategies. So all those functional strategies should support the competitive strategy
of the organization. Now what is the competitive strategy?. The competitive strategy is the set
of customers need a firm seeks to satisfy through its products and services.
Now the meaning is how differently, how differently a firm can fulfil the requirement of its
customer over its competitors. I require some product for my transportation requirements.
Now there are A, B, C, D, E, 5 different marketers offering different type of transportation
solutions. Now what is that USP of A, what is that USP of B, what is that USP of C, that
defines my competitive strategy.
So each one of us each marketer, each corporate, they look to provide some kind of different
solutions, some kind of unique solution, the unique way to fulfil the requirement of the
customer's need. Walmart is the very popular example and in our supply chain class we use
this example so often. The competitive strategy of Walmart is everyday low pricing,
everyday low pricing that is the competitive strategy of Walmart means Walmart is famous.
Walmart provides that you can get large number of products under a single roof with the
lowest possible price on a particular day. So that is the competitive strategy of the Walmart.
So now the issue is that how my supply chain supports this competitive strategy and then
only the success of organization comes. So here we see that there are different types of
functional strategies you have product development strategy, you have marketing and sales
strategy.
These are the functional strategy. And then there is a supply chain strategy. Supply chain
strategy should support by competitive strategy. It should be in sync with my competitive
strategy, it should support my competitive strategy, so that I can achieve what I want to. Now
supply chain strategies, it determines the nature of material procurement, transportation
materials, manufacturing of products, creation of services, distribution.
And it is also how consistency and support between supply chain strategy, competitive
strategy and other functional strategy. Earlier we use to have different silos in organization.
61
Silos means each one of us, if I am in the marketing department I normally uses no
communication with my product development department. If I am in product development
department I used to have no communication with my production department.
If I am in the production department I use to have no communication with my HR

department. So we use to live silos earlier. But nowadays it is not possible to live in silos. We
need to have a very strong integration of all functional strategy.
Therefore, this value chain represents that right from the new product development to
marketing sales to operations, to distribution, to services. All these things, all these functional
areas need to be integrated, need to be supportive of each other, and finance, accounting, IT,
human resource etc. need to work as supporting to these major functional areas. So this is the
value chain which links the supply chain with the business strategy of the organization.
And therefore we are very much concern that what type of supply chain strategy we adopt. So
that the competitive strategy of my organization can be achieved.
62
And for that purpose we further talk of this strategic fit where we should have a consistency
between customer priorities of competitive strategy and our supply chain capabilities
specified by the supply chain strategy. Now it is very important to understand that customer
priorities which is determining my competitive strategy. And supply chain capabilities these
things are determining my supply chain strategy.
So my customer priorities, my ability to understand my customer priorities that is the first

important role which analytics will play. The data will help me to understand the priorities of
my customer. And it is very simple to understand that over a period of time customer
priorities are changing. Customer priorities are no longer constant, I am living in India.
But in India also the priorities which we used to have about 20 years back or 30 years back,
these are no longer the priority at the moment. Our priorities are totally changing, so
priorities do change as we get more exposure as we have better purchasing power, as we have
access to latest technology particularly IT, particularly internet, and so on. So all these things
are helping me to change the customer priorities.
Now as a marketer, as a supply chain decision maker, if I am not able to capture these
changing priorities of my customer I will not be able to evolve dynamically my competitive
strategy. So the first important role of our supply chain analytics, the data analytics part into
the supply chain is to identify the changing the customer priorities, how things are changing.
63
You see when Japanese movement of quality started at that time the customer priority shifted
towards getting the quality product and throughout the world not only in India customers
started expecting to get Japanese qualities in their product. But over a period of time almost
all the manufactures throughout the world they had same technology, same management
principles, and they were able to deliver almost competitive quality levels.
Then came the Chinese envision of low cost and now a days customer expects that the quality
of the product will be as good as Japanese product, but we need price as low as Chinese
products. So priority all of a sudden is changing. And then priorities do change with respect
to the different market segments. So as you go from one market segment to another market
segment may be on any variable you will find that priorities are changing.
Sometime I am purchasing a product only for my status and sometime I purchase a product
for its functional utilities, so dynamically I need to see what are the customer priorities, so
this is very important to understand that these are the customer priorities which I am getting
with the help of my data support. The second is supply chain capabilities, what my supply
chain capabilities are, what it can do and you have the example of Maruti in India.
Maruti in India is very good example that how Maruti reached to tier 1 then to tier 2 and now
it is reaching tier 3 cities also of the country. And therefore the worst infrastructure the large
number of facilities created by Maruti is helping to a great extent to achieve the supply chain
strategy of the marketing. But there are many companies in the country, they have a supply
chain strategy in their vision, but the supply chain capabilities are not developed accordingly
and therefore you are not able to implement, you are not able to execute that supply chain
strategy.
64
So first thing is we want to develop a strategic fit between business strategy or you can say
competitive strategy which supply chain strategy must support and the supply chain strategy
the exhibit of the strategy comes in the form of capabilities. Capabilities which we develop
these are the manifestation of supply chain strategy. How do we see that this is the supply
chain strategy.
So the capabilities which were developing, these are actually the manifestation of the supply
chain strategy and this manifestation must match with the competitive priorities or customer
needs which are continuously. So changing customer priorities are to be captured with the
help of data, so this is the first use of data analytics. That data analytics will help us in
regularly monitoring the competitive priorities and you need a very good strategic fit between
these 2 things.
Then just to simplify the same statement we have this statement that competitive and supply
chain strategies must have the same goal. It means many times it is possible that people may
not understand the holistic nature of the organization and when they do not understand the
holistic nature people at their own departmental level, at their own functional level may
develop the departmental and functional goals.
So we are warning against these things, that the overall goal of the organization which is the
business goal or the competitive goal and the supply chain goal must have same objective,
must be same thing. And then we further warn that company may feel because of a lack of
strategic fit or because it processes and resources do not provide the capabilities to execute
65
the desired strategy. So we need to see that what type of strategy we are developing and how
the resources and processes support this strategy.
Just to give you an example because it is becoming very theoretical, so how our resources can
provide the support to the competitive strategy. For that purpose if I am taking a very market
intensive strategy, my supply chain strategy is market intensive. Now when I am saying that I
want to develop a market intensive strategy, supply chain strategy the meaning is that I want
to open large number of facilities in almost all the corners of my market.
So when I am talking of market intensive strategies or the market dominants strategy, I

should have large number of facilities. But then in a city like Delhi I am opening only 2
retailers, in a city like Mumbai I am opening 3 retailers, so this does not qualify to be the
market dominant strategy. When I am saying market dominant strategy in a city like Delhi
there should be more than 50 retailers.
In a city like Bombay there should be more than 30 retailers, wherever I go in Bombay,
wherever I go in Delhi, wherever I go in Kolkata I should have every corner with my retailer
and Patanjali is a very good recent example that how Patanjali is following is market
dominant strategy. And for that purpose wherever you go in even a small city every street,
every market you will find distributors, retailers of Patanjali.
But they say that we are going to be the market dominant company and in whole other city
you have only 1 distributor, one retailer than probably your supply chain capabilities are not
matching with your supply chain strategy. So whatever supply chain strategy you have you
need to develop the resources, you need to develop the capabilities in line of that and this
strategy that we should follow market dominant strategy, it should come from your
competitive strategy.
That my customer prefers those products which are closely available, which are nearly
available, they do not want to put lot of efforts in buying those products and if that is so I
should go for market dominance strategy and when I go for market dominance strategy then I
should have large number of stores, large number of retail outlets in a particular area. So all
these things must be in sync.
66
If it is there then my supply chain strategy is in line with my competitive strategy, but I say
that my customer follows the convenience way of purchasing, but even in that case my
supply chain is following the selective way of distribution. So my supply chain strategy is not
in line with my competitive strategy. So the very first thing we should understand the
competitive priority that what are the competitive priorities and according to those
competitive priorities.
I must design my supply chain strategy and according to supply chain strategy I should
design the, I should develop my supply chain capability. So data analytics is the part which
will help us in dynamically understanding the competitive priorities and accordingly you can
take all these decisions. The way the very good example of use of data analytics in
developing the supply chain capability is almost all petroleum companies in our country.
Earlier we use to have very limited number of retail outlets of these petroleum companies to
which we call as petrol pumps. But now over a period of time with the help of available data,
these companies realized that number of automobiles, number of cars, number of 2 wheelers
in the country is increasing exponentially and therefore the requirement of petroleum
products are going to increase and then they realize that we should go for the market
dominance strategy.
And for that purpose now you see almost in each street of the city highways are filled with
the petrol pumps of different of companies. So the earlier very selective strategy was there,
but now it is totally changed into the market dominance strategy. So this way you can
understand that how data analytics can help us in providing the required strategic fit between
the competitive strategy and the supply chain strategy of an organization. Now moving
further into this discussion of supply chain strategy and the competitive strategy.
67
We will see that it is a 3 step process. The first process is you need to understand the
customer and supply chain uncertainties. So as I am saying that your customer is
continuously changing its priorities and so on the capability of your supply chain also has lot
of uncertainties. Because in your supply chain there are supplier, vendors, different type of
other members are there and they all have some component of uncertainty associated with
that.
So very first thing is to understand the customer and supply chain uncertainty. And to a large
extent data analytics may help us in resolving the issue of these uncertainties. We can have
good modeling software we will discuss and those things will help us in removing the
customer and supply chain uncertainties. Then the second thing is to understand your own
supply chain.
The capabilities, these capabilities, what are these capabilities, so that is the second part. And
third part you need to link these supply chain capabilities with the customer and supply chain
uncertainty. Whatever uncertainties you resolved, you should try to fulfil those things with
the help of your supply chain capability, if you can do that it means you have achieve the
strategic fit.
So here today we close about this concept of strategic fit in our next session we will see with
the help of some kind of modelling exercise that how these strategic fit is achieved and that
will give us with help of some more practical examples the usefulness of the data analytics in
achieving the supply chain strategic fit with your competitive strategy. Thank you very much.
68
Lecture-06
Supply Chain Strategy
Welcome back. In our last session we discussed about different views of supply chain
management and we discussed the cyclic view and the push/pull view. And in push/pull view
we discuss the importance of boundary of push/pull and as a supply chain manager the data
analytics will help us to decide dynamically where to keep the boundary of push processes
and pull processes in a supply chain environment.
Now the supply chain as we discussed in our last lecture must fulfil must be in sync with the
competitive strategy of the organization. And we discussed a 3 step processor for achieving
that strategic fit between the competitive strategy and the supply chain strategy. We also
discussed that there are different types of functional strategies in a value chain right from the
new product development to the after sale services there are various value addition
component and human resource, information technology, finance, accounting.
These are supporting functions which help in achieving the objective of new product
development, marketing, distribution, logistics and after sale supports. So all these functional
strategies must be in sync, must be in line with each other and that is the need of current
business environment. If these strategies are not in line with each other so probably the
competitive strategy of the organization cannot be achieved.
Now particularly in this course we are focusing more on the supply chain strategy that what is
the supply chain strategy and how we should develop with the help of data, the robust supply
chain strategy which can be fulfilling the supply chain capabilities and those supply chain
capabilities are in line with the customer requirements and we yesterday in our last lecture
discussed that those customer requirements are dynamic.
And data analytics may help us in understanding in continuously monitoring those changing
requirements. So now let us discuss that how the various components of supply chain strategy
or yesterday we discussed that supply chain strategy is manifested, is exhibited in the form of
69
various capabilities and let us discuss what are those capabilities which a supply chain should
develop.
And with the help of data analytics, with the help of real time data analysis we will see that
how these capabilities will help us over a period of time to develop a robust supply chain
strategy helping in achieving the competitive strategy of the organization.
So very first capability a supply chain develops that is in the form of facilities. You need to
develop various facilities in a supply chain and these facilities may be of different types, there
may be different roles of these facilities, so what are the role of the facilities, what are the
sizes of these facilities, what are the locations of these facilities. These are some of the very
important questions.
We like to answer in this particular aspect of supply chain capabilities that whether my
supply chain facility is a manufacturing facility, it can be a warehouse, it can be a distributor,
it can be a retailer, so these are different types of roles, my facilities may do and I need to see
what should be the number of these facilities. How many manufacturing locations I require.
You have example like company Maruti, company of Tata motors, company like Hyundai
motors.
They have primarily one big manufacturing facility. On the other hand if you talk of
petrochemicals or you talk of some other products which are having multiple manufacturing
facilities. You have one big refinery, but then you have so many distribution points, so many
70
outlets through which you can refill your LPG cylinder, so you need to see what are the
different types of role you expect from your various facilities in your supply chain.
There are centralized manufacturing, there are examples of decentralized manufacturing. So

where the unit cost of product the rule is where the unit cost of the product is very high. You
will like to go with centralized manufacturing facility. So this is based on data. And when the
unit cost of the product is low you can go with decentralized facility of manufacturing. You
can have multiple manufacturing facilities in different markets.
So now it depends up on your target, your type of product, that what type of product you
have, and accordingly you will decide the number of manufacturing facility. Again when you
want to achieve high economies of scale, you will like to have bigger sizes of your here you
get high economy of scale, if you have a bigger size of manufacturing facility you get higher
economy of a scale, the meaning is the cost of production per unit decreases accordingly.
And if you are not interested and obviously when I am talking of high economy of scale size
should be bigger and it is more like centralized manufacturing facility. So when you have
centralized manufacturing facility you can achieve higher economy of scale and higher
economy of scale means your supply chain is efficient one. So you see how facility can help
you on moving from efficiency frontier to responsive frontier.
If you are making a bigger facility, if you are making a centralized facility you are on the left
side of this spectrum. And if you are making more facilities, if you are having a decentralized
manufacturing system and a smaller facilities of manufacturing you are moving towards the
responsiveness side. So facilities can help us in deciding the type of supply chain strategy you
want to have the type of capabilities you want to have in your supply chain.
So this is about the manufacturing and high economy of scale or low economy of scale. Then
same applies for the warehouses, same applies for the distributors, same apply for the
retailers, how many wholesalers you want, how many distributors you want, how many
retailers you want, what should be the size of these wholesalers, what should be the size of
these distributors, what should be the size of these retailers.
71
All these are questions we want to know in this particular case of facilities. Then another
important question we like to have the answer for that is the location of these points. Where
should you locate your warehouse, where should you locate your distribution aspect, where
should you locate your retailers, do you want to have retailer in most expensive location of
your city, or you want to have retailer at a low cost location in your city.
So depending upon your efficiency and responsiveness, if I want to locate a retailer let say in
Connaught place in New Delhi or south extension in New Delhi probably my supply chain is
not an efficient supply chain. But if I am locating a retailer in some of the low cost locations
may be east Delhi in Delhi. In that case it is towards the efficiency side of the supply chain.
So the location of the wholesaler, retailer, distributor, manufacturing facility will give you an
idea whether you are on the efficiency side or on the responsiveness side.
Many of the new factories, many of the new organizations are opening their manufacturing
facilities in some of the newly form states in India, to avail the benefits of taxes, to avail
different types of incentives, given by state governments given by the central government to
promote industrialization in those states. And when you are moving to those places this is an
example of efficiency. While it is more convenient, it is more responsive that you locate your
manufacturing facility in NCR.
You locate your manufacturing facility near Bombay, near Ahmedabad, which are already
very popular manufacturing locations. But the cost of land, the cost of labour, and cost of
other resources required for manufacturing are already very high on those places. So though
you will get high responsiveness but cost will be very high. So many manufacturing facilities
are located in these new states to take the advantage of efficiency in your supply chain.
So all these things that when can I get the tax benefit, where I can get the benefit of income
tax, where I can get the benefit of excise, where I can get lower cost of labour, all these things
are helping us in a dynamic environment for changing the location of even my manufacturing
which is a costly affair, which is a long term affair, but still many of my friends in this class
may be aware that.
So far we all used to consider China as a location for low cost manufacturing, so all major
manufacturing players of the world, they are looking to locate their manufacturing facilities
72
in China because Chinese are known to be the very low cost manufacturing locations. But if
you read the recent reports you will find that the benefit of China of low cost manufacturing
is now no longer existing because of higher cost of wages because of higher labour rates etc.
And now that benefit is going to even southeast Asian countries like Cambodia, like Vietnam,
and all those places, smaller places. So the benefit is shifting to Vietnam and Cambodia type
of country. So and new manufacturing facilities are going to those countries particularly
Vietnam if I say. So you can take decision related to your manufacturing facility on
continuous watch on real time watch of your data that in what type of scenario.
Where I can get the maximum benefit and accordingly you take the decision to locate your
manufacturing facility. So the first issue is about the facility that how do I handle this facility
issue, so that I can get the best of either efficiency and responsiveness. So my real time data
not exactly on day to day basis, but may be at some interval I will take decisions about my
facilities and continuously working on distributors.
I am continuously working on my retailers, so where to locate my new retailers, where to

locate my new distributors, what should be the size of these retailers and distributors, all
these things are very important decisions for a growing business and therefore data analytics
will help me in determining the locations of distributors and retailers and on a longterm basis
the location and sizes of manufactures and warehouses can also be determine with the help of
available data.
73
The second important decision after facilities in the supply chain and where data analytics is
going to help us maximum that is inventory. Many decisions related to inventory are very
crucial for the supply chain. Now whenever you have a supply chain, you are keeping
inventory of various products in this supply chain. You keep inventory at retailers end, you
keep inventory at wholesaler, you keep inventory at manufacturer.
Now the inventory level, what should be the optimum inventory level. So the level of
inventory is a very crucial decision because the Japanese system of management which is
very popular now, it says that inventory is not a value addition activity in the organization. So
that system is probably you can say is anti-inventory, we should not keep inventory because
inventory is additional cost to the system.
So in some of the places in the world where things are very close to each other you can work
with systems like JIT, where we are not keeping any inventory and things are readily
available to us in real time. But in country like India or in many other countries where one
party is geographically distant to 100 kilometer or more than that, so we need to keep
inventory for taking care of lot of uncertainties in the environment.
So inventory is a necessity you can say in case of country like India. But what should be the
optimum level of inventory. Because whenever you are keeping inventory there are 2 types
of very important cost which we are incurring. These costs are the ordering cost, and the
second is carrying cost. You can say the ordering cost is also known as depending upon
situation to situation.
The ordering cost can be loading, unloading cost, if a particular packet is coming from
wholesaler to retailer, so the loading at the wholesaler end, unloading at the retailer end, those
types of cost are the part of the ordering cost. The carrying cost or the holding cost when you
are keeping a material with you, you have blocked some of your capital in keeping that
material.
And when have blocked the material, block the capital in keeping that material, so the
opportunity cost, the benefits with that material with that amount of money you are deprived
of, so that is the carrying cost. And in other books also you can find that the cost of space
74
involved, the cost of safety involved, the cost of wear and tear involved, all these are the part
of carrying cost. So carrying cost is a very important cost.
And there are lot of businesses because of improper inventory management and the carrying
cost is increasing exponentially, business is gone down. So the carrying cost you can say is
such type of evil, it is a disease you can say which eats away all your profit, you have a very
good supply chain, you have huge demand in your supply chain, but if inventory is not
properly managed and if this carrying cost is high this will eat.
This will take away all your profits. So for a supply chain manager and particularly now
when we have tools like analytics available with us we need to pay very sincere efforts for
appropriate management of inventory, and for that purpose we will see that how analytics can
help us, the management of inventory requires that we need to adjust ourselves between 2
things.
That we want to have a good customer satisfaction also, so for that purpose we want to
achieve a good service level. So we want to have high service level and that is required for a
better customer satisfaction. And at the same time we want to minimize the cost of inventory.
So we need a balance, we need a trade off between service level and cost of inventory. If you
are increasing the service level for improving the customer satisfaction.
So in that case cost of inventory will also be increasing, that is not desirable. We should keep
cost of inventory low, and when we focus primarily on cost of inventory that it should be low
75
resulting will be the service level will also go down and that will lead to poor customer
satisfaction, that is also not desirable. So we need to make a trade off between these 2 things
that there has to be some kind of balance between service level and cost of inventory.
And it is very important for a supply chain manager that you continuously monitor, that you,
what is the expected service level for my product, And it is not a constant thing, and therefore
real time data analytics will help us to continuously change the service levels. Today I am
operating at 80% service level, let say with my own assumption I start a business and I am
maintaining 80% service level.
The meaning of 80% service level is that out of 100 times if a customer visits to my retail
outlet 80 times, 80 number of times that customer is able to fulfil his or her requirement. I
have readily available products with me and 80 number of times, 80 times I am able to fulfil
his or her requirements, 20 times I am not able to fulfill the requirement of that customer.
So that is the meaning of 80% service level. Now that 80% service level gives me some type
of cost of inventory. But my data analytics team, my marketing team capturing the data on
customer satisfaction may tell this data to be that you are operating probably on a higher
customer, higher service level. Your competitors are operating at 75% of service level and
that gives me a margin of 5% service level.
That I can reduce my service level from 80% to 76% or 75.5% and this will directly translate,
this will directly be reflected into my cost of inventory. My cost of inventory will accordingly
go down and when my cost of inventory will go down this will accordingly increase my
profits. So you can say that real time data analytics with respect to your service level, with
respect to your inventory management is very important.
And very crucial and it is directly related to the profitability of my entire supply chain. So
therefore the analytics will help me in developing the capabilities related to inventory that
what should be the optimum level of inventory in my supply chain and that optimum level of
supplier inventory will help me to get the optimum level of cost of inventory. So this is the
second important decision area where analytics will help. The first is facility, second is
inventory.
76
Now let us move to the third important area of our discussion and that will be more like your
logistics, that third area is transportation. Transportation and many of us say transportation is
logistics. So now, we are talking of that logistics and transportation that the movement from
one end to another end, from where house to retailer, retailer to customer, manufacturer to
wholesaler.
And now a days, we know that E-Commerce is a very current word, buzz word in the area of
supply chain. So where products are directly moving either from manufacturer or wholesaler
to customer by-passing the retailer. So transportation is becoming a very important area in the
supply chain and analytics will help us to a great extent in solving the transportation related
issues.
So now let us see that how transportation related issues can be solved with the help of supply
chain analytics or simply the data analytics tool. When you are getting products at a particular
supply chain stage, let us say as wholesaler, so you are getting products from wholesaler,
from manufacturer. So products coming to wholesaler from the manufacturer are coming in
big sizes, in big lots.
So you have sufficient economies of scale when products are coming to a particular stage and
when products are going out of that stage the size of each consignment is much smaller than
the inbound consignments. So you lose economies of scale in the outbound transportation. So
77
what I am trying to say that you have higher economies of scale in inbound transportation and
to certain extent you lose that economies of scale in your outbound transportation.
And this is a very crucial issue that how far you are able to reduce your economies of scale
how far you are able to sacrifice your economies of scale. Because when you are sacrificing
economies of scale it means you are moving towards this side, you are becoming more and
more responsive and when you are keeping economies of scale in your supply chain even in
the outbound transportation you are remaining to the efficiency side.
So now depending upon your supply chain strategy you can decide certainly you will lose
some economy of scale in the outbound strategy, outbound transportation. But to what extend
you are ready to sacrifice whether it is 50%, whether it is 60%, whether it is 70%, whether it
is 25%. So now loosing particular economies of scale will determine your position in this
efficiency, responsiveness spectrum.
If you loose too much you become very responsive. It means you are becoming very close to
your customer. So that you are able to fulfil the requirements of a customer immediately. But
if you are not ready to sacrifice economies of scale it means you are trying to serve a cluster
of customer with one single facility and then you are maintaining some level of economies of
scale in your outbound transportation also.
So, this is again a dynamic decision today I am operating with lot of responsiveness, today I
am operating with lot of responsiveness in my outbound distribution but tomorrow it is
possible that I may not like to have that high level of responsiveness and I may close some of
my lacks, some of my outbound lacks, and then I will try to consolidate to become more
efficient in my outbound transportation.
Vice-versa is also possible today, I am operating in a very efficient manner in my outbound

transportation. But tomorrow I feel that market is becoming more and more responsive. Data
will tell me that market is becoming more and more responsive and therefore I also need to
be more responsive and at that time I may open more lacks in outbound transportation. So
that I can reach closer to my customer and I become more responsive.
78
So the decisions with respect to outbound transportation is also govern with the help of my
analytical tool, with the help of my data, so I can take decision whether I should have more
efficiency or I should have more responsiveness in my outbound transportation. So with this
we come to end of first three important decisions which will help us to develop our supply
chain capabilities. We will discuss three more such important capabilities issue in our next
lecture. Thank you very much.
79
Lecture-07
Supply Chain Drivers
Welcome back, so now we are into the seventh lecture of this supply chain analytics course.
So far we have discuss the importance of supply chain, we have discuss the role of data
analytics into the supply chain management, what is the real time decision making in supply
chain management and then in our last lecture we discuss the role of supply chain strategy
and we discuss some of the drivers which are important in making the supply chain strategy.
Because how do we use those drivers, how we make them working in real life, so we
discussed three important drivers in our last session, these were facilities that what type of
infrastructure you are building with respect to factories, with respect to warehouses, with
respect to retailers, with respect to distributors, what will be the role, what will be the size,
what will be the location. So, that is one very important area of decision which we have to
taken into consideration for our supply chain management.
Because as we have discussed in our earlier lectures that supply chain strategy ranges from
the spectrum of efficiency to responsiveness, so how do we handle these drivers that will take
our supply chain either on the efficiency side or on the responsiveness side. So we discussed
facilities, then we discuss second very important driver that is inventory. Inventory is a very
important phenomena.
And lot of discussion will take place in supply chain management classes about the inventory
management in our supply chain and third important driver which we have discussed that is
the transportation. How do we handle the transportation because logistics is also one of the
very important elements of supply chain management, so we have discussed these three
important areas in our last lecture.
80
Now going further into this discussion of supply chain strategy and handling of driver to get
right kind of supply chain strategy, we have the fourth driver that is pricing. Pricing is the
fourth driver which will help us to take our supply chain from the efficiency to the
responsiveness side. And pricing, most of us may feel is a area is one of the actor in the
marketing domain, but pricing can play a very important role in supply chain management
also.
Now it is very interesting question and we will love to answer that how pricing can play a
role of driver in case of supply chain management. Because when we talk of four Ps in a class
of marketing when it is very appropriately said the role of pricing at different stages in the
product life cycle, but how pricing plays important role in case of supply chain management.
So it become very interesting, now as we know we have discussed in our very first class that
supply chain management is that function which ensures the proper balance between demand
and supply and we know for lot of products there is a period of very low demand and then
there is a period of peak demands.
You have example of electricity, you have example of tourism, you have example of
investments. So these are some of the areas where all of a sudden demand increases to a very
high level, a peak level and rest of the time demand remains on a low level. You can
understand like in case of tourism this is the period and this is the volume, now the volume of
81
tourists visiting a particular tourist place for 6 days a week will remain at this level from
Monday to Saturday.
But as Sunday comes you will find a quantum jump in the level of tourist visiting that
particular place because Sunday is a holiday, so you reach to a very high level of demand on
Sundays for the product known as tourism. So on for various other products you have a
particular time when demand increases to a very high level, you visit restaurants and during
evening hours around 8 o'clock to 10 o'clock.
You will find long queues in front of good restaurants because that is the time most of us will
like to visit the restaurants for our evening meals, so restaurants give the concept of happy
hours and this happy hours is nothing but using the pricing tool, using the pricing driver for
matching of demand and supply, you have a fixed capacity in a restaurant, you cannot
increase or decrease that capacity with respect to your demand.
But by creating a differential pricing mechanism, you can to some extent match the demand
and supply. So when in happy hours you reduce the prices you give extra benefits if you visit
before 7 o'clock, if you visit a restaurant before 7 o'clock on a Saturday evening they give
you one extra beer, they will give you one plus one type of offer.
So that some of the people who are price-sensitive can take that offer and then they can
equate the available supply which is there with the restaurant with the increasing demand
during the peak hours, so some of the demand of peak hours can be shifted to non-peak hours
or off hours by this kind of differential pricing. So how do use this differential pricing so that
you can match demand and supply that is one very interesting area.
You can see in some of the countries like in Singapore and in some of the European countries
on highways, on toll ways where we pay toll for using a particular highway, if traffic
increases beyond a particular density the toll rate increases and the toll rate increase is
actually to divert some of the passengers using those toll rates to those roads which are non
toll roads. So that you can reduce the traffic congestion on the toll roads.
Because toll roads, people want to use for driving at higher speeds for covering distances in
less time but if traffic congestion is there any purpose will be lost, so the pricing mechanism
82
can be used to divert, to give alternative routes to suggest people to go on alternative routes
because only those people who require emergency solutions with the help of that toll road,
can use that would effectively.
So pricing is basically used for matching your demand and supply and in variety of ways
whenever you find there is a kind of discount, there is a kind of some kind of offers for using
a product for off time or off routes you have this pricing mechanism, so pricing also give few
ideas that your supply chain is mostly on the responsive side and this is efficiency side, so by
pricing some of the customers normally you are positioned here.
But by using this pricing mechanism some of the customers can be shifted to this side. Those
who are looking for low cost solutions, those who are looking for low cost solution to take
the advantage of this pricing tool they will take their position to this side so same facility,
same infrastructure, same type of paraphernalia can be used by the pricing mechanism to shift
some of the customers from the responsive side to the efficiency side.
And then you can serve in a better manner to those customers who are actually looking the
responsive solution, those who are ready to pay premium higher prices for getting the
responsiveness from the supply chain. So, the pricing is a very effective tool, it is not only in
the marketing but also in the supply chain that we can use it for the customer satisfaction
looking into the interest of responsive or efficiency type of customers or you can say that
pricing helps us, there are two ways in which two categorical statement in which we can
understand the use of pricing.
One pricing helps us in meeting the demand and supply gap, so you can match your demand
and supply that is one important categorical statement. The second categorical statement is
that price in help us in filtering out efficiency looking and responsiveness looking for
customers from your entire customer base. So those customers who are more interested in
efficiency they will filter out to this side and those who are ready to pay premium for the sake
of responsiveness in the supply chain they will remain to this side.
So pricing can help you to filter out your customer and appropriately you can design your
further solutions you can go for new product development as per this filtering process. So
pricing is another important driver. After pricing, there is one more important driver in the
83
supply chain which helps in deciding the position of our supply chain on this efficiency
responsiveness spectrum, that is information.
Information is another important driver, which is very important and particularly after 1980s
after this IT revolution, the importance of information as a supply chain driver has increased
tremendously and therefore nowadays it is almost impossible to believe that a supply chain
can exist without robust information infrastructure. The success of companies like Walmart is
mostly attributed towards their very robust information infrastructure.
So information infrastructure also plays a very important role in the supply chain positioning
from efficiency to responsiveness, and this is one particular driver irrespective of cost you
invest in this information infrastructure. It helps you either in achieving efficiency or in
achieving the responsiveness. You need not to handle it differently, but the way customer
wants you can provide that solution immediately or at low cost.
So with the same infrastructure, with the same type of information system you can achieve
both the things efficiency and responsiveness. So we say normally in our supply chain
language we say that information driver actually has been waived efficiency type of
characteristics in a responsive supply chain. Information driver has waived efficiency
characteristics a responsive supply chain. So what I am trying to say that Walmart is an
example of highly responsive supply chain.
You find large number of products under single roof and at the same time those products are
considered to be everyday low priced product, that is the USP of Walmart and that USP is
being achieved by using the information network with the help of this information they very
appropriately handle their inventory management, they very appropriately handle the
transportation management and as a result of that they are able to achieve high level of
efficiency in inventory and transportation uses.
And because we are able to maintain large number of SKUs a skipping units, so it is a very
responsive supply chain that way, so because of information infrastructure, because of proper
information network you have in your supply chain you are able to put the characteristics of
efficiency into a responsive supply chain. The other important use of information in a supply
84
chain because supply chain is a game of mutual trust between various entities of supply
chain.
We have from manufacturer to wholesaler, to retailer, to customer. Now whatever we are

discussing this only is possible the result of these things are only possible when there is a
proper coordination, when there is a mutual trust between various entities of the supply chain.
Now earlier when it was a manual process when the information, information networks were
not very strong.
At that time it was possible that may be intentionally, maybe unintentionally some of the
information, some of the data, some of the important piece of business was not shared with
your supply chain partners. But now information networks, information infrastructure has
eliminated all these possibilities. Now because of information infrastructure is available and
all the entities are properly integrated in that information infrastructure.
So whatever is happening at any place in the supply chain at any stage in the supply chain
that will be known to all the members of the supply chain, so the very core of the supply
chain management on which it is based that is the trust, that is the corporation, that is the co
ordination. All these things can be easily achieved with the help of information infrastructure
and all kind of manual interpretations, all kind of manual passage of information can be
eliminated with proper information infrastructure.
So as we are moving into more sophisticated information sharing, we are moving into
collecting more data related to better supply chain management the importance of
information as a driver is continuously increasing and probably this course of supply chain
analytics is based mostly on the ability to share the information across this supply chain. So
that the manufacturer, the wholesaler, the retailer.
All these entities can take decisions in real time environment. So information is a very
promising, a very important driver now for the success of supply chain management and then
another important driver which is going to take very important call for India as a nation that
is sourcing. In our supply chain there are large number of activities, you develop new
products, you manufacturer them, then you distribute them through and network of
wholesalers and retailers.
85
And for that purpose you need to have some kind of transportation arrangement, so that these
physical products can move from one stage to another, from that is stage to further and
further. So these many activities are involved. Now it is almost impossible for any single
organisation to have enough competency in all these activities which are required in a supply
chain.
I may be good in manufacturing, I may be good in developing the network, I may be good in
marketing a product but it is very difficult that I am excellent in all the fields. So now it is it
time, it is the time where I concentrate only on those areas where I am excellent and I leave
or I take help of other people who are excellent in their respective areas and therefore the
concept of sourcing comes.
But I am limited in my supply chain to those areas where I have core competency, where I
can do that work in most excellent manner, world class manner, so I may be a good
manufacturers, I may be a good distributor, I may be a good transporter and for rest of the
activities we take the help of different type of parties and these are known as 3PL third party
logistics because they are not the primary contributor to my supply chain.
But they are providing a service I have sourced them to provide particular level of service in
my supply chain. In this supply chain to understand the concept of this 3PL third party
logistics, this manufacturer can be the first party, this wholesaler can be the second party,
now if I want to transport some of the products from manufacturer to wholesaler, so either I
can send my own truck, I am owing trucks, I am owing lorries, so I can send my own lorries
from this end to this end.
So this is a direct transportation between first party and second party, but sometime I feel that
I am not a competent person or it is not right for me to handle this transportation activity also.
So in that time I can involve somebody from outside, I can take help of Gati, I can take help
of DHL, I can take help of some other transporter who can help me in transporting my
products from this end, this facility to this facility and this is third party logistic.
Sometime I feel that this manufacturing is not my cup of tea, I am good in distribution, so I
can outsource manufacturing also and you will see in large number of our FMCG products if
86
you see the wrappers you will find that the product is manufactured by somebody else and
marketed by somebody else and we know normally the one who is marketing the product.
So in that way, the manufacturer is outsourced now, so depending upon my core competency
because I am a good marketer, so I want to limit it myself only to the marketing activity and
not to the manufacturing activity. So in this particular case the manufacturing it outsource, so
as I said that in whatever activity I feel that I am competent, I will do that activity on my own
and rest of the activities I can outsource.
Now outsourcing has its own plus and minus. Sometime outsourcing is very good, but
sometime when you see economy as a whole, sourcing cannot be as good as providing the
employment to a local economy etc. Many a times you outsource your activities to some
other country because you find you can get cheap labour, you can get easy resources, but at
that time the employment type of issues from your economy go to some other economy.
So that is the kind of political issues which are there in case of sourcing, but if I limit only to
the supply chain point of view, sourcing can help me in achieving efficiency also, sourcing
can help in achieving responsiveness also. So depending upon the competency of outsourced
partner, it can be useful in achieving efficiency or responsiveness. Sometime my outsource
partner is very cost conscious, so in that case with outsourcing I can achieve the efficiency.
Mostly in the case of transportation if I am taking help of somebody in transportation in

outsourcing my transportation activity normally it is helpful in achieving the efficiency and
you see when we are outsourcing some kind of activities which are emergency oriented, in
that case it is for achieving the responsiveness. So depending upon what type of partner I am
choosing I can achieve efficiency or responsiveness through sourcing.
It can help you in reducing the cost of your product or it can help you in providing the
solution at a very fast rate. So efficiency and responsiveness can both be achieved by taking
their partner, appropriate partner in your supply chain activity. Sometime you do not want to
outsource a particular thing, because you know that your partners will not be able to achieve
either than required level of efficiency or required level of responsiveness.
87
You have the example of companies like DHL etc, they own, their own aircraft to achieve a
particular level of responsiveness, so they feel by owning their own transportation feel you
achieve a particular level of responsiveness. On the other hand you have the example of
Indian Post. Indian Post uses this facility of Indian Railways for transporting their packets
from one place to another place and through that they achieve high level of efficiency.
Because of low cost of transportation from the Indian Railways the same is available to
Indian Post and therefore post office is able to maintain the efficiency through outsourcing
and DHL on the other side is not outsourcing to achieve the high level of responsiveness. So
how do we handle the outsourcing partner, in what case we should use or what case we
should not use that will define your position on this efficiency responsiveness spectrum.
So with this way we saw that how we need to handle our 6 important drivers to have a
strategic location on the supply chain spectrum from efficiency to responsiveness, because
you will not take decision about a particular driver, you need to take decision about all six
drivers together, starting from facilities, inventory, transportation and then these three
pricing, information and sources. So these are the six drivers and you need to take decision
for all the six drivers in a single decision environment. So that you can use these drivers
effectively for better results for better achievement of the strategic point of view of your
supply chain.
So finally if I want to summarise we can say that this efficiency to responsiveness these
things are dependent on effectively handling the 6 drivers, so if I put these 6 blocks here
88
representing 6 drivers, so you can say that these 6 drivers help me in achieving my supply
chain strategy and that supply chain strategy is basically from this efficiency to the
responsiveness side and then this supply chain strategy will help me to achieve my business
strategy.
The overall business strategy which I have that will guide by supply chain strategy and that
supply chain strategy will be achieved through these drivers of facility, inventory,
transportation, pricing, information and sourcing. So you need to see that how are you using
collectively because it is very important to understand that all these drivers should be handled
in one single direction.
If you handle for example I want to have a very efficient supply chain and today I am taking
decision about facility, so if I want to have a efficient supply chain, so facilities should be
centralised, I should not have large number of distributed facility, so that I can maintain
economies of scale. But on the other side when I am taking decision about inventory I am
keeping large number of inventory.
When I keep large number of inventory obviously my responsiveness will increase and my
cost will also increase. So these two decisions or not in sync with each other, facilities are
more towards the efficiency side, but inventory more towards responsiveness side. So there
will be a conflict between handling these two drivers of facilities and inventory. So therefore
it is very important.
That all these driver should be handled in such a way that you get a particular position for
your organisation on this efficiency responsiveness side and there should not be any conflict
between these drivers. Otherwise there will be a problem and you will not be able to achieve
any particular position for your organisation. So with this we will like to close the today
session about the drivers of supply chain.
And how these drivers will help us to achieve a particular position on this supply chain
spectrum and therefore we will see the use of data to take appropriate decisions about all
these drivers because the supply chain management is all about taking right decision with
respect to facilities, with respect to inventory, with respect to transportation, pricing, sourcing
and information.
89
So in our further classes, in our further lectures we will see that how the data, real time data
will help us in taking decisions with respect to these drivers. Thank you very much.
90
Lecture-08
Developing Supply Chain Strategy
Welcome back, so now we are moving into the eighth lecture of this course and recently we
discussed about the role of drivers in making a supply chain strategic. So now moving further
into the discussion of supply chain strategy.
We will see that how the first process of the supply chain strategy management that we need
to understand the demand and supply uncertainties. And you see the demand and supply
uncertainties are related to variety of things. Now some of the uncertainties which are there
from the demand side in a supply chain. Let us see a customer can vary the quantities
demanded.
Now we see in the last discussion that how during a particular time of the day that demand of
food products may increase, demand of seats in restaurant may increase. How the demand of
cinema tickets on a weekend show may increase. So quantity of products required at
particular time may increase or decrease. So you have uncertainty related to range of quantity
increases.
91
And as this range of quantity is increased or decreased by the customer, the implied demand
uncertainty as a result of that is that wide range of quantity implies that greater variance in
demand. Means if you have large number of SKUs, so in that case this range of quantity will
create further higher variance in the demand uncertainty. So that is one very important thing.
And now a days you can see a lot of product categories the uncertainty related to quantities
are there and particularly if I take you to the side of humanitarian supply chain in that
particular supply chain you have huge amount of uncertainty with respect to quantities. So
quantity changes and this quantity changes is even more severe when you have wide variety
of products to be offered in your supply chain.
Then you, customer need is also with respect to lead time. Now customers are continuously
expecting that lead time, when I order the product and what I received. That difference, that
time gap is known as lead time. The time between you order and you receive the product, that
time is known as lead time. Now there was a time 20 years-30 years back when I used to
book a Bajaj scooter.
And I used to get deliveries after 4 years or 5 years or 6 years of that time. But nowadays
what is expected I should have money in my pocket and I should walk to a retailer shop and
immediately I must get the product. So now we are continuously expecting reduced lead
times. I am not interested in waiting for my product and this particular aspect created a
landmark change in the supply chain management of Dell computers.
Before that, Dell was taking sufficient lead time of around 2 to 3 weeks to deliver their
products, but as Dell could understand it at a very appropriate time that now customers are
not going to wait for the product and immediately they change their supply chain and went to
the retail market. So lead time is continuously decreasing and now as a result of that when
customers are expecting low lead times there is less time to react for the orders.
Today I am giving the order and you have one day or even less than one day to fulfill my
requirement. So you do not have much time and you need to develop the capabilities in your
supply chain. We all are looking, a certain degree of responsiveness in my supply chain. And
therefore this is creating another kind of, you can say a stress in the supply chain that you do
92
not have enough time to react or to understand or to satisfy the customer requirement you
have to do every requirement fulfilment on war footage.
Then variety of products required are continuously increasing. You can feel in your last 10
years of period or 15 years of period that how the variants of products are increasing.
Because we all are living in marketing era. This is the era of marketing and in the era of
marketing we say the customer is king and we are going to fulfill each minor requirements of
the customer.
And therefore we are creating continuously more and more variants of the products. So the
variety of products are increasing and this increased variety of products are creating another
kind of pressure because when you have more number of SKUs, the stock keeping unit, you
have so many varieties of soaps, you have so many varieties, flavours, pack sizes of biscuits,
you have so many varieties of namkeens.
So all these things are variety of products that is increasing and then you need to understand
the demand per product is becoming more disaggregated. So that is also a challenge to the
supply chain and you need to when I am talking in strategic manner, you need to understand
that how to fulfill this disaggregated demand. Because for each small product you need to
develop the supply chain solutions.
Then numbers of channels are also. Earlier we used to have simple brick and mortar type of
supply chain. But nowadays you have small Kirana shops, you have big malls and then you
have E-supply chains also. So there are different types of channels through which producer,
through which companies are reaching to the customer. They are following the conventional
wholesaler, retailer and customer model.
They are following directly from manufacturer to big retailers those who are the organised
retailers and then to the customer. And then there are companies directly from manufacturer
to customer. So numbers of channels are continuously increasing and customers have option
all the customers can go for different types of channels. So these are also creating a type of
pressure on the supply chain that you need to maintain, you need to manage the multiple
channels which are available in the market.
93
So what is happening E-channel, E-channel like E-supply chain, the big malls and small
kirana shops. The overall demand of the product, overall demand of the supply chain is now
distributed over these different channels. So you need to handle all these channels because
your overall demand is now disaggregated over more number of channels. So as you evolve
more number of channels you loose economies of scale.
So that is another challenge, because you have X number of customers total and now X
number of customers are distributed over 3 different type of channels. So each channel will
take care of if it is uniform distribution X by 3, X by 3 and X by 3 number of customer. So
that total customer demand is now disaggregated to more number of channels. So obviously
you lose some of the efficiency in managing the different number of channels.
Then you have another important area from the customer that is innovation. Customers
continuously expect the innovative products and as we all can observe the rate of innovation
is continuously increasing. The new products are coming at a much faster rate and we had a
time when we used to have black and white TVs and black and white TVs remain into the
market for more than 3 decades in Indian market I am talking.
But after 1980's, when colour television came to Indian market we are seeing in last 35-36
years a very fast change in the colour TV Technology. You had variety of colour TVs where
CRTs were there, then after that we have LCDs, LEDs, High Definition, full HD, then TV
with Wi-Fi, smart TVs. So in last 35-36 years you can see the rate of innovation in colour TV
technology.
And it is not only limited to colour TV technology you can see in the information technology,
in the area of your mobile communication, in the area of automobile, in the area of your
household appliances, everywhere you talk you will see that rate of innovation is
continuously increase and this rate of innovation is putting another pressure because the life
of the product is not much and as a result of that new products tend to have more uncertain
demand. When a product has passed or when product has a longer maturity cycle.
94
This is not a class of marketing so I will not go in much detail. But you can understand or you
should know that this is a generic kind of product life cycle. We normally abbreviate it as
PLC, Project Life Cycle. On x-axis we have Time and on y axis we have Volume. Earlier in
previous time we used to have a sufficiently long maturity period, where during this period
demand was almost very certain, it was almost static kind of demand.
But nowadays because of this innovation you will not find a longer maturity period, a product
comes and reaches to a particular level and immediately declines. Because a new product
comes into the market. So in graph, this is the earlier time and this is the present time. And as
a result of that you have more uncertainties in handling the demand of new products. That
how a new product will feature into the market.
You have the example of iPhone 5, iPhone 6 and iPhone 7. So Apple is continuously
introducing new iPhones into the market. But we all know the response of the market towards
iPhone 7 is not as good as it was for iPhone 5 and iPhone 6. So you cannot take decisions on
the basis of past data. You need to see that each new product will come with its own
uncertainties.
And this is happening because of rate of innovation and customers are becoming lover of
innovative products, but more innovative products are putting it different type of pressure on
the supply chain uncertainties that you have uncertainties related to demand of the new
95
innovative products. Then you have expectations from the customer about increased service
level.
We want as soon as products demand come to my mind, as soon as I walked into a shop
retailers place I must get the product immediately. So I want a very high level of service for
my expectation. Now because of this high level of service requirement the companies supply
chain now has to handle unusual increase in demand. There is all of a sudden a very high
increase in demand and all the customers are expecting same high level of service level.
And therefore it again becomes a challenge for the supply chain that how do we handle that
sudden increase in demand and all the customers are expecting same level of service.
Because they all are ready to pay extra premium, they all are ready to pay the additional cost
for getting the higher service levels. But you have limitation of supply chain. So all these
things, all these customer needs ranging from quantities to the service level.
All these things are creating different type of implied demand uncertainty and these implied
demand uncertainty is to be handled for developing a proper supply chain strategy. So now
moving further into this because of this implied demand uncertainty now you can understand
from the spectrum of this implied demand uncertainty.
That on this side you have a very predictable supply and demand. And this is like my earlier
PLC, the previous PLCs. Some of the products like I take the example of salt, the common
salt it is still like earlier PLC. So you have a very certain demand and supply of this type of
96
product. So it is coming on predictable supply and demand side of this implied demand
uncertainty where the implied demand uncertainty is very very low or almost absent.
On this side of the spectrum the implied demand uncertainty is almost absent. Now coming to
this side of the spectrum to the right side of the spectrum, hear you have very high level of
uncertainty in the supply and demand like as I mentioned about i7, iPhone 7 and it is a new
phone just launched into the market and this product has a very high level of uncertainty that
how customers will respond to this type of new product.
And therefore I have placed this i7 to the right side of this is spectrum. So now you have
these two extremes where on the left you have no uncertainty, suppliers are also having a
very routine kind of demand and they are habitual, they have developed a good infrastructure
to meet that kind of demand. Here you don't know what type of demand it will be. So all the
suppliers are in a state of confusion that should we produce large volume, should we produce
low volumes.
If we produce large volume and customers do not respond that, you have huge inventories. If
you do not produce in large volumes and customers are also expecting high service level,
they want product immediately i7 is launched today and I want today the product. But since I
am very much cost conscious so I have not stored enough inventory of i7 so most of the
customers will not be happy because I am not able to provide the required service level.
So you have lot of challenges on highly uncertain supply and demand side of this spectrum
where you see that all these products will have to face and therefore we will talk of flexible
supply chain that how initially you need to have a very responsive supply chain, a flexible
supply chain. So that you can make changes as per the requirement of your uncertainties.
Then somewhere in the centre, not exactly in the centre but somewhere in the centre place of
this spectrum.
You have somewhat sudden, it is somewhat uncertainty, you have a balance type of place
where you have some kind of predictable supply and uncertain in demand or uncertain supply
and predictable demand or somewhat uncertain supply and demand. So you have possible
combination of uncertainties and predictabilities of supplies and demand. And therefore some
routine kind of models like you take the example of Scorpio.
97
You take the example of Alto, you take the example of WagonR, and these types of you take
the example of CD 100 motorcycle of Hero. These types of examples can fit in at this type of
place where you have some existing routine kind of model which is already popular in the
market. So that product will fit into this place. So where you have a very predictable supply
and demand obviously you cannot make huge profits.
Where you have very unpredictable situations either you can make huge profits or you will
go down. Because of not able to achieve the desired level or because not able to answer the
desired level of uncertainties. But here in this case in the central location where you have
some uncertainty is in your supply chain may be either because of supply side or because of
demand side.
But this is a position which is very hard on position. Because in current environment of
innovation where PLCs are continuously shrinking even in this stage you are able to maintain
a longer PLC for these products and then only this type of positioning is possible for your
product. So it requires a very integrated effort from all the functional area, not only from the
supply chain.
But from the product development team, from the marketing team, from the distributors point
of view, from the after sale service point of view. All those who are involve in delivering
value to the customer their important role is there to get this position on this implied demand
uncertainty spectrum. After understanding the concept of this implied demand uncertainty
you can see that as a result of some of the researchers.
98
When you have low implied demand uncertainty, your product margins are low as we saw
just now in the case of salt, you have low product margin because it was position on the left
side of the this implied demand uncertainty spectrum. When you have high implied demand
uncertainty your product margins can be very high as evident in the case of i7, which is
position on the extreme right side of the implied demand uncertainty spectrum.
The average forecasting error is very low as low as of 10%, in case of low implied demand
uncertainty. But if you are positioned on the right side of this is spectrum your answer
forecasting errors can be as high as 40% to 100%. So that is very unfortunate just to quote
you some data. When in Bombay now Mumbai, this Bandra Worli Sea Link was conceived.
So the initial forecast for that Bandra Worli Sea Link was about 12 lakh cars per day 1200000
hours per day.
But if I give you the present uses it is just 45000 cars per day. So that is the type of errors
which are possible in case of high implied uncertainty product. So this is a kind of I can say
blunder in calculation or in the forecasting. But it is always possible when you are dealing
with a product having very high implied uncertainty. Then stock outs, the number of times
products goes out of stock and a customer wants that product.
So in case of low implied demand uncertainty products it is just 1-2% very low. Most of the
time means the opposite of this is you are able to achieve 98 to 99% of service level, a very
high service level you are achieving in case of low implied uncertainty products. But in case
99
of high implied uncertainty products your stock out rate is 10% to 40%. It means your service
level can decrease as low as up to 60% in case of high implied uncertainty.
And if the service level is 60% you can correspondingly understand the customer satisfaction
level, means out of hundred times only 60 times customers get the products, 40 times they are
not able to get the product when they are visiting the retailer or retailer is visiting the
wholesaler or wholesaler revisiting the manufacturer. So that is the reason of low customer
satisfaction in case of a new product because of this low service levels.
Then because of forecasting error and because you are piling up the inventories you need to
clear those inventories and end of season marking down pricing is now a days we see very
often. So in case of low implied uncertainty you hardly see this type of activity. So it is
almost zero, you never find any kind of end of season sale on product like salt or grocery
items or wheat items. There is no such concept.
But you find lot of end of season sale for garments, for shoes and for so many other product,
because of forecasting errors and because of these reasons you are piling up inventory. So 10
to 15%, 25% of the time you have this type of marking down of the prices. So you can see the
only thing which is favouring the high implied uncertainty is the high product margin. But
you have correspondingly very high risk of forecasting error, you have very low service level
of 60%.
And you also can go up to 25% of end of season marking down. So all these things are big
challenge for high implied uncertainty. But it is almost unavoidable. The reason is because
whenever you are launching a new product, the implied uncertainty is always very high and
since it is the time of innovation, it is the time of shrinking PLCs. So to remain in to the
market, to remain competitive into the market you have to be launching new product.
But you have to be very careful and now it is very important to understand that the data
analytics. The data analytics can help us to change these figures. These figures of forecasting
error can be reduced from 40% to 10% or 15% if I use proper forecasting methods. The more
adaptive forecasting methods, more adaptive forecasting methods if I use probably I can
reduce the forecasting errors.
100
If I can reduce the forecasting errors both these things, the stock out rates and the marking
down will also be reduced. Because when I am able to correctly forecast, so I will stock only
that much product with me and this correspondingly will help me to achieve higher service
levels and to reduce by marking. So both these things are actually related with the forecasting
errors and with the help of data analytics with the help of adaptive forecasting.
I will be able to somehow reduce this value of 40% to 15 or 20% so that is one big advantage
which I am going to have with the help of data analytics into the supply chain management.
Now once we understood the uncertainties related to demand and supply in my supply chain
the next important point is to understand the supply chain. That how my supply chain is
going to answer these uncertainty, how I am going to full fill the gaps which are there in my
supply chain. And that is the second step in developing the supply chain strategy canvas. So
now let us see that how does the firm best meet the demand.
What are the different typical you can say coping mechanism available with the firm through
which firm is able to meet its demand, like in the class of drivers we discuss that pricing is
one such mechanism through which firm can meet the demand and supply, when the
dimensions which are describing the supply chain and how it is responsive supply chain,
those dimensions whether it is through pricing, whether it is through facilities, whether it is
through a distribution, whether it is through the inventory etc.
101
And supply chain responsibility to respond wide range of quantities the number of products,
the how much you are demanding, so how my supply chain is responsive to that. When
reduce lead time, then variety of products, then ability to innovate new product continuously
and to meet a very high service level. So I need to see that how my supply chain is able to
answer all these abilities which we require from the step one of our strategy development
process.
Now it is also very important to understand that to achieve a particular level of

responsiveness there is a cost associated to that. When you want to achieve a particular level
of responsive you need to invest into developing the supply chain capability. So there is a
cost associated to develop the particular level of responsiveness. Efficiency we have already
discussed that it is inversely proportional to the cost of making or delivering the product to
the customer.
And as we just discuss that increasing responsiveness results in higher cost and it lowers the
efficiency of the supply chain. So efficiency and responsiveness are somehow inversely
related to each other and now we will see that to develop a strategic fit of the supply chain,
we need to put the supply chain capabilities on the spectrum of implied demand uncertainty.
So today we close our discussion at this point.
And in our next lecture we will see that how do we map the supply chain on the
responsiveness spectrum that and later on we will see that how this responsiveness spectrum
102
is embedded in that imply demand uncertainty. So thank you very much for this eighth
session of our supply chain analytic course.
103
Lecture-09
Strategic Fit in Supply Chain
So welcome back, we are discussing the development of supply chain strategy, we discuss in
our last session about the implied demand uncertainty spectrum. We discuss that there are
products where uncertainties are very low. We took the example of salt for that category and
we also saw that on the right side of the spectrum there are products having very high level of
uncertainties and we took the example of iPhone 7 that how that high implied demand
uncertainty is there in that type of product.
And then we took the example of some of the existing automobiles like CD 100 motorcycle
of Hero Motor corp. like Alto, like WagonR, like Scorpio. These types of products where we
have somewhat uncertainty and somewhat stability. So that is the first step in the
development of supply chain strategy. Then we also discuss about the capabilities of supply
chain. How my supply chain can answer the required questions which are coming because of
uncertainty is in my supply chain.
And now we are trying to develop on the basis of capabilities in my supply chain a particular
spectrum of responsiveness verses efficiency, cost as we have discussed in our last lecture
cost is inverse to efficiency. If it is a high cost supply chain it means low efficiency, if it is a
104
low cost supply chain it means high-efficiency. So it is a relationship between responsiveness
and efficiency of the supply chain.
And this is that cost responsiveness efficiency frontier. This curve is the cost responsiveness
efficiency frontier that you have a particular level of responsiveness with a particular level of
efficiency and when you are trying to increase the responsiveness of your supply chain the
cost of supply chain or the efficiency of the supply chain has to be trade off. So this is that
cost responsiveness Frontier.
We want low cost of the supply chain or you can say high efficiency, high supply chain and
high responsiveness so this is the most desired frontier. This is the most desired frontiers
where you have high responsiveness and high efficiency, low cost means high efficiency. So
this is that spectrum on which we want to locate our supply chain but most of the time in real
life we see that if you have high responsiveness in your supply chain you have low
efficiency. But this is the most ideal situation which is desirable in any kind of supply chain
high responsiveness and high efficiency. So this is that curve.
Now finally we are coming to step 3 of the supply chain strategy development that we are
developing a strategic fit in our supply chain with respect to uncertainties of demand and
supply and with respect to efficiency and responsiveness that is the capabilities of the supply
chain.
105
So now let us see how this is developed. You have highly efficient supply chains, this is that
responsiveness efficiency spectrum and this is highly responsive supply chain. Now you can
see on this spectrum that companies like Apple, companies like Samsung which are highly
innovative, continuously introducing new products are highly responsive to the markets
requirement and that is why these are placed on the right side of this spectrum.
On the other hand we have integrated Steel Mills which are continuously producing same
type of products so these are highly efficient because of their particular label of integration
which they have achieved and in between you have companies like just for an example very
popular low cost apparel company Kidley, and this towards of the efficiency side. Then you
have towards responsiveness side most of the automotive production companies.
Because they are continuously introducing new models, they are producing more variants of
their product. So these are towards the responsiveness side. So you can actually have
intermediate locations depending upon efficiency and responsiveness and their relative ability
to make their supply chain efficient and responsiveness. So from this point to this point, you
can have various intermediary locations and accordingly you can find on your own.
Different examples which you feel are suitable and I must give you this exercise at this
moment that try to see different type of examples around you and put those examples on this
responsiveness spectrum at different location. See where Maruti fits into this, see where Hero
fits into this, see where post office fits into this, see where Amul fits into this, see where
Mumbai Dabbawala fits into it, so try to see various examples around you.
106
Try to see the supply chains of new start-ups, new start-ups are coming up and where they
are, where they are on this responsiveness spectrum, so this will be a good hands on learning
for you and you can see that it will be very interesting as an outsider, as an expert of the
supply chain. Now you will be able to understand that whether the supply chain is towards
efficiency side or towards responsiveness side. So this is the responsiveness spectrum. In our
last class we discuss the implied demand uncertainty spectrum, now we are discussing this
responsiveness spectrum.
And now we have actually clubbed both these things. In this figure we have clubbed the
responsiveness spectrum and implied demand uncertainty. This we discussed in the previous
lecture. Today we are discussing this responsiveness frontier that from the efficiency to
responsiveness side, we have this spectrum and this is the uncertainty side to certainty side.
And when we are trying to combine these 2 spectrums you develop this zone of strategic fit.
This is very important.
Now as a supply chain manager, I always want to keep my supply chain in the zone of
strategic fit. This is the place, this is the zone where I will like to place my supply chain.
When I am inside this zone, when I am inside in this zone it means my supply chain
capabilities are enough to respond to various uncertainties which are coming to my supply
chain. But if I am not in this zone then probably I am not able to develop that type of
capabilities which are required to answer the uncertainties which are coming in my supply
chain.
For an example if I am towards this uncertainty side, if I am toward this uncertainty side so if
I project this uncertainty into this graph, so appropriately I need a more responsive supply
chain. So more responsive supply chain is the answer of uncertain demand and if I have a
certain demand then the efficient supply chain is the answer of that certain demand. And for
any intermediate location you can appropriately find a suitable level of responsiveness of the
supply chain.
If in case of uncertain demand you are having huge uncertainty in your supply chain because
of maybe more innovations, because of quantities, because of varieties etc. But the
capabilities are towards efficiency side. This is a purely a mismatch. This is purely a
107
mismatch and you are not going to achieve any strategic advantage of the supply chain.
Rather just type of supply chain arrangement is bound to give you failures.
So it is very important to understand that my supply chain capabilities which are exhibited on
this responsiveness spectrum are in line, are in sync with the implied uncertainty spectrum.
So if I have a more certain demand, if I have a routine demand, if I have a fix demand and in
that case I am developing a responsive supply chain useless, no point. Because if I am having
a certain customer may not be interested.
Rather customer is not at all interested in paying me any extra premium and that is going to
be a wasteful responsiveness of the supply chain. Because when I am developing
responsiveness, when I am having more responsive supply chain we just discussed
responsiveness comes at a cost. So it means I am providing a costly solution, a high cost
solution to my customer, which a customer of this point where uncertainties are low never
likes and therefore this is also a mismatch.
So if you can understand now that if you have uncertainty in the demand and supply you
should have a responsive supply chain and then you are located somewhere here in this
responsiveness zone of strategic fit. Then if you have more certainty in the supply chain then
you should develop efficient supply chains and you are located at this point of the zone of
strategic fit.
So whether you are located here or you are located here it is the zone of strategic fit where
you are located and therefore this is also a suitable position for the supply chain strategy and
this is also a suitable position. But if you are located somewhere here or somewhere here or
anywhere outside this zone probably your supply chain is not going to give you the kind of
advantage which you are expecting from your supply chain.
So this course on data analytics and supply chain is going to help us to take decisions that
what type of uncertainty are coming in my supply chain, because these uncertainties are not
constant. Uncertainties are going to come and new type of uncertainties are going to come.
So you continuously need to be very much watchful about the uncertainties. And I must warn
you, I must caution you that a company's position, a supply chain's position on this
uncertainty spectrum may change as you move from one market to another market as time
108
changes as you launches new products so with respect to all these factors your position on
this uncertainty spectrum may change and therefore you have to be watchful about changing
positions of your supply chain on this spectrum. And when you know that how your position
is changing appropriately you need to change or you need to update rather it is more
appropriate to say you need to update your supply chain capabilities because then only you
will remain into the zone of strategic fit. Otherwise if you have a particular level of
capabilities in your supply chain.
And let say, you have developed a product a new kind of automobile and initially during its
launch and growth stage you are toward the uncertainty side and therefore in the beginning of
the product life cycle you require more responsive supply chain. But as luck is there and now
the product is moving into the maturity side, so you need to develop capabilities in your
supply chain that you can come from this responsive side to efficiency side.
Because as you are going to the maturity of the period you are moving from uncertain
demand to certain demand. So you cannot keep all the time same level of capabilities in your
supply chain and Maruti is a wonderful example of this concept. When Maruti first came into
India when we have a very limited car market in this country where only few Fiat cars and
Ambassador cars and Contessa cars used to be there.
At that time when Maruti came so if I see from this point of view so initially it was an
uncertain demand. It's a different matter that supply chain was not so responsive at that time
because of various licensing reasons. But over a period of time Maruti's is demand for their
109
800 became more like a certain demand it became a commodity type of product and when it
became a commodity we were looking to change from responsive to efficiency.
And Maruti did that wonderfully and developed its network throughout the country to
provide more efficient solutions with respect to their supply chain. And as a result of that you
have a very strong network of Maruti giving highly efficient solutions to the distributors and
therefore supply chain of Maruti is another source of competitive advantage to Maruti apart
from their huge penetration to the car market, the new products, the low maintenance of their
products etc.
But supply chain is also providing a very important competitive advantage to Maruti. So
therefore what I am trying to say that the role of this zone of strategic fit is that over a period
of time the direction of movement because whenever a new product is coming into the
market it is at this stage of uncertain. So normally this is the starting position for most of the
products in this zone of strategic fit.
And over period of time we move to watch the origin from here we move in this direction, so
that over a period of time as uncertainties are reducing we need to change the supply chain
capabilities from responsive to the efficiency. But data analytics, the supply chain analytics
will help us to understand the rate at which these uncertainties are diminishing.
And appropriately we get time to change the capabilities of the supply chain. So therefore this
understanding because ultimately the whole objective of this course or the entire supply chain
management is to plan our supply chain positioning in this zone of strategic fit. So this
particular point is to be emphasised clearly boldly that what is the objective of this particular
zone of strategic fit.
And if you are in this zone of a strategic fit it is going to give you clear advantage that you
will always be having that type of supply chain capabilities which are required to answer
different types of uncertainties in your supply chain. So I hope that the purpose of this supply
chain strategy development is now clear to us, the process is clear to us and how drivers help
in development of the supply chain strategy that is also clear to us.
110
So after achieving the strategic fit we can discuss these two examples one is the example of
post office which is a very efficient supply chain and second is the example of a very
responsive supply chain that is DHL. So you can see if I can point out that somewhere here
we have the Indian post office, you have almost a constant demand in certain level of demand
is there which you can represent through almost a straight line.
If I want to represent the demand of Indian post office so it is more like this way. So over a
period of time the demand of product in case of post office is this horizontal curve. So post
office has developed a very efficient supply chain because that certain demand is there so you
can locate post office somewhere here. On the other hand DHL is a product, DHL is a
111
company where uncertainties are very fine because of reduce lead time, expectation of the
customer is very low lead time customer is expecting very high level of services.
And all these things are putting us towards high uncertainty in the demand and therefore you
see the DHL, DHL has developed to answer this uncertainties, DHL has developed
capabilities to suite the responsiveness and DHL will come somewhere here. So both these
organisations are taking the advantage of supply chain very strategically. DHL to answer
uncertainties develop the responsive supply chain and the location is here.
Post office in the same category because of the certain demand the efficient supply chain
located here. Both these organizations are into the zone of strategic fit. So you can discuss
more such examples where you find that same supply chain, same category of products can
be located either at the origin or at the extreme of the zone of strategic fit. Now once we have
understood the concept of this zone of strategic fit.
Now it is too important takeaways from this whole discussion that there is no right supply
chain strategy independent of competitive strategy. Now competitive strategy I derive from
my market what type of product to which type of customer I am going to offer. So post office
is looking as a welfare organisation, so it is looking for low cost market and therefore my
supply chain has to follow this competitive strategy of the post office.
On the other hand DHL is looking those customers who are very much time sensitive and this
is the competitive strategy of the DHL. So my supply chain strategy must response to this
type of requirement of DHL. And then another important take away is there is a right supply
chain strategy for a given competitive strategy. So if I want to have a customer who give
more importance to the time that I have to develop a responsive supply chain.
And if I am looking for a customer who gives more importance to money in that case I have
to develop a very efficient supply chain. So there is a particular type of right supply chain
strategy for a particular type of competitive strategy. So it is the role of supply chain manager
or the supply chain executive to understand that what is the competitive strategy and how do I
develop those capabilities in my supply chain.
112
So that my supply chain strategy in sync with the required competitive strategy. Otherwise if
my supply chain strategy is not in sync with the competitive strategy in that case there will be
a serious disaster in the organisation and we will not be able to achieve that competitiveness,
that advantage of the supply chain and the whole purpose will be lost.
Now some of the quick comparison between the efficient and responsive supply chain. So we
discuss that these are the two extremes of supply case strategy. And when we see some of the
items on which we can compare the efficient and responsive supply chains. So the first thing
is the primary goal. The primary goal, the primary objective of efficient supply chain is
lowest possible cost, you want to reduce the cost that is the most important objective.
And we have seen this in the case of Dabbawalas, you have seen this in the case of post
office, you have seen this in the case of public distribution system. So these are some of the
examples where we want to have efficient supply chain and cost is one of the important
considerations in those efficient supply chains. In case of responsive supply chain we want to
give quick response.
Now the meaning of quick response is that normally to understand it is with respect to time,
that you need to fulfill the customers' requirements with minimum time. So how fast you can
fulfill the requirement, how much you can reduce the lead time that is very important in case
of our responsive supply chain, so this is the primary objective. Then with respect to product
design strategy. We because now the primary objective is clear.
113
So rest of the things will be guided by these primary objective. In case of efficient supply
chains all the things are guided with maximum utilisation with low cost like those parameters
and in case of responsive supply chain all other items will be guided by how fast you can
deliver, how much you can reduce the lead time and therefore in case of product design
strategy.
In case of efficiency driven supply chains, you want to achieve product design with minimum
product cost, you try to design products where cost is very low. And here in case of
responsive supply chain you want to provide more modular solutions, you want to provide
more modular solutions so that you do not think of cost rather you think of that your product
should be more as per the requirement of the customer.
So it is highly customised and for that purpose you allow postponement, you go up to the
stage that once the order comes then only we will supply the product. Because you are
thinking that we need to make some kind of modular arrangements and then when the order
comes you should assemble the products and supply as per the requirement of the customer.
So we go with the postponement kind of strategy in case of product design for a responsive
supply chains.
Then the pricing strategy it is very simple in case of efficiency we have low margins because
not much uniqueness is there and low rate of innovation and there are routine items. So
product is already in the maturity stage. So for all these things you don't have a very high
margin in case of efficient supply chains. While in case of a responsive supply chain you can
go with very higher margins because you are taking lot of risk.
You are ready to take care of a lot of uncertainties in the supply chain and therefore higher
margins are possible and I just said that i7 type of products where you have a high
responsiveness as you take very high margins initially. Then the manufacturing strategy here
the focus is maximum utilisation of the assets. That is the concept, that economies of scale
that drives our manufacturing strategy.
On the other hand in case of responsiveness we look for flexibility that how we can use our
capacity in a more flexible manners, if more quantities are required you use more capacities,
if less quantities are required so you should be able to use less capacity of your entire
114
production system. So the capacity utilisation you want to have a flexible system in using the
capacity in case of responsiveness.
Inventory strategy we want to go with concept like JIT or we keep minimum inventory in
case of efficient supply chain and in responsive supply chains we keep sufficient amount of
safety stock so that you achieve high service level, so you keep buffer inventories with you,
that is also important part because inventory is a very important driver in our supply chain
strategic locations. The lead time, you can have lead time but you can reduce the time only
when you increase the cost.
And it is not expected and since in case of efficient surprises because the demand is very
certain it is a routine type of product. So normally you have very low lead time because of
good forecasting practices available in case of efficient supply chain, so lead time is normally
not important issue in case of efficient supply chain. But it is a very important issue in case of
responsive supply chain.
Because in responsive supply chain we want minimum lead times. So we want to reduce lead
time aggressively, so that our customer are always satisfied and for this purpose whatever
cost we are incurring, whatever additional cost we are incurring we are ready to pay for that,
so this time lead time strategy is very important for the responsive supply chain. Then
supplier selection strategy in case of efficient supply chain we look for cost.
And quality is not that important, but since our suppliers have developed over a period of
time. So we expect a particular level of quality, a particular level of quality is expected
because over a period of the suppliers have been developed. While in case of responsive
supply chain we expect flexibility, we expect speed, we expect quality, all these things we
expect when we are selecting a supplier.
Because we need that supplier can supply one product also, supplier can supply 100 items
also as per the requirement when we are introducing a new product into the market. But if
suppliers says I require only one product, one component for developing a prototype. But if
my supplier says no I cannot reply one item, I can supply at least 50 items. So probably that
supplier is not my choice for a responsive supply chains.
115
I need a supplier if I want one component, the supplier should have capability to supply
single component. If I need 100 components the supplier can supply me 100 components so
that type of flexibility, if I need a product in one day supplier can supply me in one day. If I
need a product in 3 hours supplier should be able to supply me in 3 hours. So that type of
challenges with respect to speed challenges with respect to flexibility.
And obviously because I am introducing a new product. So I expect very high level of quality
also because I have to make my mark into the market. So all these things are important when
I am selecting a supplier. And then transportation strategy. So in transportation strategy in
case of efficiency I look for low cost modes of the transportation. So post office is again a
very good example.
Post office is relying mostly on Indian Railways which is a very low cost of transportation
board and in case of a responsive supply chain we respire take care of our supply chain
through more responsive or faster mode of transportation and therefore companies like DHL
and Gati and all these overnight Express type of courier services are using air as the mode of
transportation because that is much faster as compared to rail.
And therefore they are responsive and post offices efficient supply chain. So these are the
comparison, this is the table which gives you in black and white the different type of strategic
decisions which we take with respect to efficient and responsive supply chain and as I said in
the beginning the decision are primarily governed by the objective of these supply chains in
case of efficient it is cost in case of responsive it is time.
So with this we are closing this lecture and now we are clear that how do we use supply chain
drivers for getting a particular kind of supply chain strategy. In our next class we will discuss
first lecture with respect to the data analytics practice and that is in the use of forecasting
techniques. Thank you very much
116
Lecture-10
Demand Forecasting in a Supply Chain
Welcome back, so far we have discussed about supply chain strategy and at different stages
of our discussions in previous classes. We discussed that how data analytics will help us in
taking our decisions in a better way. So now we are moving actually into the use of data
analytics. We are starting with forecasting that how you can have better forecasting in your
supply chain. Now forecasting is a very important tool because rest of the planning of your
entire supply chain depends on the forecasting.
So now we will see that how forecasting is going to help us in achieving the results and these
are the contents of our discussions.
117
Now as I said that forecasting is the basis for all strategic and planning decisions in a supply
chain. It is useful even in the push phase or in the pull phase of the supply chain process. And
there are uses of forecasting in all activities of your value chain either in case of production,
in case of marketing, in case of financing, in case of human resource and we know that the
current environment. All these things are interrelated.
You have a very close coordination if you don't have you will not be able to achieve the
desired. So it is very important to have it close coordination between production, marketing,
finance, HR etc. And then only you will have a proper outcome of your organisation and
forecasting actually provides input for all these activities in the organisation. So forecasting is
a very important primary activity for the development of our strategic and planning process
for the supply chain.
118
Now as we discussed if you remember in our last lecture we had a comparison of low implied
demand uncertainty and high implied demand uncertainty. In case of low implied uncertainty
the forecasting errors were just 10%. And in case of high uncertainties these errors were from
40% to 100%. So what I am trying to say that you will always have some kind of errors.
Forecasts are always wrong.
It means this is a very harsh statement to write but what I am trying to say that you do not
have the exact value of the forecast, these are the projections, these are just the projection.
And with the help of better tools, with the help of more sophisticated models we are trying to
reduce this level of error. And therefore in the development of forecast we calculate 2 things.
One, is the expected value that what is the expected forecast, what is the expected demand of
this product.
And at the same time what is the level of error associated with that level of demand. So that
you know that my forecast has some kind of weather three percent 4%, 5%, 10%. So that is
how you can take better decisions if you know both these things. So whenever we are going
to calculate the forecast we will calculate both these things. Then another important thing
related to forecast, because forecast can be done for different time horizons.
You can do forecast for your immediate requirement means maybe what will be the weather
tomorrow, so probably you can forecast this with more prediction, with more accuracy and
when I say what will be the weather after one year or after one month, after 6 months. So
119
these forecast will have less amount of accuracy, less precision. So the forecast for long term
are less accurate and forecast for the short term are more important.
So what I am trying to say that time horizon is very important in case of forecast time horizon
is more important normally the forecasting techniques can be classified on the basis of time
horizon. You have long term time horizon where we do forecasting for more than 3-4 years
and in that case you have different types of planning approaches and probably only few
decisions we take based on those long term forecasts like development of new facilities.
This type of issues can be handled on the basis of long term planning horizon. Because you
know that in South Asia the market is going to come up, so you want to locate a factory in
that area, so it is a long term forecasting decision. But then there is a intermediate time
horizon where you can take decision for around 6 months to 1 year time horizon and here
also you take decisions which are strategic in nature.
And then you have time horizon where you take decision from one day to 3 months time
period. And when you take decision for such a short time period these are not very strategic
decisions, these are more operational decision, you have developed a Canvas you have
developed a plan and now you are just execute in that plan based on this short of forecast
which is for the immediate uses.
And most of the time with the help of available tools with the help of available data analytics
techniques we get very accurate results for the short term forecast, probably long term
forecast results may not be that much accurate. I discuss the example of Bandra Worli Sea
Link in my previous class where long term forecast was the at around 12 lakh cars will be
running on that bridge per day.
But we saw that how imperfect, how inaccurate that forecast was, now the current data says
that only 45000 cars are crossing that bridge on a particular day, so now you have more
accurate data available with you and with the help of this more accurate data you can manage
the short-term issues, the immediate issues coming up on that bridge. So the forecasting is
very important with respect to time horizon that which time horizon you are taking.
120
And accordingly, you need to take into account a particular type of forecasting technique.
Then aggregate forecast, you see in our last class we discuss that number of SKUs are
increasing, number of distribution channels are increasing, so know if I have more number of
distribution channels, if I have more number of SKUs in my supply chain, in that case my
demand is distributed over those number of channels on those number of SKUs.
I am having a car, I am an automobile company and I have 5 different types of variants of this
car. So now overall demand of car that is X is now distributed in five different variants and
this I am taking that the demand is uniformly distributed for all these 5 variants. Now in each
variant you have some amount of imperfection with respect to forecast. So when I am talking
of disaggregated demand the inaccuracy the procedure is less.
But when I am talking only for car and I am taking the decision only for this X, I will have a
better forecast, a forecast with less error. So sometime I need to plan for each component then
I have to go for different different SKUs, but sometime we can consolidate the demand of
different SKUs and we can take aggregate demand and then forecast, so it will have less
amount of forecasting error. So these are another important characteristic of forecasting.
121
Now coming to forecasting methods what type of forecasting methods are available to us
because we have different types of forecasting methods which can be used depending upon
different type of time horizon. The forecasting methods are primarily classified into two
broad categories and these broad categories are one is qualitative and another is quantitative.
These are the two broad categories of forecasting methods. So qualitative methods are very
subjective, you take the interviews, you have certain qualitative research tools and these are
based on purely opinions of some experts, but sometime when we don't have any kind of data
we have to rely on this qualitative method. But whenever we have data we go with
quantitative methods when data is there these qualitative are purely opinion based these are
purely judgemental based.
122
But quantitative forecasting methods are data based. So now you can have different type of
quantitative methods and I have listed few here you have the most popular method of
quantitative forecasting that is the time series analysis. Time series analysis is the most
popular way of forecasting and time series can be used for immediate requirement as well as
for intermediate requirement. Qualitative methods are mostly used for long term forecasting.
If I am putting the time horizon for these things, so qualitative methods for the long term
forecasting and quantitative methods for the immediate and intermediate, so far these two
time horizons may be you can say up to 1 year we can use quantitative methods and beyond
1 year we go with the help of qualitative methods. In quantitative time series methods are
most important methods, the widely used methods and you have static time series methods.
And you have adaptive time series methods. So now in our analytics course we will see the
use of adaptive methods that how the model itself takes care of new data, the most recent data
and how the model itself evolves, the formula itself evolves over a period of time. So that you
have the best of the best forecasting. So time series where we have the lot of historical data
available with us and with the help of that data we do this type of analysis.
So use of data become very important in the development of time series models what type of
adaptive method is useful that is also very important because each set of data has its own
characteristics, so we need to understand that what type of characteristics are exhibited by a
particular set of data and accordingly we need to develop methods, models to forecast. The
second type of quantitative method is the regression analysis which is very popular name or
in our language of forecasting we call it causal analysis, cause effect analysis.
So cause effect analysis is basically you identify some kind of independent factors, you
identify some independent factors which are affecting the demand and then you develop
mathematical relationship between those independent factors and demand and now from
some independent sources you know that how the values of those independent factors will
change over a period of time.
And then you substitute those values and get the value of demand corresponding to changing
values of those independent factors, so this is the causal relationship and as I mentioned that
it is the relationship between demand and some other factors to develop the forecast, so what
123
I am trying to say that in causal you identify certain factors, these factors can be X1, X2, X3.
X1 can be the income level, X 2 can be the advertising expenses and X3 can be the let us say
our cost of the product.
Now these are the three independent factors which can affect the demand of my product, so
my products demand is being represented by Y, Y is the demand of the product and I try to
develop a relationship of this nature a1 + b1X1 + b2 X 2 + b3X3, so developing this kind of
relationship that how these factors are going to affect the demand of my product, that is
causal analysis.
So in this case, I am interested in knowing the values of a1,b1,b2 and b3. Once I determine
the value of a1, b1 and b2, b3 it means I have developed the relationship and now from some
independent sources I will determine the values of X1, X2 and X3 will substitute that values
in this relation and correspondingly for that period I will determine the demand of the
product.
So this is also a very interesting predictive method for demand forecasting. You have
predicted from some other sources the X1, X2 and X3. The interesting part is that you have
used some pass data to determine this relationship, but you are using future data future data
of X1, X2 and X3 for determining values of Y. So it is a kind of more interesting important
method.
And as we go ahead with the help of our computational tools we keep on changing the values
of a1, b1, b2, b3, so that our relationship becomes more dynamic, it becomes more adaptive
to changing values of Ys and Xs. This relationship I have shown as a linear type of
relationship but during our course of discussion we will see that this relationship can be
nonlinear also.
So as we go for more and more complicated models we will find any kind of general
relationship which is prevalent in between the demand and these other independent factor.
Then we have other quantitative type of tool which is simulation tool. Here in case we are not
able to develop a proper mathematical model for our demand forecasting purpose, so in this
case we can go for the simulation exercises also.
124
And this is also very interesting way of forecasting that you have more future data to use for
predict, you use lot of random future data and with the help of that random future data you
have some kind of projections that is also possible. So we will see we use in our course the
time series methods, causal methods and simulation methods for the forecasting purposes.
Now when we are talking of, so now the focus is only on this quantitative part of the
forecasting, we are not going to touch the qualitative forecasting part in our discussion so we
will focus only the quantitative part. Now when we are talking of quantitative forecasting
method, as I am saying that the data which we have it has different types of characteristics
and we need to identify those characteristics and accordingly we will use appropriate method
of forecasting.
Now in the any kind of data there are two types of components, one is systematic component
which is represented as S and another is the random component that is R. Now with the help
of our models, with the help of our mathematical formulas we can determine the systematic
component. The random component as the name suggests is random in nature and our
limitation is there that we cannot predict, we cannot determine with the use of our analytical
tools the values of these random component.
So these random components are attributed to watch forecasting error and now the duty is
that we should try to use good model, so that we can find the values of the systematic
component more accurately. So that the errors are only because of the random component,
125
but if because of our casual approach, because of our limitation if we are not able to use good
methods to determine systematic component.
So you can understand there are already certain errors because of random component which
you cannot determine and some errors may come because of your wrong choice of method of
knowing the systematic component so error will further magnify. So the first important thing
is use appropriate method for determining the systematic component. Now what let me give
you some idea about the systematic component and random component?
Like in case of a product when I say that demand is almost constant so this is the time and
this is the level of demand and when I am seeing the demand is almost constant so this is how
you represent the demand, but actually demand is not this straight line, this demand is
actually fluctuating around this straight line. Demand is actually fluctuating around this
straight line.
So sometime it is going above the straight line, sometime it is coming down this straight line.
So these variation around this straight line these are the random components. At this I cannot
determine that how much randomness is there. The straight line is the systematic component
here, so this I can determine if it is a horizontal, so if it is on January 2017 it is at a particular
level let us it is 50 units.
And I say the demand is horizontal, so in May 2017 it will be projected as 50 only. But since
there are certain random components, so because of those random components in May 2017
126
it can be either 55 or it can be either 45 also. So these are because of the random component
in the demand data. So we have to clear, we need to understand that how much demand is
coming because of systematic component and how much demand is coming because of the
random component.
Now in systematic component there are three types of characteristics, you have level, you
have trend and you have seasonality. These are the three types of characteristics which are
present in the systematic component. This example where I am showing a horizontal line, this
example is the example of level data that you have almost horizontal type of demand data.
What is the meaning of deseasonalized demand that we will discuss slightly later. Then you
have the second type of data where you have trend. Now trend is this type of either you can
have positive trend or you have negative trend. So when you are having a continuous increase
in the demand of your product which is normally during the growth stage or the launching
stage of the product.
So when the product is in the growth stage you have a positive trend demand is continuously
increasing over a period of time, so you have this type of systematic component. When the
project is in the decline stage at that time you have this type of systematic component where
the trend is negative, demand is decreasing over a period of time continuously. But again it is
not this type of straight lines, straight line increasing, straight line decreasing.
Actually the data is like this and similarly like this, you have systematic component and you
have some type of random component. But just for understanding the systematic component I
am giving only the straight lines. Then you have another type of component in the systematic
data that is the seasonality. Now there are products which exhibit this behaviour of
seasonality, products have a particular period, a particular season where the demand increases
to a peak level.
And in rest of the period demand is at very low level. So now the demand is like this and a
particular season comes and demand increases to every high level and then it is like this, so
you have winters and when in winters the demand of woollen clothes increases to a very high
level, you have the period of Diwali demand of crackers increase to a very high level and so
127
on you have lot of seasonal product and in the seasonal products demand increases to a very
high level.
So now your demand data may exhibit this type of seasonality behaviour, that every time a
particular period comes and demand increases to a very high level and again it is not this type
of straight curves there will be actually this type of fluctuations, so these fluctuations again
we will attribute to the random component and we need to see that how far we are able to
filter random component from the systematic component to have a better modelling better
exercise related to forecasting.
So we saw that horizontal trend and seasonality. Now in level data we have written the word
current sesonalized demand. Now desesonalized means that when I remove seasonality factor
that demand is increasing by 30% in a particular period, so that is a particular seasonal effect,
so when I remove that component of 30% so it should behave like this type of level data, this
type of horizontal data. So this is a case of seasonal demand.
And when I remove seasonality I get this type of level data and when I have the increasing or
decreasing trend then I have a trend data and now for level and trend to give you an example
we can again take the help of product life cycle. In product life cycle we have the situation
where you introduce a product and after that a period of growth comes during the period of
growth you have actually increasing trend.
Then the period of maturity comes and during the period of maturity you have the level data
and then phase of decline comes and during the phase of decline you have negative trend. So
therefore you can very well relate the three systematic components level positive trend and
negative trend with the help of this product life cycle where positive trend growth period,
level data is the maturity period and the decline trend, the negative trend is 1 the product is
coming out of the market.
So this understanding will help us initially to screen the method which type of method. If I
am in the growth stage so those models which are helpful, if my product is in the growth
stage I know as a marketer that my product is right now in the growth stage, so as a marketer
when I am going that my product is in the growth stage. So I will select a method which is
helpful in dealing the positive trend.
128
I should not select a method which deals with level data or which deals with negative trend.
So this understanding of product life cycle with the systematic component of my demand data
will help me to initially screen my method. If I am in my maturity period of demand data in
my product life cycle, so I should not select a method which is useful for the trend kind of
components. I should select a method which can handle the level data.
So this way I can filter out my models, I can filter out my methods and this initial knowledge
of various components which are available in my forecasting data will help me a lot in
selecting the right type of methods for the forecasting purpose and this as I mentioned the
expected component is the systematic component, the random component that is the
deviation from the systematic components.
So all the fluctuations which I have mark these are the results of those random component
and then another thing is the forecasting error that is the difference between the forecasted
value and the actual demand and we always use this forecasting error to update our
forecasting models and that is the beauty of adaptive forecasting system in time series we
have the adaptive forecasting system where our model is continuously updated with the help
of most recent data.
So we now have understood the usefulness of forecasting, the various methods of forecasting,
the relationship of those methods with respect to time horizon and then we started discussions
about the time series analysis we just saw the various components which are present in my
time series data and with the help of the example of product life cycle we related the
systematic components, the level components, the trend components and the seasonality
component in my product life cycle data. Thank you very much.
129
Lecture-11
Bullwhip Effect and Time Series Analytics
So, welcome participants, we are already discussing the role of analytics in the supply chain, and
as part of the discussion, in our last session discussed the role of demand forecasting in the
supply chain. And now we are moving to discuss some of the quantitative techniques which are
the part of analytics to discuss the forecasting in supply chain environment. As, we have
discussed in last many sessions that forecasting is one of the primary building block for supply
chain decisions.
And for that purpose it is very important to have a good forecast. We already know that
forecasting is not the 100% accurate. There are always some errors involved, because we are
predicting for the future, but how far we can predict correctly, how far we can minimise that
error, that is what we are looking in a supply chain environment. Because, one of the very
important factor which we have discussed in our earlier sessions also.
And, it is worthwhile to mention that again that is this bullwhip effect. This is one of the
important you can say parasite, to the profit of a supply chain, because of poor forecasting,
130
because of lack of trust in a supply chain, we do individual forecasting at different stages, and as
we move from the customer side to the manufacturer side, you keep on pilling the inventories,
and as a result of these inventories, the profit of your supply chain goes down.
So, now with the help of forecasting and forecasting particularly in a supply chain is to be the
collaborative forecasting. We have different individuals, manufacturers are there, wholesalers are
there, retailers are there. So these are the individuals, and if I talk of country like India, where we
have large number of supply chains which are unorganised. If I am a Maruti, if I am Hero, if I
am Amazon, so, I have an organised supply chain.
But, in most of the cases we have highly unorganised supply chains, and, in those cases, the lack
of trust, the lack of information sharing between various entities, result into this kind of bullwhip
effect and because of this as I mentioned the profits of your supply chain goes down. So,
therefore it is very important to understand that we do forecasting where you can have one
forecast for all the partners, all the entities, all the actors of your supply chain.
And, that is what we are going to see that what are those techniques through which we can do
this type of collaborative forecasting, and we can minimize these types of bullwhip effect,
particularly the role of IT is worth mentioning in these types of activities, because of IT
involvement, information technologies involvement, I am ready to share the information from
the retailers and to the manufacturers and at the same point of time.
131
So, we will see that what are those techniques, now for starting because what are the different
types of components of time series data, what are the different methodologies which are
available we have discussed in our last session. Now we move ahead with some of the
techniques which are available for the time series analysis and for doing the forecasting in
collaborative environment. Now we have this data available with us on the screen.
This is the sales data and for 2014 to 2016, we have this data available. And now with the help of
this data we will see that how can be forecast. We have discussed in our previous session, that
time series analysis is one of the most important type of forecasting technique which is available
to us. Now in that particular technique, we have also discussed in the last lecture that there are
two types of you can say components of your demand data.
132
One component is the level component, and another component is the random component. Now
this random component we attribute to the forecasting error, and level component is something
which can be determined, and random component cannot be determined, now it is important for
all the students to understand that in all forecasting methods, we are only looking whatever type
of model we are going to discuss.
Because, slowly and slowly we will start with a very simple model and then we will move to the
complex models, we are only determining the level components, because this is something which
we can determine. The random component cannot be determined and therefore this is attributed
to the forecasting error. So, you have determined that the forecasting for second quarter of 2017
will be let say 42000.
This is what you have forecasted, but the actual demand happens to be 42500. So, this difference
of forecasted value and the actual demand 500 is attributed to the random component. So, we
will not be able to actually forecast, what is going to be the actual demand. So, that is a limitation
of the forecasting you can say, now let us try to see that how in the best possible manner we can
determine the level components. Now, for that purpose let me start with the simplest method of
time series.
133
And this simple method is known as moving average method, this is the simplest method of time
series. Now, the data is here, data is here for past many periods. So we collect this type of data
and with the help of this past data. We will try to predict the feature and when I am saying the
moving average method which is the building block of time series forecasting. In that case I take
there are further a category of moving average method, which is known as simple moving
average.
This is simple moving average and in this simple moving average, moving average means some
period I am taking. So let us say if, I take a moving average period of N = 3 that is the period of
moving average. So what I am going to do for the forecasting of the next period I will take the
average of demand of most recent three periods. So here most recent three periods are 13000 +
32000.
And, the latest period is first quarter of 2017 that is 41000. So, I will take these three most recent
periods and I will take the average of these three periods and that will be the forecast of in fact
this is the forecast of second quarter of 2017. So, this way I will calculate the forecast for the
second quarter. If I am moving average period is of 3. Now for a better forecast more accurate
forecast I will, like to increase the period of moving average.
134
If I take N = 4 and for calculating the same forecast for the second period of 2017, I will take the
average of most recent 4 periods that is in this case second quarter of 2016, third quarter of 2016,
fourth quarter of 2016, and first quarter of 2017. So, If I am taking the average of these four
recent periods it means my period of moving average N = 4. So, to have a better forecast I need
to have higher value of N.
And that is the simplest method of time series forecasting where I am taking simply the average
of most recent periods, the method is very simple. Therefore, it has very limited applicability,
wherever we have simple method it is various categorical statement I am giving you that the
simple methods have limited applicability. So, we need to see that what are the limitations of this
particular method, one limitation is that.
Now you are seeing in this where N is 3, I am taking 13000, 32000, and 41000. Now here just by
observation, you can see that there is a kind of seasonality in this data but I have not consider
that seasonality in this data, and I have given equal weight, to all 3 past periods. One logic says
that, I should give maximum weight to the most recent period, but, in this particular method.
I am giving equal weight to all 3 past periods. So, this is one limitation of this method, and,
therefore I am not able to give due regard or due weight to the happenings or to the conditions of
the most recent period, and I am giving equal weightage, so that is one limitation. The second
limitation is, as I am saying for better forecast we need to have higher values of N.
Now, for higher values of N, I need huge amount of past data, you talk of simple automobile
company, one automobile company making a car deals in more than 2000 different types of a
smaller components, now when you want to forecast for those 2000 different types of
components you need to have data for all those components. Take an example of company like
Amazon.
Supplying more than 6 billion items SKUs. Now you can imagine the amount of data, these
companies need to manage for the forecasting to use this simple moving average method with
higher values of N. So therefore it is very costly, time consuming affair to have higher values of
135
N, because of better forecast. So these are the two very important limiting factor for this
particular method. So, now let us go for slightly improvement on the simple moving average
method.
And then we have other method which is known as weighted moving average method. So, now
let us see what is this weighted moving average method, and here in this weighted moving
average method we give more weight to the most recent period.
Again take an example where the period of moving average is 3, in this case also the period of
moving average is 3 and is equals to 3, but, here I give more weightage or maximum weightage
to the most recent period. So, the first quarter of 2000. If I am forecasting for the second quarter
of 2017, so I will give maximum weight to the first quarter of 2017 and my weights will
decrease.
As I go ahead from the most recent period to the furthest period. So, I have weights like W 1
which I will give to the demand of first quarter of 2017 + W 2 that is the weight given to the
fourth quarter of 2016 + W3 which is the weight given to the third quarter of 2016. And in this
way we will determine the forecast, but I have to be careful that my W1 is greater than or equal to
W2, we should be greater than or equal to W3.
136
And, another condition should be that W1, W2 and W3 sum of these 3 weights should be equal to
1. So, by this let say I give W1 as 0.5, W2 is 0.3, and W3 as 0.2. And then if I substitute these
values Ws these values of weights in this particular equation I get a forecast where I am
overcoming one of the limitation of my earlier method, that I am giving more weights to my
most recent demand and my weights are decreasing, as I am moving away from first period to
the further periods.
Again, the issue of better forecast will remain here in this case also, and, I can increase, by
improving the values of N. So if I keep N equals to 4, so, I will have W 1, W2, W3, W4 and
accordingly I can follow the same rule to have weights differently. There is no set rule, to give
the values of these weights, normally experts on their own, give the weights and with the
experienced, with the exposure in a particular field.
With the exposure of a particular market, with the exposure of the external environment, experts
are good enough to give these weights. Though there are some types of fuzzy mathematics
systems are available to decide these weights when we invite some experts, and this experts may
be in a group, and can think of what should be the appropriate weight, but here also there is a
problem.
In this method, we have overcome one of the limitation that we are giving maximum weightage
to the most recent period, and the weights are decreasing. But here also you have two important
issues, one there may be some reason, there may be a temporary reason because of which
demand of a particular product in a particular duration has increased.
Recently in the October-November month of 2016 there was a problem around deepavali time in
NCR reason, when a smog was there, and the result of that smog the demand of air filters, the
demand of mask has increased drastically in that area. That is what a regular feature, that was a
very temporary phenomena but, due to that temporary phenomena that demand has increased
drastically for those products. But now if I am following the time series method of forecasting.
137
And, if that October, November data is there, where the demand has increased all of a sudden to
a very high level, and I give maximum weightage to that period. So, what will happen that, the
same will carry to the next period also, but actually in the next period that phenomena of a smog
is not going to be there, and therefore demand will going to be on a very low level. So, in some
this exceptional situations I should know that I should not be assigning the maximum weightage
to the most recent period.
Because of the temporary nature of particular phenomena and therefore, we need improvement
over this weighted moving average method also, that these types of temporary phenomena can
also be handled. And therefore we move to the next category of time series analysis, where we
can very well handle these types of temporary phenomena and these category of time series
methods are known as exponential smoothing methods.
That there are certain reasons which may be permanent like, if I talk of government employees in
India. So, there was sixth pay commission and now over last 2 years employees are moving
towards seventh pay commission and as a result of that the purchasing power of employees,
government employees are increasing and therefore the demands of certain products may
increase, but this phenomena is a permanent phenomena.
So, I need to incorporate this phenomenon in my forecasting model. So, there are certain things
which are of temporary nature like the smog problem in NCR, I should not include those things
in my forecasting method, and there are certain permanent issues like increasing the purchasing
power of your customers. And that must include in my forecasting model. So, as a forecasting
manager, as a supply chain manger I must understand that what is to be included permanently in
my forecasting methods.
And what is to be excluded in my forecasting method, and the next method which we are going
to discuss the exponential smoothing methods will take care that particular aspect, that what is to
be included and what is to be excluded. Moving average methods are very simple calculation
wise and these methods have therefore limited applications as I discussed.
138
But, just to start the concept of the moving methods, just to start to understanding that how
methods can be adaptive. Because as you go to second quarter of 2017, you have new data
available and then with that new data if you have N equals to 3 again. So, that second quarter
sale, the first quarter sale, and the fourth quarter sale of 2016, these 3 sales will be used to predict
the sale for third quarter of 2017. So, this method is automatically an adaptive method.
Whatever is the latest data available with you, that latest data is used for forecasting for the
future. So, the method is continuously evolving, you do not have a static forecast for all the
periods, whatever new data, whatever latest data is available with you, you use that data for
moving ahead. So, that is the beauty of time series analysis, therefore the moving average
methods whether it is simplest type of time series analysis.
Help us to understand what is the meaning of adaptive forecasting. So, whatever is happening,
whatever changes are happening there, those changes you are including in your forecasting
method and continuously you are updating you forecast for next period.
So, now moving to the second category of time series analysis, this is the most popular type of
time series analysis which is known as exponential smoothing methods. Now as we discussed in
our earlier sessions, that our time series data may have different type of components, time series
data may have trend, time series data may have seasonality, time series data may have some kind
139
of cyclic fluctuations. So, I need to see that what type of components my time series data may
have.
And, therefore accordingly I will use my particular method of forecasting. But the philosophy of
exponential smoothing method is that just to elaborate that this is for the demand data, for some
of the periods. Now, you can see that this is not a straight line, this is a slightly zigzag line. Now
this zigzag line is around a particular dotted line, which is a straight line. So now in exponential
smoothing method.
I will try to smoothen, the fluctuations of the actual demand data so that it can be represented by
this dotted line that is what we are doing. So, we use different types of a smoothening constants,
in this exponential smoothing method and by using those constants we will try to smoothen the
fluctuations of actual demand data. So, that it can be represented by this dotted lines.
And, therefore, we need to understand which type of data we have and accordingly the number
of smoothening constants if there is no other component available in my data. So, I use only one
smoothening constant α alpha, when no other component is available, I use just for level
activities. I use two smoothening constants if I have trend also in my data and I use three
smoothening constants one for level, one for trend, and another γ gamma for smoothening the
seasonality component.
So, depending upon type of data I have, like this data very clearly shows the kind of seasonality I
have, the demand in the first quarter of each month, each year increases to very high level, like
here it is 34000, here it is 38000, and here it is 41000. And then in the second quarter of each
year like here it is 8000, here it is 10000 and here it is 12000 decreases to a low level. So, this
data just by observation without going for any kind of mathematics you can see that.
The kind of data we have represented here, is showing a type of seasonality into this. Now when
this types of seasonality is built in to the data, we need to use different type of a smoothening
models, when only trend is there we need to use different type of smoothening model and when
140
nothing is there only plane horizontal demand is there, and some kind of regular fluctuations are
there. So, only one type of smoothening constants are required that is α.
So, now let it may show you the representation of different types of curves which are possible
because of these different type of components which are present in the demand data and you will
appreciate that as my data will change the type of model will also be required to change. Now we
can have nine possible combinations let me clear the board first, so that we can understand it in a
better way.
Now this is the demand data where know this X axis represents this Time, Y axis represents the
Demand. So there is no trend, no seasonality. So, demand is represented as a horizontal straight
line with respect to time. Now you can have few more curves and these curves will help you to
understand different types of components. Now first is there is a linear trend, linear trend like
this over a period of time.
And then you can have multiple trend or the ratio trend, so which is represented like this way
where demand is increasing in a multiple manner, and in all these three curves figure number 1,
figure number 2, figure number 3. You have only trend component present, here there is no
trend, here is linear or additive, here it is ratio or multiplicative. Then come to figure number 4,
here there is no trend but some kind of seasonality is present.
141
That demand increases to a high value in a particular period but there is no trend. Then you have
demand where you have linear trend but linear seasonality also. So, here your curve is
represented like this. And here you have linear trend but ratio seasonality, so around this curve
you can think of this type of actual demand data. Then you have here I can write is the linear
seasonality, and then you have ratio seasonality. In ratio seasonality you have demand increasing
in season like here 34, 38, 41.
This type of increasing seasonality data is like a constant increasing data 4000, 3000 so it is more
like a linear seasonality this type of seasonality, but it is possible that it is here 30000, 34000
then here it is like 60000, then it is 70000 or 750000. So the demand may increase in a ratio term
also in the seasonal periods. So this is the ratio seasonality and when this ratio seasonality is
couple with the linear seasonality.
So you can think of the curve something like this. So which is eighth and then you have the most
extreme case where you have the ratio seasonality and ratio trend. So you will have the actual
curve around this like this. So this is the ninth figure. So in all you have these nine different
types of demand data sets which are possible and in our next session we will discuss the
exponential smoothening method for some of this model and rest of the models you can develop
on your own.
But we will see that how exponential is smoothening method can we used for some of the
models which we have discuss some of these different characteristics of demand data which are
possible. So there are two important duties which are there, one is to identify the characteristic of
your data that just by seeing it is almost impossible that what type of characteristics your data is
showing and once you identifying the characteristic then appropriately use that model which is
suitable for that particular case.
And the third thing will be the use of well use of α alpha, β beta, γ gamma as for the case may
be. So these are the two three important things which we will discuss in our next session. Thank
you very much.
142
Lecture-12
Experimental Smoothing Method of Forecasting
So, welcome back in the previous session we were discussing about time series analysis, and
we discussed one of the simplest method of time series that is moving average methods. We
discussed 2 methods of moving average, the simple moving average method and the weighted
moving average method. Then we will also discuss that limitations of these moving average
methods are responsible for their limited application into the real field of demand forecasting.
So, we have improved method which is known as exponential smoothing method, we also
discussed in our previous session different types of models incorporating various types of
trends, and various types of seasonality component in to the demand data. And, now we will
move further to see how we can use the exponential smoothing method for those different
types of components in our demand data.
For an example we have this sample data with us and we will go with the method of
exponential smoothing for this sample data.
143
Now, to understand the basic exponential smoothing method, where we have only horizontal
component in our demand data, and there is no other component in the demand data, what is
happening actually we are right now in the month of January, and February is approaching. In
the month of January we have some level or you can say the forecasted demand at on the
basis of their forecast, you are doing the forecast for the February.
Now, in the forecast of February there will be some errors and the actual demand will be D t.
Now, on the basis of this actual demand of for February we will determined the level of
February, and this level of February is the forecast of March. So, this is how the system will
work. Now, it is up to us, that how much uncertainties how much variations of a particular
period, we want to include in getting the forecast for the next period.
And therefore, if we see that there are certain changes in the external environment, which are
of permanent nature, we will incorporate more deviations in to our forecast and if you see
that there are certain temporary changes in the forecast, then we will not incorporate those
changes in to the forecasting model. And therefore this exponential smoothing method gives
us that flexibility that how much variations you want to include into the forecasting method.
Actually, as we have discussed in last two sessions, that we have 2 components of forecast,
one is level, and another is the random variation. So in the basic exponential method we are
only interested in determining the level component, and the random component cannot be
144
determined in viable mathematical model. So, whatever is the level of the current period that
is actually the forecast for the next period.
So, in the basic exponential method as I mentioned that the level of February is the forecast
of the March. So, if I say if I represent level wise S t, the current level I am representing St, for
this current level is actually the forecast for the next period, this is the forecast for the next
period ft,1. So, now how do we determine the current level that is the issue, because if I
determine in the current level that is the forecast for the next period.
Now, the current level actually, when I am taking this exponential smoothing method so,
what I am trying to do that I will take in to account some of the fluctuations of my the
demand, and for that purpose this is the current level and therefore S t-1 represents the previous
base, previous level, and some of the fluctuation of my current period, I will like to
incorporate. And if, this is the new demand - old base, St = St-1 + α.
This becomes my expression to calculate my updated base value or which I can also write as
α(Dt - St-1) or you can write it as αDt + (1- α)St-1. So, this becomes the expression for my new
base, St. And now let us see that how do we use this for our this type of data, this data is
there, and let us try to use this formula on this data. Here, we have a data of actual demand
for the first quarter of 2017 as 41000.
So, my current Dt is 41000. The forecast I am assuming, the forecast I am assuming for this
period when I was in the fourth quarter of 2016, I did a forecast of 40000 for the first quarter
145
of 2017. So, let us say my S t-1 is 40000, so, now if I determined, if I determine the base for
this period, if I determine the SI,2017, that is actually the forecast for second quarter of 2017.
And, please ensure that I am not taking consideration, any kind of trend seasonality in this
data, though this data exhibits some kind of seasonalities, but for the sake of example I just
considering that seasonality into this data right now. So, now I am determining S I,2017 which
is if I go by this expressions is αDI,2017 + (1-α)SIV,2016.
So, if I go with this expression I can calculate S I, 2017 and SI, 2017 is nothing but the forecast for
the second period of 2017. So, I get the forecast of second quarter of 2017 using this method.
Now here the importance is that what should be the value of this α, theoretically speaking the
value of α can lie between 0 to 1.
But, practically the common values of alpha which we used are from 0.05 to 0.30, these are
the common values of α is 0.05 to 0.30, but, it can be zero or it can 1 also. Now the meaning
of taking different value of alpha. Let us say if I take α equals to 0.1, if I take α equals to 0.1
what it means. If I take α equals to 0.1 the meaning of this alpha is that I am taking only 10%
of my current demand.
I am taking only 10% of my current demand, and I am discounting 90% deviations, and I am
90% of my previous base. So, that is the meaning of α equals to 0.1, this means that a smaller
values of α has more smoothening effect. If you have a smaller value of alpha, it gives to
more smoothening effect, and larger values of α should only be taken in that case when you
have a shifting base.
When you feel that there is a change which is of permanent nature, and you want to improve
that change in your forecasting model, then you should go for higher values of α. For an
example as we have in discussed in previous session also, that when a new pay commission is
coming and purchasing power of people are increasing. So, this is a permanent kind of
change, and as a result of that permanent change in you have purchasing power.
Your value of α met permanently shifted. So, in that case, you can go for higher values of α,
and when you take lower value of α you are having more smoothening effect. Now, if you
take for an example two extreme cases, you take extreme cases when α is equals to 0, and α
146
equals to 1. So, if you substitute in this equation in this case where α equals to 0, so, it means
your current base is equals to old base.
The meaning is that you do not want to include any deviation of the current period in to your
current base, like the issue of a smog during October and November in NCR area, that is a
temporary type of phenomena. So, because of that whatever fluctuations in demand has taken
place, you do not want include the fluctuations permanently in to your model and as a result
of that you will take the small values of α.
And, maybe you can take α equals to 0 in some cases, then α equals to 1. This will lead
αequals to 1 will lead to St = Dt. Now St equals to Dt, in this particular case you see, that you
have totally shifted your base, you have taken a new base, a new demand is your new base,
so, you do not want to take your old base at all into your consideration, so, it is a jumping
base type of scenario.
And, maybe in case of pay commissions, may be in case of rehabilitizations, large level of
rehabilitation or something of that sort whenever happens. So, you can have every high value
of α, so, these are the extreme cases, but, normally α lies between 0.05 to 0.30. So, 0.1, 0.2,
0.25, 0.15 these are very popular values of α. And on the basis of that you can do this
calculation, and you can get the new base.
For this period first quarter of 2017, and that will automatically become the forecast for the
next period. So, this is our simple method of basic exponential method. Now as we discussed
in the case of moving average method for better forecast you need to have long historical
data, now that requirement is reduced here, you only need data, you don't require this long
data with you. You only require data of just last 2 periods.
And, that current data helps you in getting the forecast, so, you are getting forecast with less
data and which is more accurate, which is more adaptive, which is more you can say
customised as per the situation whichever is happening in the market. Now once we have
understood the simplest form of exponential smoothing method. Now we go to different form
of smoothing method where we will see that how you can customise your model.
147
To suit the requirement of trend and seasonality component in to the demand data. And, here
we are using only single smoothening constant α, in those cases we may use more than one
smoothening constant because this smoothening constant is only smoothening the
fluctuations of your level data, but when you have trend. So, you require one smoothening
constant to smoothen the fluctuations of your base data.
And, one to smoothen the fluctuations of your trend data, when you have seasonality in to
your demand data, then you require 1 smoothening constant to smoothen the fluctuations of
your seasonality component also. So, depending upon the type of the characteristic of your
demand data, you will require as many number of a smoothening constant. So, now let us
move to second of smoothening method.
Where we have the trend also in our data, now trend can also be of two types, one is linear
trend or additive trend, and the second is ratio trend or multiplicative trend. These are the two
types of trends which are possible. The meaning I show you if I have the historical data with
me in that case, you can see this for this is period, this is column A, this is column B, and let
me have the data for some past periods.
Now, here I am starting with 20,22,26,27,29,31 like that and in this case we have 20, 24,29,
34, 39, 44. So you will see that in both these cases you have a trend. But, here in the first case
here the demand is increased by two units, then by 4 then by 1, then by 2, by 2, so, you are
having a kind of additive some constant figure or a fluctuating figure is being added in to the
demand of previous period.
148
So, it is more like a linear trend some almost constant thing is being added. Here demand
increase by 4, then 5, then 5, then 5, then again 5. So here the demand is increased in a more
ratio type of field that it is multiplying to the previous periods. So finally from 20 to 44 it has
just doubled. So that it is kind of ratio you are getting over period of time. So it is multiplying
effect and here it is additive effect.
So depending upon what type of trend you have you can make changes in your model which
we are going to discuss. Now going further into the calculation part of this since we have in
our component trend also and already that linear part is very much present. So two solve such
type of cases.
Let us see if I take the first data with us. So you have this trend component initially, there is
no trend and then you have the trend of +2, +4 you have +1, +2 and you have again +2. So
these are the trend data which you have. When we are the developing the forecast in this
particular case. So now the FII,2107 will be the updated base of the current period that is S I, 2017 +
the updated trend of current period TI,2017.
That will make the forecast for the next period. So I need to make updated base for the
current period and I have to make the updated trend of the current period and when I add both
this things. I will get the forecast for the next period. So if I generalised this relationship. So I
will say that Ft,1 is nothing but St + Tt that is the forecast for the next period is the updated
base for the current period and the updated trend for the current period.
149
Sometime it is possible but in case of trend data, I will like to forecast for two, three periods
ahead from the current period. So in that case if I want to forecast for M period ahead. So in
that case this value of trend is consider as constant value and then I will do like this F t,m = St +
mT. I am taking a particular case of linear trend that is why I am just adding up these things.
If it is a ratio trend then the multiplying effect will come into picture accordingly. Now let us
see how do we do that.
So first we need to calculate the updated base and then with the help of updated base we will
determine the updated trend also. So the updated base S t we already know how we determine
the αDt + (1-α) St-1 that is in the simple exponential method. Now with the trend it will
become αDt + (1-α)(St-1 + Tt-1) that is the previous forecast and that is the current demand.
And now I also need to do the updation in the trend data and for that purpose I need to
calculate the Tt and Tt is nothing but the difference of my current base and the previous base.
So Tt = St - St -1 + (1 - β)Tt-1, because I am taking the second coefficient of a smoothening of
β which is for the trend purpose (1 - β)T t-1. So this is the updated trend value and then finally
the forecast Ft,1 = St +Tt.
So this is how we will do the forecast when trend is available, the calculation of linear trend,
the additive trend, is then St-St-1 and if it is to be multiple trend, if it is to be ratio trend it will
be St /St-1, please be careful that if it is a ratio trend it will be S t /St-1. And, since I am
considering the case of linear trend it is St-St-1. And, this is the old trend, then I update the St +
Dt and I get the forecast for the next period.
150
And, if I want to use this trend value for getting the forecast for 2, 3, 4 periods ahead then this
expression will work, at St + mT, and now this model is ready, and now I can use this model
for getting my values for next period. Here as I discuss about α the same discussion apply for
the β also, the values of β also varies between 0 to 1. The popular values of β are from 0.05 to
0. 20, because the fluctuations in trend are not much.
So, we use lesser values of β and we want to discount maximum fluctuations of trend values
and it is very rare, that you use very extreme values of β. So, normally β values are less than
α values, but, since both these are smoothening constant, so, academically, theoretically their
values can vary 0 to 1. So, β can also 0 to 1, but, the popular values are from 0.05 to 0.20.
So, now if I apply this equation these two equations on this piece of data, so, you can say that
for seventh period, for F7, I need to apply S 6+T6. I need to calculate S6 and T6 and with the
help of S6 and T6, I can directly get the value of F7. And, S6 will come from this equation for
S6 I require S5 and for T6 I require T5, and when I use these expressions, I can directly get the
values of my required forecast for the next period.
Then, in case of ratio the only thing which I just told you, this value will change, this will
become St/St-1, in case of ratio, this calculation will change to S t x Tt. And, this calculation
will also change to St-1 x Tt-1. Rest of the model will remain as it is. There will not be any
change in any other component with this model. So, with this you can handle both this types
of trend, you can handle the linear trend, you can handle the ratio trend.
But, now just by seeing nobody can give you the answer whether this A has linear trend, or
whether B has this linear trend, you may think in your mind that how do you can say that here
you have 20 to 31, you can have a ratio of 1.5 something like that, here you are moving from
20 to 44 you have ratio of somewhere around 2.2. So, why cannot we apply a ratio model in
this case or why cannot we apply a linear model in this second case.
Because each time 24 to 29, 29 to 34, 34 to 39, 39 to 44, demand is increasing by very
constant value 5. So, why cannot we apply a linear model in second case and why cannot we
apply a ratio model in the first case, practically speaking, theoretically speaking I do not have
151
any answer for that, only my model will tell the answer, model means whether I am
calculating using St-St-1or St/St-1.
And, after determining the forecast I will calculate the forecasting errors and, whichever
model will give me minimum forecasting errors, that is suitable model for my particular case.
So, now our next topic of discussion is the forecasting error. So, that we can understand the
meaning of selection of different type of models. Without understanding the forecasting
model it is almost impossible to select the right kind of model.
Because, once you select the model the appropriateness of that model will also depend on
proper selection of values of α and β. So, all these things are very much you can say in a
family, the kind of model the selection of values, α and β and these models should produce
minimum forecasting error. So, in our next class we will discuss more about forecasting
error, and then we will also take a case of third type of component that is the seasonal
component in our demand data.
And, how to handle that seasonal component in the data so, that case will also take, and this
data will be useful in that case also, the because, here you can see we have the seasonality in
our demand data present, so, how to handle the seasonal component in the demand data, and,
then we can move to the most complex case where we have seasonality as well as trend, both
these things in our demand data. So, how to handle such type of cases, that we will see in our
next lecture, thank you very much.
152
Lecture-13
Measures of Forecasting Errors
Welcome back, we are already discussing various forecasting models for the supply chain
decision making. We have discussed so far, different types of time series method in which we
have studied the weighted moving average methods, and then we also realised that there are
certain problems with those moving average methods. And as a result of that we discussed about
the smoothing methods and we discussed exponential smoothing methods for helping us with the
problem of excessive data.
And how we can handle the fluctuations which are happening in the most recent periods. In our
earlier classes we discussed about simple exponential method, and now today we are going to
discuss a complex exponential smoothing method where we have some kind of seasonality.
and that seasonality we are assuming is of ratio seasonality type. Now in our earlier two cases we
have discussed the basic exponential smoothing model where there is no trend, no seasonality.
Then we also discussed a case of linear trend exponential smoothing model and now we are
153
moving to the third category of exponential smoothing models, that is the ratio seasonality
model.
Now to understand this model we have some hypothetical data available with us, and with the
help of this data we will see that how this ratio seasonality model can be operated. Now the point
of sales data available with us from January 2016 to July 2017. This is the actual demand, and in
our notations this actual demand is represented as Dt.
That is the actual demand which is shown in the second column of this table. Now as we have
discussed in our last two classes the demand is considered to be fluctuating around a base value.
So in case of a seasonal factor also there is a kind of seasonality effect, and actually the demand
is fluctuating around these base values. So we will determine these base values which is rather a
more smoothened curved.
And actually the demand is represented by this zigzag curve which is the fluctuation around this
base value. So the third column of this table represents that base value. Now for the purpose of
our understanding we have assumed we have just assumed the base value of December 2016 to
be 30 and the further calculation is shown here. And, since we are talking of 2017 only so, there
is no need for the base value of this period.
154
Now, since we already have assumed that is type of demand data has some kind of seasonality.
So in this particular case we are aware that seasonality is there. So whenever seasonality is there.
We need to de seasonalize the demand and for de-seasonalisation purpose, we need to divide the
demand of individual periods by the average demand of 2016. So if you see the average demand
of 2016 is 24.07 that is the 24.07 that we get by summing the demand from January 2016 to
December 2016 divided by 12 and you get the average demand as 24.07.
Then for de-seasonalisation purpose we need to divide the demand of individual period by this
average demand, for the case of January 2016, we divided 19.36 by this 19.36 which is the actual
demand of January, by the average demand of 2016 and you get the seasonal index, and by doing
this calculation you have the seasonal index of all the periods 2016.
155
For each month of 2016, you have the seasonal index, so the calculation is very simple 19.36
divided by the average demand of 2016, that is 24.07, you get 0.804, similarly 25.45 divided by
the average demand you get 1.057, and so on for rest of the periods of 2016, that 19.73 divided
by average demand of 2016, and you get the seasonal index. So now when you multiply, the
actual demand of these periods, by these seasonal index.
You get the deseasonalized demands and this deseasonalized demand will be used for our
calculation to get the forecast, for any period of 2017, but here comes a question, in our mind
that here I know, that I am going to discuss a ratio seasonality model. So I know that the demand
is seasonal in this case, but if you see this data just 19.36, 25.45, and so on, for entire 2016.
It is very difficult from naked eye, that you can say that this is a seasonal demand data, it is
almost impossible for any human being just by observing this data, you can say that this is a
seasonal demand data. So, how do you need to have a deseasonalized demand or what is the need
of calculation of signal index, since, I am saying in this particular case you know that we need to
calculate seasonal index.
We need to determine deseasonalized demand, but actually in the reality because of our inability
just by observing if it is a very clear pattern like if it is 19, 19, 19, 19 and then in the month of
May, June, July demand increases to 40, 45, 50 and then again demand comes to 20, 25 like that.
156
So you can even by naked eye say that demand has kind of seasonality factor it increases to a
high level in a particular period and then it decreases.
But here in this particular case, the data 19, 25, 19, 21, 20, 25, 23, 28, 26, 25 the data is so
closely linked for each consecutive period. That the seasonality is not visible just by observation.
So, how do we decide, whether we are doing a right kind of thing. And the second thing here I
am saying that we are discussing the ratio seasonality model. Now how do I say that this is ratio
seasonality model.
Why not it is a linear seasonality model. That is also a question, so, for this purpose we must
know, whether we are applying the correct model or not. And, for that purpose, we need to know
about the forecasting errors. Forecasting errors will help us to determine whether we are using
the right model or not. And, for that purpose we will discuss briefly not in much detail about
some of the possible errors, which are there in the forecasting and what are the measures we used
for the forecasting errors.
So in case of forecasting errors we all know that forecast are the tentative values for the future
and the actual demand like this and the forecast which we are not given. So let us say for the
January 2016 if I am forecasting in the December 2015. So forecast can be lesser than 21 but the
157
actual demand is 19.36. So, the difference of actual demand and forecasted value is known as the
forecasting error.
So, forecasting error let us represents for a particular period as e t. et is the difference of demand
and forecast for the same period, et = dt - ft. So the difference of demand value and the forecasted
value for a particular period is known as forecasting error. This is very simple, no complication.
But this major alone is of limited use. This definition of forecasting error has limited application.
So we have developed certain other measures of forecasting error.
And, now we are going to discuss those other measures of the forecasting errors. The other
measures of forecasting error that is the simplest one and the average error is defined as the
average of forecasting error like in this particular case, you have forecasting error for January,
February, March, April, may, June, July, and let us say if in the month of December, I am
calculating what is my average forecasting error. So that is the average of forecasting error for N
number of period.
So if I am doing forecast for N number of periods. So the average for errors for these N number
of period, is my average forecasting error. It is normally assumed or it should be for a large
sample size, the forecasting error should be 0 or should be very close to 0. But, it is also possible
we all know, that for one period there is a problem with average, we all know, but it is possible
that for one period.
The error is +10, and for other period the forecasting error is -10. So, if I take the average of
these two forecasting errors, so +10 and -10, the average error will be 0. But, in both the cases, in
both these periods, the forecasting error is individually very high, but the average error will be 0.
What I am trying to say that eJan is +10 and eFeb is -10.
If I am taking the average of these two periods, that will be (e- Jan + e-Feb)/2 and (+10-10)/2 will be
0. So, by definition it looks that average error is 0, so I am having a very good forecast. But,
actually because of different signs of these forecasting errors, I am getting the 0 forecasting
158
error. But, nevertheless in large sample it is assumed that there will be certain plus values and
there will be some negative values.
And with the positive and negative values you should have a 0 average error. So, that is what we
expect. Now as we have discussed that there is a natural limitation of average value, and this not
only in the case of forecasting error. The problem of average is always there, that because of the
sign you will have this type of issue. So, there is a second measure of forecasting error, which is
known as mean absolute deviation. Very popularly known as MAD.
MAD is a very popular measure of forecasting error to eliminate the problem of the sign in the
forecasting error. Now, we take the absolute values of error to calculate the mean absolute
deviation. So, it remains like that only, that you are calculating the average. But, the average is of
absolute values. You take the absolute values and then you determine the average of that. So,
whether it is +10 or -10.
If I am calculating mean absolute deviation it will be absolute values and you get the 10 as mean
absolute deviation. So, for same values we just saw that the average error was 0. But, mean
absolute deviation is 10. So, mean absolute deviation is a very popular measure of forecasting
error. And this mean absolute deviation helps us in determining, whether we have taken it
correct forecasting model or not. Whether we are considering ratio seasonality it is appropriate or
not.
If my MAD values are higher than my model is not appropriate, some other model may be more
suitable in my data. So, with different types of model, I will calculate the MAD value and
whichever model will give me minimum values of MAD. I will select that model for my
forecasting purpose. And at the same time let me also tell you about smoothing constants.
In our last classes we discussed the use of smoothing constant. We will use 3 type of smoothing
constants alpha, beta and gamma. Alpha for smoothing the fluctuations of the base, beta for
smoothing the fluctuations of trend, and gamma for smoothing the fluctuations of the seasonality
159
components. So, what should be the correct value or what should be the most appropriate vales
of alpha, beta and gamma.
That also can be checked, that also can be realised using this mean absolute deviation, because
the selection or to take a particular value of alpha, beta, gamma, there is no mathematical
calculation, there is no formula just by observation, just by practice we take alpha, beta, gamma.
The thumb rule, we have already discussed in our earlier class, that the smaller values of alpha,
beta, gamma. The values of alpha, beta, gamma we discussed varies from 0 to 1.
The smaller values of alpha, beta, gamma these values have more smoothing effect. But, in case
there is a quantum jump and we want to have a different base or a different level of trend or a
different label of seasonality, then you take higher values of alpha, beta, gamma. The smaller
values of alpha, beta, gamma gives more smoothing effect. So, that is the thumb rule. But, as we
know the alpha, beta, gamma normally should be between to 0.05 to 0.30.
Though values can lie from 0 to 1. But in most practical purposes alpha, beta, gamma varies
from 0.05 to 0.30. So, what should be appropriate value of alpha, beta, gamma in my case for
that also we use, we take the help of mean absolute valuation. And, by putting different values of
alpha, beta, gamma you calculate the MAD values and whichever set of value give you
minimum values of MAD.
That is your selected value of alpha, beta, gamma for a particular set of data for a particular base
of forecasting. Then there is one more method also, one more measure of forecasting error also
available to us. That is mean square error and it is abbreviated as MSE-Mean Square Error. In
mean square error you take the square of forecasting error and then you take the average of those
forecasting square.
The purpose of MSE, because of the square factor the purpose of MSE to penalise the very high
forecasting errors. So, in case because of square effect, if there is a higher forecasting error, by a
square of that term it becomes even higher. And, it immediately comes to know that there is a
160
particular set, there is a particular case of data, where the forecasting error is very high. So, this
gets you additional information other than MAD.
But, not very popularly in use, but can be used combination with MAD. Then one more
forecasting error measure is there, that is known as mean absolute percentage error. With stands
for MAPE- Mean Absolute Percentage Error. Now, mean absolute percentage error is one more
additional measure of forecasting error. In which what we are doing, we see that how much the
forecasting error is an offset of our actual demand.
And, that is determined by this method, and then you take the average. So, this is one more
measure of forecasting error. But, out of these four the most popular is mean absolute deviation,
and with the help of this mean absolute deviation we will determine that which method to use
and which method is more suitable, what are the values of alpha, beta, gamma which are suitable
for our particular case.
So, now we are not doing all these calculation for this given set of data. We assume that ratio
seasonality is best suited for this particular case. But otherwise in all practical cases we need to
do the calculation of mean absolute deviation. And, once we have done this calculation of mean
absolute deviation. Then only we can determine that this model is suitable or these values are
suitable. Now, let us see once we have assumed that ratio seasonality is best suited for this
particular case.
161
So, then we go for deseasonalization. As we have already discussed by dividing the actual
demand, by the average demand of a particular period, and then we determine the seasonal index
of different periods. Once we have determined the signal index and since it is a ratio seasonality
model under the case of exponential smoothing, so, we need to see how to determine the
forecast, and since it is a ratio seasonality case.
So, now we will go in the reverse direction like I have the actual demand up to July 17, and I
want to determine what will be the forecast for august 2017.
162
So, now in case of a ratio seasonality model, in case of ratio seasonality when I want to calculate
the forecast for August 2017, I will use the most recent base. The most recent base is of July
2017. That is the most recent base, base of July 2017 and I will multiply since, it is a case of ratio
seasonality, so actually the base gets me deseasonalized demand. Deseasonalized forecast you
can say. And, then I want the forecast of August 2017.
So, I will multiply this deseasonalized forecast with the seasonal factor of august 2016, so that is
1.178. So, I will multiply this by the seasonal index of august 2016. Now my job here is to
determine the values of base of 2017. So, I have the base of June 2017. I will use these values of
30.0273 to get my updated base for July 2017.
And, already since, I have just started this problem. So, because of the initial calculation I have
the seasonal factor for august 2016. But, this seasonal factor for august 2016 will also be used to
get the updated seasonal factor for August 2017. And, that August 2017 seasonal factor will be
used for calculating or for determining the forecast for August 2018. So, the point which I am
trying to make seasonal factor will be used for the previous years. And, the base is used of the
current years.
So, if I want to write this equation in a generic term, the equation will become like the F t, 1. The
period current and one period ahead that signifies Ft, 1. So, if today I am in July 2017, Ft, 1 means
it is August 2017. Then July 2017 is my current base multiply by I Aug,2016 , so August 2016 will
be, when I am July 2017 so, if this is a data which is presented on monthly basis.
FAug,2017 = SJuly,2017 x IAug,2016
This data is present on monthly basis. You may have data presented on quarterly basis, you may
have data presented on the half yearly basis also. Here is the data presented on the monthly basis.
So, you have 12 months in a cycle. So, if I am in July 2017, one cycle is of 12 months. And,
cycle is represented by capital L in our case. So, if I am writing t-L here. It means it is July 2017,
it becomes July 2016.
163
And, I want the base value of august 2016. So, +1, this makes me this equation in to the general
form . When, I want to determine the forecast for the next month. In case of when today I am in
July 2017. If today I am in July 2017, and I want to determine the forecast august 2017. This is
how I will write the equation. And, if I want to use this formula.
Ft,1 = St + It-L+1
If today it is July 2017. So, this is for august 2017. Now, I want to determine if today again I am
July 2017. But, now I want to determine the forecast for December 2017. If I want to determine
the forecast for December 2017, or for any other month October, November, December etc. So
in that case I write that I want to determine the forecast of M period ahead. And, that will be S t,
that is the base of July 2017 multiplied by It-L+m.
Ft,m = St + It-L+m.
That signifies that I want to determine the forecast of mh period from the current period. So with
this we get the most generalised forecast using the current base, and the seasonal index. Which
are available to us at the moment. So, we will see now in our next class, that how to calculate
this updated base value and how to calculate this updated seasonal index, and then how to
determine the forecast for the any next period in case of ratio seasonality model. Thank you very
much.
164
Lecture-14
Tracking Signal and Seasonality Models
Welcome back, we are discussing the forecasting methods in a supply chain environment. And
so far we have discussed different types of time series analysis methods. We have discussed
about the simple methods where we are taking simply average of the past data. And on the basis
of that average we are determining the forecast for the coming periods. Then we also discuss
about the limitations of those average methods.
And we discussed about a smoothing methods, where we consider the demand is fluctuating
about a base value. And in those a smoothing methods we discussed the some type of very
fundamental method where there is no type of characteristic available in our demand data. We
started with basic exponential smoothing model in which only there are fluctuations around a
base value, a level value.
Then in our last session we discussed a slightly more complicated model where some kind of
trend was also available in our demand data. And with the help of trend and base we actually
determine the forecasted value for the coming period. Now we are moving to a more complicated
model where seasonality is included in our demand data.
165
And here we have seen in the last class, that some data is available to us. Where, we are saying
that ratio seasonality is present in this demand data. We discussed in the last class, that it is
almost impossible from the naked eye to determine the type of data, which is available with us.
The second column in this table represents the actual demand of 2016 for various months.
And for 2017 from January to July. Now from this data it is almost impossible that somebody
can say there is no component it is a horizontal demand data. Somebody can say, this data
represents trend as well as seasonality. And now I am claiming that this data has ratio
seasonality. So, for this purpose that whether it has ratio seasonality should be use the ratio
seasonality model, should we reuse the trend model, should be use the trend with seasonality
model.
We discussed in our last class. That there are different types of forecasting error measures. And,
out of that we discussed MAD, that is Mean Absolute Deviation. This is the most important
measures which helps us in selecting the appropriate type of forecasting model. We discuss the
procedure to calculate the Mean Absolute Deviation. We discuss other type of forecasting error
measures also like average error.
We discussed MSE that is Mean Squared Error and then we also discuss about MAPE which is
Mean Absolute Percentage Error, but out of various types of forecasting error measures we said
166
that MAD-Mean Absolute Deviation is the most important type of model is the method which
helps us in determining the appropriate type of model for the forecasting purpose. And at the
same time since we have talking of smoothing methods there are smoothing constants alpha,
beta, gamma.
What should be the appropriate values of alpha, beta, gamma. That will also be determined using
these methods of forecasting error measures. But, along with those methods of forecasting error
measures. We also have a more measures which will help us in deciding whether the model
which we are using. This model has any kind of biasness or not. Now the meaning of biasness is
that this is the actual demand 19, 25, 19, 21 etc.
Now, if continuously if I over forecast my demand is always less than the forecasted value or
vice versa. Demand is always more than the forecasted value. So, in both these cases I will say
that my forecasting model has some kind of biasness. Either I am regularly over forecasting or I
am regularly under forecasting. In a good forecasting model, in a appropriate forecasting model
we should have a fluctuating kind of situation.
Sometime, I may over forecast, sometime I may under forecast. So, that the nut story should be
more like the concept of average error. It should be on a 0 side. So, that is one thing which is
determined with the help of one more concept which known as tracking signal.
167
Now, this concept of tracking signal will give you fair idea. Whether, is our forecasting model is
moving in one direction or it is a balance kind of model. Now to understand the calculation of
forecast this tracking signal, in case of forecasting. This tracking signal is calculated as RSFE
divided by MAD. RSFE this stands for Running Sum of Forecasting Error.
MAD we have already discussed in the last class, that is Mean Absolute Deviation. So we need
to calculate Running Sum of Forecasting Error and MAD. And, with this ratio we can calculate
the tracking signal. Now I can demonstrate you a sample calculation of tracking signal. And then
with the help tracking signal calculation we will also understand the physical significance of the
biasness of the forecasting model.
So, now let us have a very sample of tracking signal. You can have the real time calculation also.
The reason of discussing this tracking signal in this class is so because that we are discussing the
real time decision making. The concept of tracking signal will help us in improving or in
adapting the forecasting model in the real time situation. So this tracking signal calculation,
which is I am showing you for a sample data. Let us have six periods.
And for each of these periods let us forecast, let us assume some forecast value , that I am
forecasting 1000 units for each of these periods. Now the actual demand for these periods, let us
assume some values of the actual demand for these periods. And these demands may be 950,
168
975, 1050 then again 950, 1025 and 1100. Let us say these are the demand values of over the
period. Now with the help of these forecasting values and the actual demand data I can determine
my forecasting error which is actual demand.
This is column number 1, this column number 2, this is column number 3. So forecasting error
will be column number 3 – column number 2, that is my column number 4. Now, 950-1000= -
50, 975-1000 = -25, 1050-1000 = 50, 950-1000 = -50, 1025-1000 = +25, 1100-1000 = +100. So
these are the forecasting errors. The absolute value of these errors, or you can say this value will
be without sign.
So these are 50, 25, 50, 50, 25 and 100. Now the next column will be running sum of forecasting
error that is my column number 6. The running sum of forecasting error is actually the
summation of various elements of column number 4. So this is -50. Then you add -25 in to this,
becomes -75. You add 50 in to this so, that is of + sign. So, -75+50 it becomes -25. Then further
-50 it becomes -75, +25 in remains -50, and +100 so, it finally end with up with +50.
So, these are Running Sum of Forecasting Errors. Then you also calculate with the help of these
Dt values your MAD. That is your column number 7. Now what is the MAD, MAD is actually
this et divide by the number of periods. So as you go up you need to add up the values of,
absolute values of forecasting errors. So here it is 50 x 1 = 50. In the second case 50+25 that is
75 and 75/2 = 37.5. Then in third case 50+25+50 it is, 125/3 = 41.6.
Then in the next case 50 + 50 + 50 + 25 = 175, 175 divided by 4. For the fifth period it will be
175 + 25 that is 200/5 = 40. And for the last period 200 + 100 that is 300, 300/6 = 50. So, you get
the values of Mean Absolute Deviation in column number 7. And then in column number 8, we
calculate the values of tracking signal. And that will be determined by dividing the values of
column number 6 with the values column number 7 respectively.
And -50/50 = -1, -75/37.5 = -2. Then -25 for third period divided by 41.6, here the value is again
in -, but it is somewhere less than 1, and we can have fine tuned calculations also. For the fourth
period, it is -75 divided by 175 by 4. For fifth period it is, -50 divided by 40. For the sixth period
169
it is, 50/50 = 1. So, now you see you started tracking signal from the first period that is -1, it
increase to -2.
Then it decreases it remains about only about 0.5. Obviously of with negative sign, here also in
the negative sign, here it is slightly more than 1, -1`and it is coming + sign in the sixth period.
So, if I plot the values of tracking signal, on a graphical plane, so, this is time on the x axis, we
have the values of the tracking signal. So, if this is the first period, second period, third period,
fourth, fifth, sixth like this.
So for these periods we can start with -1, -2 and then in the third period it is coming very close to
0 value. And then in the fourth period it is actually 75, 300 divided by 175. So it is around 2- less
than -2 it is around -1, and in the sixth period it is crossing 1, and it is becoming +1. So you see
you have some type of fluctuations, though all the fluctuations are in the negative side from
period 1 to period 5.
You have all fluctuations which are in the tracking signals but there are fluctuations sometime it
is going to -2 sometime coming to 0.5 values also. And, then in the six period it is coming to +1.
So here we can say the forecast sometime it is over estimated sometime it is under estimated. So
it is a more like a reasonable forecasting model. But if it is like this where you have continuously
this type of data.
If you have this type of data which is represented by these red dots, where your forecasting
model is giving you a tracking signal which is represented by this red dots that means you are all
the time moving into negatives and it this means you are all the time doing the over forecasting.
This red data is a symbol of over forecasting. Your demand is not that much. But you are doing
over forecasting.
On the other hand, if you start in this fashion and then you move this way, the dots representing
the green line, if this type of tracking signal is there where continuously the value of tracking
signal is moving upward. So this type of tracking signal is representing under estimation where I
am forecasting less, the demand is more and the physical significance is that I am not able to
170
fulfil the demand of the market. Some of the demand I am not able to fulfil my service level is
low.
And as a result of that the customer satisfaction is also low. When I am doing the over
forecasting it means I am piling excessive inventory in my supply chain. I am keeping more
inventory in my supply chain which is not required because, I am not touching that level of
demand which my forecasting is suggesting. As a result of that this excessive inventory in my
supply chain has eaten away my profits.
So, both these lines which are represented by green or by red are not desirable. My line ideally
should fluctuate above this X axis. Above the time X axis, still if you see the this line the line
which we got as a result of this data that line is still acceptable. Because, there are certain
fluctuations in this line, but if fluctuations are not there we will not be able to accept or we will
say that our forecasting model has some type of biasness.
And, that is not desired, so for that reason this tracking signal is a very important tool it is very
important you can say outcome which will help us in determining whether the model is having
some kind of biasness or not. And therefore this concept of tracking signal must be used in
combination with the errors forecasting error measures particularly MAD and tracking signal
gives you biasness, and since we plot this tracking signal in real time environment.
So this tracking signal data will help us for any kind of adaptation of the data. If we require any
kind of adaptation this tracking signal will help us that you can modify your model, you can
develop your model you can take care of whatever happening in your recent data. So, that your
model does not exhibit any kind of biasness. And with this now we are coming back to this
original problem which we started and now let us see how to go ahead with solution of this
model.
So we were talking of we right now just for this sake of understanding have just assumed thus
the model or the data is showing the ratio seasonality it is exhibiting the ratio seasonality and
with this assumption we are going ahead to understand the process of solving this ratio
171
seasonality model. Now in this particular case this model needs to be de seasonalized first
whenever we have is seasonal data.
Whenever we have a seasonal data we need to de seasonalized that and for that purpose as we
discussed in the last class we first need to take the average of past demand. So in this case we are
taking only 2016 data, for the purpose of initialization of the solution process. We are not taking
this 2017 data to start the solution process. I am taking only the first 12 cells for starting the
solution process. We will take the average demand of 2016 and the average demand will be this
sum of actual demands of 2016 divided by 12.
And that comes to be 24.07. So this is the average demand for 2016 now we will divide the
actual demand of any periods like of January 2016, 19.36 by the average demand and that will
give me the seasonal index. So like we have shown the sample calculation here. The actual
demand of January 2016 is 19.36 divided by the average demand. I got this seasonal index as
0.804. And so on we do this calculation for all the months of 2016 from 0.804 to 1.128.
172
So all these seasonal indices are calculated by dividing the actual demand. These actual demands
by the average demand of 2016, that is 24.07 and students can try these calculation to get this
seasonal index. Now, these seasonal index of lot of significance, because this seasonal index will
be used for calculating the forecast for the 2017. So we discuss the development of the formula
for seasonal index.
And now with the help of the seasonal index and the base values we will determine the forecast
for the 2017.You already have the data up to July 2017 in this table. So we will like to determine
the forecast for August 2017 and to determine the forecast for August 2017. In our last lecture,
we discuss how to develop the formula and now we are directly showing you the developed
formula which we saw in the procedure in the last lesson.
173
So this forecast will be the forecast for August 2017 which is represented by F t,1. Ft represents the
current period so current period is July 2017, so comma 1 means 1 period ahead, so that becomes
August 2017. So I am forecasting for August 2017 and today I am in July 2017. So I will be
using updated base for July 2017 which is represented by s t that is the updated base value of July
2017.
And this It-L+1 this ‘L’ represents the period of seasonality. In this data you have monthly data for
2016. So, you have the demand data available for each month. So, in this case the period of
seasonality. Since, monthly data is given is 12, sometime you have quarterly data available to
you. So, in that case period of seasonality ‘L’ will be 4. If you have half yearly data available
data to you in that case the period of seasonality is 2.
So ‘L’ will be 2, so depending upon the type of data available to you ‘L’ will be in case of half
yearly data it will be 2 in case of quarterly data it will be 4 and in case of monthly data it can be
12 and even if you have fore night data ‘L’ can be 24 also. So, depending upon the type of data
available to you. The values of ‘L’ will change, so if weekly data is available to you so in case of
weekly data ‘L’ can be 52 also.
So, depending upon since we are considering the monthly data. So, ‘L’ is 12. Now the formula
says t-L so if I am talking of July 2017. So t-L will be July 2016 and +1 it means August 2016.
174
So, for forecasting in the month of July 2017 for August 2017. I am going to use the base value
of July 2017 and the seasonal index of August 2016 and their multiplication will give me the
forecast of August 2017 and for this purpose, I require two things.
One is the updated base value of July 2017 which I am going to calculate from this formula and
the second is It-L+1 which is the seasonal index of August 2016. So, that seasonal index of August
2016 is already available to me that is 1.178. The updated base value of July 2017 is not
available to me. So that I need to calculate and once I multiply this value whichever will come
here with the value of August 2016, 1.178.
That will give me the forecast which I will put here for August 2017. So, this value which I need
to get here multiplied by 1.178 that is the seasonal index of August 2016 will give me the
forecast of august 2017. Now for that purpose for getting the updated base value of July 2017, I
will use this formula now in this formula you see what I am taking since this is the base value so
here It-L this represents again the deseasonalized demand of July 2017.
So I will whatever Dt is the actual demand when I divide this Dt by seasonal index factor I get the
deseasonalized demand or you can see the base value. So whatever is the current base value and
then what ever is the previous base value. One period before, so I will take a part of that and
then. So that alpha is the smoothing constant. A smoothing constant multiplied by the
deseasonalized demand of July 2017 multiplied.
St= α (Dt / It-L) + (1- α)St-1
Then you have the addition of St-1 the remaining component comes from the base of June 2017.
So, if I see from this table I take this value 30.0273 this is S t-1. This is Dt 29.14 this is Dt. This Dt
will divided by 0.988, to get the my deseasonalize demand of July 2017. 29.14 divided by 0.988
multiplied by alpha value + one component of my updated base + a part of that will comes from
30.0273.
And, this will be multiplied by 1-alpha. So, with this way I will get my updated base value for
July 2017. When, I multiplied this by the seasonal index of august 2016. I will get the forecast of
175
august 2017. But at the same time I will like to update my seasonal index also though it will not
be used immediately.
It will not be used immediately it will be used for 2018. But, I continuously need to update my
seasonal index also and for that purpose this equation will be used of seasonal index where I get
the seasonal indexes here. This is Dt upon updated St which is multiplied by the a smoothing
constant of seasonality that is gamma and this is the t-L. This is the seasonality index of the same
period one cycle before t-L.
It = γ(Dt / St) + (1 - γ)It-L
And this updated trend of July 2017 will be used in year 2018 not now, and we get the updated
base value and we already have the old seasonal index multiplication of that will give us the
forecast for August 2017. So with this our ratio seasonality model where we have some kind of
seasonal component and this seasonal component is ratio in nature. So we have got this
forecasting of that method.
So we are closing our this session at this end. Now in our next class we will take this discussion
to a more complicated forecasting model where we will have a trend component as well as
seasonality component in built same data. Thank you very much.
176
Lecture-15
Forecasting using Multiple Characteristics in Demand Data and Inventory Management in
Supply Chain
Welcome back, so far we are discussing about the various forecasting models in the supply
chain, and today also we will continue with the same discussion of forecasting methods in the
supply chain. In our last session we discussed about the use of tracking signal in the decision
making of the forecasting methods. Then we also discussed a case where we had the data, which
is given to us for year 2016 and 2017.
And we assumed initially that this data is representing ratio seasonality in it is case. And as we
discussed that we need to have proper use of MAD as well as tracking signal for determining
suitabilty of a forecasting model. And, it is almost impossible for anybody, that with this type of
data, you can determine the characteristic of your data.
Sometime you have demand data like 50, 60, 50, 60 and then all of a sudden it increases to 90,
100 and then it remains 90, 100 and it comes to 50, 60 that way. So in that case you can say that
there is a clear visibility of some kind of seasonality, but when you have this type of data which
177
is 19, 25 again 19, 21, 20 like that. So it is no way that just by seeing the data you can analyse
that this has the ratio seasonality or linear seasonality or trend or no trend or no seasonality kind
of thing.
So, we discussed in the last session in detail that how to use tracking signal and how to use mean
absolute deviation to determine the suitability of a particular type of forecasting model for the
given set of data. Now we discuss the ratio seasonality in the last session. We will go further in
this session beyond one type of component. In last to last session, we discussed only trend with
our sample data. Then in the last session we discuss seasonality with sample data.
Now we will consider for the same set of data, that this data has trend as well as seasonality, so
now we are assuming that this data does not have simply ratio seasonality. This data has trend as
well as seasonality.So, now we are going with this type of model.
(Refer Slide Time : 03:07)
Where, this model has trend and seasonality, in case of trend we are assuming that this is the
linear trend, this is the additive trend. You may be aware that in one of the session we discuss
that there can be 2 types of trend. One is linear trend, which we are assuming here and linear
trend is determined with the help of St - St - 1. This is the formula to determine the linear trend.
And this formula of linear trend or additive trend is St - St - 1.
178
This is linear trend which is also known as additive trend. There can be another type of trend
also, which can be ratio trend or multiplicative trend. In case of ratio or multiplicative trend the
trend component is determined with this type of ratio. This is the formula. So, if I take the ratio
trend here, if I take the multiplicative trend here. So, this way of calculating the trend will
change. This St - St - 1, will become St / St -1. And rest of the model will remain as it is.
So, we need to know that what type of model can be customised using this one of this general
type of model. This is the most general type of forecasting model which we are discussing now
and since this is the most general model you can see that we are using all 3 types of smoothing
constants. In last two, three sessions we have discussed about the smoothing constants alpha,
beta, gamma. This is one model where I am using all 3 types of smoothing constants.
Alpha, to smoothen the fluctuations of the base value ‘S’. Then beta to smoothen the fluctuations
of the trend value. And, then gamma to smoothen the fluctuations of the seasonal component
seasonal index. So all three types of alpha, beta, and gamma are used here time and again. We
have discussed that the values of alpha will lie between 0 to 1, value of beta will lie between 0 to
1 and values of gamma will also lie between 0 to 1.
But, the popular values of alpha, beta, gamma will be from 0.05 to 0.30. Now the lower values
again we have discussed have more a smoothing effect. You want more smoothing effect used
179
lower values of alpha, beta, gamma. But, sometime because of some happening in the
environment whatever change, whatever disturbance has taken place in very recent past. I want
to incorporate that disturbance completely in my model, completely in my forecasting model.
In that case I want to use higher values of alpha, beta, gamma whatever type of you all want to
see that there is a complete shift in my base. I will use higher values of alpha. So that the most
recent disturbance or most recent happenings are totally captured in the base or so on for the
trend as well as for the seasonal index. Now let me start again with my the last formula that is F t,1
that is the forecast.
If I am in the month of July 2017 and I want to forecast for the August 2017. In that case, F t,1
represents July 2017 comma 1 represents August 2017. And in this case I will use the updated
base value of July 2017 + updated trend value of July 2017 this is S t+Tt. Then It-L+1 this we
discussed in the last session also. This is the seasonal index of August 2016. So, these are the
values of July 2017, this is the value of August 2016 multiplying this S + T by the seasonal
index.
Ft,1 = ( St + Tt)(It-L+1)
I get the forecast for August 2017. So, this is how you can interpret this generalised formula
which is represented in terms of Ts and Ls. Now when I am using this formula I need the
updated values of St and trend. So these are the updated values of S t and t, but this value this
updated value of seasonal index It. This is not to be used in the current forecast, when I am doing
forecasting for August 2017.
And I am calculating this It which is for July 2017, so this is not to be used right now, this will be
used in the next cycle for 2018. So I will used this formula where this D t / It- L represents the
deseasonalized demand of July 2017, and alpha is the smoothing constant, whatever factor of the
current demand I want to incorporate that will be represented by the alpha. Then this is the
St-1+Tt-1 that is the previous base that is the base of June 2017 and this is the trend of June 2017.
180
Now once I have calculated the updated base value then it is a kind of iteration. I will use this
updated base value here to calculate my updated trend value. You see you check first I need to
do this calculation of St, and with the help of this St the second calculation will come for Tt that
is St will be used here. This β( St - St-1 ) which is for one period previous the beta multiplication
of that.
This is the factor of my current trend, and this is the old trend, and this will I will get the updated
value of the trend. So, this St and Tt, this I am using here in the final forecasting multiplied by the
old seasonal index of one period ahead. And, this seasonal index this updated seasonal index of
July 2017 which is a smoothened with the help of this gamma a smoothing constant will be used
for next cycle.
Now, when I am using this Ft,1. I can also use this equation for a generalised kind of forecast like
this is for one period ahead but I can use same for m period ahead. I am in July 2017 right now.
But, I want to forecast for two periods, three periods, four periods ahead using the current data.
So that is m period ahead and in this case I will use the updated base value whichever is
available to me + this trend Tt.
I will take the value of t as a constant now and this I will write simply as t, and I will multiply
this by this M and same in case of seasonal index, I will take the this 1 will be replaced by m. So
this will become t-L+m. So the point which I am trying to make here that if it is today it is July
today is July 2017 and I want to forecast for October 2017. For August we have already seen.
But now I want to apply the same formula for October 2017.
Ft,m = (St + mT)It-L+m
So, FOct,2017 when today is July 2017 will be SJuly,2017 this is the updated base value + now what is
the value of m August 1, September 2, October is 3. So M is 3 in this case 3 and this T will be
for the July 2017 x T of July 2017 close the bracket and the seasonal index of t-L+m though
seasonal index of October 2016 that will be determine my forecast for October 2017 when I am
today in July 2017 and similarly you can calculate for November.
181
You can calculate for December, you can calculate for September when you are in July 2017.
Obviously there is one thing which can come to your mind that. When I am forecasting for
higher values of m my forecasting errors will be high that is true. Because I am using slightly
distant data for doing this forecast, So with higher values of M my forecasting errors are happen
to be more.
And when I used only for one period ahead, I have more accurate forecast because I am doing
forecasting with the most recent data in this case. But nevertheless many a times this forecast is
also very useful, because it will help us in doing the appropriate planning for our supply chain.
So this is also not to be discarded and this is also many a times very useful for doing the initial
planning of the supply chain decisions.
So with this we discussed large number of forecasting models using the time series analysis
particularly exponential smoothing methods and if you remember we have discussed nine
different types of models out of 9 different types of models we discussed in our class four
different types of model. But the same type of models can be customized like with this general
model you can use this model for any other type of forecasting model.
Because, of different ways of representing your trend, different ways of representing your
seasonal index and accordingly you will have your forecasting. But depending upon your ability
to identify right kind of characteristic in your demand data and the secondly the appropriate
values of alpha, beta, gamma. So what we do normally with the different values of alpha, beta,
gamma we prepare a table of MAD.
And whichever set of alpha, beta, gamma gives as lowest value of MAD, that set of data of
alpha, beta, gamma will use for a particular data characteristic but in a real time environment
when I am talking of more pro-activeness. So, many a times it is possible that during the course
of our forecasting. If we are having our tracking signal going into a particular direction of either
over estimating or under estimating we need to change the values of alpha, beta, gamma
intermediately also.
182
So, with this we have a very good discussion about our forecasting techniques. The next
important decision area in the supply chain is about the inventory. Once you have decided about
the forecast. That this is the final forecast in your supply chain so next issue is about the
inventory management. In the last class when we were discussing about the tracking signal. We
discussed that if you are continuously doing the over forecasting.
If you are continuously doing the over forecasting in that case you will always be having some
kind of inventory in your supply chain and that inventory is not desirable. Because, it will take
away your profits. So, now the second important area where analytics will help as that is the
inventory management. So now we will move to our discussions of inventory management
where we will see that how inventory is to be managed.
To start our discussion we have a very simple case where we have only a supply chain with two
installations, though there will be many installations in a supply chain. But just to start our
discussions of inventory we are considering a supply chain with two installations one and two.
Over a period of time we will build more installation in this supply chain, but to start the
discussion we have two installations.
Now you consider that installation 2 is the retailer and installation 1 is a wholesaler. So,
installation 1 is a wholesaler and this is retailer, so wholesaler is responsible we all know
183
wholesaler is responsible for supplying products to retailer. So, whatever inventory is managed
at the retailers end that is coming from the wholesaler.
So now if you go to your operation management classes you know there is a method of inventory
management which is very popular which is known as EOQ Economic Order Quantity model.
Now we will start this process of inventory management with that EOQ model. In EOQ model
you may recall that we keep inventory and inventory reach to a particular label and then we start
consuming inventory. So now we are seeing the same kind of inventory management for this two
stations.
And, this can be represented with the help of this picture. Now, picture shows you that this is the
inventory management at installation 2, where at installation 2 we are keeping inventory of Q
items. At a time we ordered Q items at installation 2 and these Q items are coming from
installation 1. So, now what we are doing at the installation 2, you are ordering Q items at a time
then you are consuming these Q items, and over a period of time.
These Q items are consumed completely and then again you have already placed a order, so that
by the time you reaches this zero level again your stocks are replenished to the original Q 2 level
and again you have the full stock and you start consuming these by the time again you reach us
184
to zero level. Your stocks again replenished to the original Q 2level. And this is a very popularly
known sawteeth pattern of inventory management.
So, at installation2, you can clearly see the sawteeth pattern of inventory management where you
are taking the items for consumption and you are receiving all Q2 in one lot and there is a slow
rate of consumption which you are consuming. Since we are seeing that installation2 is retailer,
so you are selling products at retailers end, and maybe this period is one month period. So, you
are consuming the Q2 is stock in one month.
Let us assume this so this is a very popular sawteeth pattern. The EOQ model economic order
quantity model is based on this sawteeth pattern. Now the inventory at installation1, at
installation1, I am procuring Q1 item from the manufacture. The wholesaler is procuring Q1 items
from the wholesaler. Now what happens when the owner of installation1is procuring Q 1 items
from the manufacturer.
So, if you see these two figures so whenever Q 1 items are coming to installation1 out of that Q 1,
Q2 items are coming to retailer. So this installation1is receiving Q 1 items and out of that Q2 items
are given to installation2. So the installation 1 is left with Q 1 - Q2 items and it will remain with Q1
- Q2 items for sometime and whenever the next order is required by the installation2.
In that case this further Q2 items will be supplied to installation 2, so it will further reduced to Q 1
- Q2 and then it will remain for sometime Q1 - Q2 and then again when there will be a
requirement of further Q2 items at installation2. So it means supply the remaining items to
installation 2and this is stop will come to zero level at the installation 1. So what we have taken
just for the sake of understanding that Q1 is four times of Q2.
Q1 is four times to Q2 because Q2 , Q1 is able to supplied four times to Q 2 out of it is originally

stock and then it will remain like this for some period like 0. And, whenever the next order is
expected at the installation2, so this installation1 will adjust it is supply requirement in such a
way that whenever this next order is expected as installation2. It will receive it is original Q 1
185
order from the manufacturer on that very day. And, out of that Q 1 items Q2 will immediately go
to the installation2. And it will remain with Q1 - Q2.
And, the pattern will further be repeated. Now, when you see the inventory management at
installation1, you are seeing these thick lines, so these are the physical stocks which are available
at installation 1. But, then you also see this dotted line you see this dotted line, and this dotted
line is being plotted taking the corner points of these 2 stocks. And, the Q 1 level, the original Q1
level the item which the installation 1 is receiving from the manufacturer.
So, this line which is the line almost of this type. And, when I go for second round of receiving
the order of Q1 quantity at installation1. So, this will again come somewhere here. And, you will
see that this pattern of dotted line can be represented at installation 1 also. So, what I am trying
to say that you have same sawteeth pattern at installation1 as well as at installation 2. So, we can
apply this EOQ model very well at the installation1 as well as installation2.
Now, for applying the EOQ model there are certain assumptions, and let us discuss those
assumptions, one assumption is that Q2 is always either equal to Q1 or less than Q1 . Because, Q2
is coming from Q1 , so Q2 cannot be more than Q1 in any case. Depending upon the model it is
quite possible that Q1 may be equal to Q2 . But, Q2 can never be more than Q1, this is one
assumption. The second assumption is that there are 2 types of cost which are involved in this
model. One cost that is the cost of ordering.
So, at this level you are requiring Q2 items, and this level you are requiring Q1 item. Then the
second assumption is that there is a cost of ordering, whenever, you are putting an order, so,
there is a cost of ordering receiving the order, loading, unloading etc. So, there is a cost that is
all the cost at installation2 by represented by subscript 2, and all the cost at installation1 are
represented by subscript 1. So, there will be cost associated at 1 and 2.
And, then there will be the holding cost, whatever items you are holding, you are blocking the
money of your organisation in holding the items. You are putting some kind of insurance
charges, you are putting some kind of rent on keeping those items. So, all these cost are
186
associated with the holding cost. So, you are putting the holding cost also at each of these
installations, whatever there in your supply chain.
So, the second cost is the holding cost, so holding cost and the order cost or in some literature it
is mentioned as setup cost also. So, holding and setup cost at each of these installations is
constant. So, there is no change in the holding and setup cost. So, we will take that is as our
second assumption. Then third assumption is that whatever order we are placing at installation2
or at installation1.
We are receiving the complete supply, there is no short supply, if I am giving order of Q 2
quantity, Q2 means let us say 100. So, I am receiving all 100 in 1. At Q 1, let us say, I am giving
the order of 400 units, so I am receiving all 400 in a single lot. So, that is another assumption,
there is no short supply. Then there is a determined time when I am going to receive the supply.
Today I am giving order and after four days I am receiving the supply.
So, the four day period is known as lead time. The time when I placing the order and the time
when I receiving the supply this duration is known as lead time. So, lead time is also fixed.
Otherwise this model will have some kind of problem. So, I take this assumption that whatever
my supplier whatever my wholesaler says, that lead time is a constant. So, that is another
assumption.
Then there is we have seen in the model that there is a constant slope of each of these teeth. The
slope represent the consumption rate of my items. So this consumption rate is also constant. In
this diagram, though in practical terms it is always impossible to say that you have a constant
rate of consumption at the retailers end. There will be some fluctuations.
In our last classes of forecasting, we have discuss this in very detail, that you will never have
these types of straight line of the consumption. There will be some kind of zigzag lines. But, for
the purpose of modelling we assume that the rate of consumption is almost constant. So, that is
also one assumption we are taking. Then one more assumption, we take for the simplification of
our discussion.
187
That there is a direct integer multiplication of Q 2 to get the Q1. For the sake of this model
building I am considering that Q1 is a direct multiplication of Q2 . So, in this case like Q1 is 3Q2.
So, here in one Q1, I am giving order of 3Q2. You can say, you can argue why it is not 2.5 or why
it is not 3.5. if it is 2.5 or 3.5 in that case I will not be able to match exactly with the cycle of my
Q2.
So, there will be some kind of extra inventory holding at my installation 1, so to avoid that I take
some integer multiplication of Q2 to get my Q1. So, these are the some of my assumptions we
have taken more I will recommend that we can go through any class of inventory management
where basic EOQ model is start, and all those assumptions are valid for our case. So, we are
stopping this discussion at the moment. And in our next class we will see that how do we
optimise the inventory management in case of supply chain. Thank you very much.
188
Lecture-16
Inventory Management in a Supply Chain
Welcome back, in this course of supply chain analytics, so far we have discussed about various
types of forecasting models which we used for taking decision with respect to planning of
various activities in the supply chain. In our last session we also started one very important issue
in the supply chain that is inventory management. Inventory management is a subject which
most of us have studied in the classes of operations management, but inventory management
when we study only for a standalone entity.
That inventory management is very different type of inventory management. In case of supply
chains where various entities are linked with one another, inventory management alone can make
or destroy your supply chain decisions. So, it is very important in a supply chain case, that we
take very appropriate decisions with respect to inventory management. Obviously, the feeding,
the input for taking decisions with respect to inventory is provided by the forecasting.
That is why we discuss forecasting in the beginning of these various decision making activities.
Now, we are into this inventory management, and in operations management class I expect that
most of us have already studied, various different types of inventory management models.
Now, we will see here, in this particular course that how we can extend those inventory
management models in a supply chain environment. Supply chain environment when I say, it is
being characterized by having different types of entities in your supply chain.
189
This is a very simple case, we started with in our last session, that here we have 2 installations. In
this supply chain we have 2 installations, installation1, and installation 2. In installation1 you can
say can be the wholesaler, installation2 can be retailer, or installation1 can be a manufacturer and
installation 2 can be a retailer directly. So, just to start the discussion that how do we move from
our conventional inventory management where we discuss only a particular installation.
In that case in a conventional system we consider only a particular station and we manage the
inventory only for one single entity. Here, if we do that, means if we start managing the
inventory separately for these different entities. We will lose the very essence of supply chain
management. So, therefore it is important that we should consider, that whatever decisions are
taken.
This should be taken in the interest of entire supply chain. Right from the day 1 when we started
this course, we are talking continuously about collectivism about taking decision where we can
take the interest of entire supply chain. And, therefore we will use principles of or the formulas
or the models developed for single installations. But, here we will extend them in a supply chain
environment.
So, to start that case, we are taking only 2 installations as I just explained. Now, in case of this
particular model we can understand, that there are different type of relations between 1 and 2.
190
But, one thing is very interesting, we all can easily appreciate that if 1 is a wholesaler, 2 is a
retailer. If 1 is a manufacturer, 2 is a retailer, if 1 is a supplier of some type of raw material, 2
may be the manufacturer.
So, what I am trying to say that as we move from this side to this side in a supply chain. As I am
moving from left to right in a supply chain. I am adding more value to my product, and therefore
many times supply chains are also known as value chains. Because, with our formal movement
with our movement from left to right we are adding value to our products.
And, therefore supply chains are the value chains and as I am adding value to my product from
left to right the correspondingly the unit cost of the product also increases, because of the value
addition. So, if you remember in our conventional model of inventory management. We use to
have 2 important variable cost.
And, you can recall that these 2 important cost are, one is order cost, and the second is holding
cost. These are the 2 important cost, which we consider in our most popular inventory
management model that is EOQ model. That is Economic Order Quantity Model. In that EOQ
model we consider these 2 types of cost. Now, the holding cost, I will like to emphasise on. The
holding cost is that cost when you are keeping inventory with you, you incur some type of cost
on keeping that inventory.
191
And, that cost may come from variety of sources. For keeping some items in your stock, you
have taken a warehouse on rent. So, that rent which you are paying for keeping the inventory
may go for this holding cost. You are paying for the security of the item, some type of insurance
premium. That insurance premium will also be the part of the holding cost. Then, there will be
some kind of opportunity cost also.
You have blogged your capital in keeping the inventory, and as a result of that you are losing
some other opportunities of using that capital, so that is also added in to the holding cost. So,
holding cost may come from variety of sources. And, therefore holding cost is always you must
have learned in operation management classes. That holding cost is always represented as a
percentage of the total material cost. The cost of the material which you are buying for which
you are keeping the inventory.
So, it is always represented as a percentage of that material cost. The order cost or in some
literature you will also find the name setup cost. So, the setup cost or the order cost. This setup
cost results because of a your receiving the order, whenever you give an order, so the setup cost
will come, because there will be some kind of loading, unloading, some kind of communication
involved.
So, these are the order cost. Whenever in a plant you are working, and whenever you are
changing from one product to another product. So, you need to change your tools, so, you need
to change your dyes, and all those things. And, as a result of that we call it setup cost. So, like
just to explain these 2 costs more clearly to you. Let us take an example in that example I want to
drink, cold drink after my dinner daily.
And there are 2 possibilities. One possibility that I take my scooter and after dinner daily I go to
a general store, take one cold drink. So, that is one way of keeping the low inventory. So,
whenever I require on a daily basis I go to a general store take a cold drink. On the other hand I
can also do that in one go I can purchase the entire carat of the cold drinks. Maybe of 12, maybe
192
of 20, and then I consume these cold drinks one by one daily, and I need not to visit that general
store again and again on the daily basis, for next 10, 12, or 15, 20 days.
So, in that case going to that general store and coming back, the amount of fuel that I am
consuming that will go to the order cost. And, since I am not keeping any cold drink in my stock,
so holding cost is absolutely 0 in this case. On the second criteria, when I am purchasing the
entire carat for the cold drink, for let us say one month stock. So, I need to go to that general
store, to purchase cold drink only once.
So, my ordering cost has reduced substantially here. I am paying the ordering cost only once, and
the holding cost has increased tremendously now. Because, I have kept the stock of 30 cold
drinks with me in one go. So, I have paid some 200 rupees or like that for that purpose, and for
that purpose my 200 rupees is blocked. I cannot use that money for any other activity, and it is
also possible that in this leakagement takes place.
Some wear and tear may take place, so I may loss my one cold drink, so that loss will also come
into the holding cost. So, I have to take a call, I have to be optimum that my ordering cost and
holding cost should have some kind of balance, and when I achieve that balance these 2 extreme
situations I discussed. In one case I have only ordering cost, no holding cost.
And, in other case I have very low ordering cost almost negligible, and my holding cost is very
high. So, I want to have a balance between these 2 types of cost. So, that is what we do in our
inventory management systems. And, this very beatifically we have already learn in our
operation management class using the EOQ formulas. Economic Order Quantity formulas we
have already learned.
So, now we will apply the same knowledge in this particular case. Now, as I was mentioning
about the holding cost. Holding cost normally we represent with h. So, whatever we are doing
for installation1, we will use subscript one for that purpose. And, whatever we will do for
installation2. We will use subscript 2 for that purpose. So, we have the holding cost at
installation 1 as h1, and holding cost at installation 2 as h2.
193
These are the 2 holding cost we are already incurring. Now, because as we are coming from left
to right, as I said we are adding more value to our products. So, product is more valuable from 1
to 2. And, since holding cost is paid on the material value of the product. So, always and always
you can understand that h2 , h2 because, it is on the right side of installation1, so h2 will always be
greater than 1.
You can understand why h2 will always be greater than h1. Because, h2 you have added more
value by the time product reaches this stage 2, installatiion2. You have added more value to that
product, and therefore h2 will be more than h1. Then another important thing is we have also
discussed there will be the order cost. So, order cost let us represent as k. So, there are ordering
cost at installation 1 as k1 and order cost at installation2 as k2.
These are our ordering cost. Now to further understand that how EOQ model will operate,
because now we are going to the stage of making the model for these 2 stage inventory
management. And, for that purpose in our last class, we discussed this particular graph where we
had the system of inventory management like this.
You are procuring or at a time you are giving order of Q 1 items at stage1, and Q2 items at stage 2.
Now, Q2 item you have received here, we discussed you received Q 2 here, and the installation2
let us say is retailer. So, the rate of consumption of these Q 2 item is being represented by this
194
slanted line. This slope line represents the rate of consumption of these Q 2 items. Then after
sometime this Q2 item will finish.
And you will have 0 inventory level. So, this is the point where we achieve the 0 inventory level,
but because of our modelling, because we know many things we have already ordered in advance
that by the time you go to this 0 level. You get a fresh stock of Q 2 items. And all of a sudden as
soon as you are touching this 0 level your new stock comes and your inventory level again
reaches to the original Q2 level. So, this vertical line represents the replenishment of the stock.
So, if I say in this particular graph if you see in this particular graph we have these types of
curves.
So, these lines which are straight lines, these are the replenishment lines. And, the slanted lines,
these lines are the consumption line these lines represents the rate of consumption of your Q 2
stocks. This is the Q2 level, and these slope lines are the consumption line. So, we have planned
in such a manner. That if I am going to have my 0 inventory on let us say 30th of January.
So, immediately on the same date I will receive the fresh supply of Q 2 items. So, that my
replenishment line shows that instantaneously my stock will again go to the original Q2 level. So,
the time, and for that purpose if this is the time I am going to receive fresh the supply. So, I
195
somehow order at this point. Here I have ordered this is time order placed. And, this is the time
order received.
The difference between these 2 times, when I am placing the order and when I am receiving the
order, this difference is known as lead time. The time between placing an order and receiving the
supply, so I already know about my suppliers, that what is the lead time. My supplier, my
predecessor can supply me in 2 days, 3 days, 5 days, 7 days, 15 days I know the capability of my
supplier.
And accordingly I decide the lead time if my supplier can supply in 5 days. So, I will know that
after 5 days my stock is going to be 0. So, I will give my order 5 days before. If my supplier take
7 days, so I know that after 7 days my stocks will go to 0 level. So, I need to order 7 days before.
So, as per the capability of your supplier, as per the lead time they take to supply the product I
decide the date of placing the order.
Nowadays because of influence of Japanese manufacturing systems in various domain of

operations management, and so on in case of supply chain management also. More and more
companies, more and more supply chains are looking to reduce this lead time. And there is a
concept coming and many of you may be aware also, that we want to have 0 lead time. We want
to achieve the 0 lead time also in our supply chains.
And, if we achieve that 0 lead time, that today I order, and immediately I got the supply, that is
just in time. As soon as I require the product I immediately get it. So, we are continuously
looking to shorten the lead times and, for that purpose it is very much important to develop the
capabilities of my supply chain partners. In such a way that they can supply me with minimum
possible lead time. Because, of so much volatility, because of so much uncertainty, because of so
much complexity, ambiguity.
Because, of all these reasons I do not want to keep long lead times. Things may change all of a
sudden, and if I have already place some order, and today I do not want, because the things have
changed. So my supplier will force me to honour the orders. But, if lead time is small, I can give
196
order only when it is required. So, it is the capability of the entire supply chain. That we need to
develop for reducing this lead time.
So, these system is there, and this system which is being represented here also is known as saw
teeth pattern of inventory management. Shows, sawteeth type of management between the
replenishment lines and the consumption lines. So this is the sawteeth type of pattern and
whenever you this sawteeth pattern.
At that time you can apply EOQ model of inventory management, or you can say also that
sawteeth pattern enables you to apply EOQ model of inventory management. So, now at
installation 2, it is very clearly visible that I can apply the EOQ model. But, now let us come to
installation1. At, installation1, because installation1 is responsible for supplying products to
installation2.
All the supplies to installation2, that come from installation1. Now, therefore for better
inventory management what we have done that whatever supplies we are getting at installation1.
Or whatever we are producing at installation 1. From that whenever we receive a new supply, a
part of that goes to installation2. And, we are left with some less amount of inventory.
And, that is what we have exactly made on this graph. Here what we have taken that this at
installation 1, the items which we are procuring these are Q 1, and installation 1. Just for the sake
of understanding, we normally need to have and you will soon understand, that the item which
we are procuring at installation 1. I am telling you a very simple relation between what we are
procuring at installation1 and, what we are procuring and installation number 2.
There is a simple relationship between the quantities which we want to procure at installation1
and installation2. Now at installation2 we are procuring Q2 items and it is very simply now clear
to us that you can apply the EOQ model to determine the values of Q2. Now what to do with Q1
and what is this N here. For this purpose just for the understanding purpose we have taken, n
equals to take 3, and equals to 3, and therefore Q1 becomes 3 Q2.
197
We are procuring 3 Q2 at installation1. So this is the Q1 we procure whenever we order this size
of Q1 and out of that whenever I receive a supply of Q 1 immediately Q2 supplies go to the
installation2. So I am left with Q 1-Q2, and I will continue with Q 1-Q2 for some period. Then in a
mean time at a installation2 these Q2 items are finished. So I am left with Q1 - Q2 here, so I will
supply again Q2 items from this Q1 - Q2 to installation2.
So, I am left with Q1-2 Q2 this and then after sometime these Q2 items at installation2 are also
finished. So, I am left with then I will supply remaining Q2 also and I am left with zero
inventory. I have no inventory at installation 1, so this is the zero level from this period to this
period there is no inventory at installation 1. Then again after sometime installation2 requires Q2
items, so what I have done.
I have adjusted my procurement cycle at installation 1, in such a manner that whenever there is a
fresh requirement of Q2 items at installation 1 only then I will procure Q 1 items at installation1.
And, like this in this particular case when these Q2 items are required here. So I am procuring
again Q1 items here, So you see where my finger is. The tip of my finger represents the Q 1 a
stock and the out of an Q1 stock again I have supplied Q2 stocks to installation2.
And I am left with this much that is Q1- Q2 and again the same process will be repeated this step
process is repeated. Now, it is simple to see that at installation 2. You have the sawteeth pattern
at installation1, you have this type of a step pattern of inventory management. So, at installation1
we do not have the system of EOQ, because of sawteeth pattern is not available there. But then
there comes a new concept that is the concept of Echelon stocks.
So, for we discussed only about installation stocks, that, whatever inventory is available at this
particular installation, installation1, installation2. So, whatever inventory is available at
installation2, I am discussing about that only. So, these thick lines are representing that
installation a stock. But, now we are going to introduce the new concept that is Echelon stock.
Now, Echelon stock is that when I am in the supply chain environment, so whatever inventories
are available at this stage, and all downward stages all right stages in my supply chain, put
198
together at a particular time that is known as Echelon stock. So, in these particular cases only
two stocks are there. So whatever inventory is available at installation2 that installation stock as
well as Echelon stock is same. Because, there is no right side entity in this particular case.
In case of 1 since right side is 2, so the Echelon stock will be whatever, is available at 1+
whatever is available at 2 at a particular time. So by this new definition we have a very
interesting thing that if you see on a particular time of starting time of a new cycle that day of a
starting time of a new cycle. Let us on this very first day So, you have the physical stock and
installation 1that is Q1- Q2.
And, physical stock at the next stage at installation2 is Q 2. So, total a stock on the day 1 becomes
Q1- Q2+ Q2 that is Q1 this point the top point of this curve. Then, we are continuously consuming
these items and at this particular time or at this particular time you see your total stock is Q 1-2
Q2. And, what is there 0, so you have this much inventory available in your total stock. Here
installation 1 + installation 2.
At this particular point you have 0 inventories here, as well as 0 here also. 0 here and 0 here so 0
+ 0 becomes 0 here. So, in a 3 cycles of installaton2, first cycle, second cycle, and three cycle, so
in these 3 cycles the inventory at installation 1, has moved Echelon inventory at installation 1 has
move from Q1 to 0. And, similarly if you see other cycles also, the same pattern will be repeated,
and when I join this Q1 point which is physically not there.
But, now since we have the concept of Echelon stock. So, this Q 1 can be become conceptualized.
And, this dotted line which is joining Q 1 to 0. And you can see this dotted line will pass from
these corner points of the steps, and this dotted will be followed in subsequent replenishment
cycles also. And, then you can see that these dotted lines are also making a sawteeth pattern at
installation1.
And therefore we can apply EOQ model at installation 1 also. So, we stop here, in this session
here, and in next class we will see that how this EOQ model will be applied at stage 1, and stage
2 of the inventory management. Thank you very much.
199
Lecture-17
Multi Echelon Inventory Management
Welcome back, so we were discussing about the inventory management in our last session, and
we started discussions about 2 stage inventory management in the supply chain. And, we
discussed that how directly at stage 2.
We can apply EOQ model because of this sawteeth pattern, but directly, we saw that we have
this type of step pattern at stage 1. And, therefore we cannot apply that EOQ model at stage 1 in
this supply chain. But, then immediately it was told to us, that there is a concept known as
Echelon inventory stocks. The Echelon inventory stocks where, you have physical inventory at
stage 1 + physical inventory at stage 2.
And, when we used this Echelon concept then we saw in the last session, that we can have these
types of dotted slanted lines. And as usual what we discuss in the last class, that these vertical
lines are representing the replenishment stocks. The stocks when it is coming to you, the
replenishment stocks. And these slanted lines, the line with slopes these are representing the
consumption lines.
200
So, same way, now we have in our stage 1 also, these vertical lines as our replenishment lines,
and these virtual lines, these dotted lines are the consumption line. And, therefore you have the
sawteeth pattern at installation1 also, and now because of the sawteeth pattern at installation1,
you can apply EOQ model at installation1 also. So, now this agreement is being made.
That we can apply EOQ model at both these installation 1, and installation 2. Now, we are told
that how to apply EOQ for a single stage, for a single installation. So, now let us start doing that
only, and as we have discussed that all the assumptions which we take for developing basic EOQ
model will apply here also. The only additional assumption, the only additional condition which
we discussed in the last session.
That, because of the value addition as we are discussing in a supply chain environment the h 2 the
holding cost which we are incurring at our right hand side of the supply chain, will be more than
the holding cost which we incur at the left hand side. So, as we are coming from left to right, the
holding cost values keep on increasing. We will discuss one more concept about handling this
increasing holding cost because, now we have discuss about the echelon stocks.
So, the concept of echelon stock will help us in understanding the combined optimisation also.
But, before we start this combined optimisation let me go with the separate optimisation. When,
we handle the inventories at installation1 and installation2 in separate manners.
201
And that is how we do the inventory management, when we are handling 2 cases separately, then
at installation1 let us say the total cost of inventory, total variable cost of inventory is being
represented by C1, and now C1 is composed of ordering cost or the setup cost and the holding
cost. So, the ordering cost is the number of times you give the order. Your total requirement is d,
which is at the end of the supply chain.
This is the total requirement d, and each time okay, let us discuss first for installation2, so C 2 is
our, because we have discussed that our Q 1 means nQ2. So, right now we are discussing the
inventory management for our second installation that we are starting this process.
202
Because, here it was very clear that how we can apply that basic EOQ model. So, for that
purpose we are starting with our installation 2, so, at installation2, C 2 the cost is d by Q2. And,
the cost of order per order is k2. So, dk2 / Q2, is the ordering of setup cost. Then the second cost is
the holding cost. And, EOQ development we have discussed that it is always paid on the average
inventory stock you have.
C2 = dk2 / Q2 .
So, Q2 is the total, 0 is the least, so, Q 2/2 is the average inventory and the holding cost is h 2 per
unit per year. So, this becomes the expression for calculating the total variable cost, and we
already know to determine the value of Q 2. You need to differentiate it and the apply principles
of maxima minima ,equate that differential equals to 0. And, then you get the value of Q2.
So, this will result in to Q2*, that is under root 2dK2 / h2. This is the economic order quantity
value, which we should have at installation2 and, this Q 2 value if I substitute back into this. If I
put Q2 back into this. This will result the C2min. Which will be I request all participants that they
should do this, and put this value of Q2 in this equation. And, then see that what C2min is coming.
So, I wait here for a minute, and you can practice, and you can check that your result should be
under root 2dk2/h2, so I am giving you directly the results. But, I request all the participants that
if you substitute these values of Q2, you get this C2. This is at stage 2. Now coming to second
level, we want to determine the value of Q 1 also. The Q1 is nQ2 and, for that purpose I write
expression for C1 also. That is the total variable cost at installation1.
d because, that is the rate of consumption of your supply chain so d will remain as it is k 2 will be
change by k1, and Q2 will be change by Q1. So, this is the ordering cost or setup cost at
installation 1. Now, comes the part of holding cost. Q 2 /2h2, now when you see you are procuring
the inventory immediately you are left with Q1-Q2. In this case we took Q1 equals to 3Q2.
203
So, as soon as you procure the inventory your 1Q 2. You say if it is nQ 2. So, out of nQ2, your 1Q2
is gone to second stage. So, you are left with nQ 2 –Q2. You are left with at stage 1. You are left
with nQ2 – Q2. And, then finally you reaches to the 0 level. So, the average inventory is at stage 1
is (nQ2 - Q2 )/2 that is the average inventory you have.
And, that we will write here that is (nQ 2-Q2 ) /2. This is the average inventory we are going to
have at stage 1 multiplied by the holding cost of stage 1 h 1. And, again now you see, the first
expression is in Q1, the second step of this expression is in Q 2. So, we can convert the whole
expression in terms of Q2 only. So, I want to remove this Q 1 from my expression. So, it can be
written as dk1/ nQ2 + ((n-1)Q2h1)/2.
So, this is the expression I have for the total variable cost at installation1. Now, out of this total
variable cost Q2, I have already determined here, so I need to determine only unknown quantity.
So, can you have a guess, the only unknown quantity you can see in this expression is this F. So,
I will differentiate this total variable cost at installation 1 with respect to n.
Because, I want to determine the value of n now, and this when I am putting the value of Q2 what
I have received from here, that I am going to put here. So, n will come, if I calculate the n will
come, and after rearranging my various mathematical terms. n will come k 1upon k2 into h2 upon
h1. So, this value of n will come, and here though in this expression and in this model
everywhere.
We have assumed that the value of n will be an integer 1, but here when I am doing this
differentiation, when I am doing this calculation. I am not taking that integer aspect into
consideration. Here, at this moment or I can write this n as n*. So, when I am writing this n*, this
n* is any kind of fractional values. This can be any kind of fractional value.
And, there is a procedure we can follow, since we have taking this assumption that n will be
exactly integer. Now, why n should be in an integer, so that is every obvious question, and it
204
must come to our mind, that why we need to have n exactly an integer. The reason for n to be an
integer is that if n is an integer so, our procurement cycle at stage 1. Can be matched with
procurement cycle at stage 2.
If n is not an integer what may happen that the procurement cycle at stage 1, will differ in it is
timing, then the procurement cycle at this stage 2. What may happen that either this right now,
you see this straight line is exactly matching with this straight line like this. So, you see as soon
as you receive this stock here, you pass Q 2 portion to stage 2, and, this is happening only because
n is an integer in this case.
n is an integer in this case only because of that it is happening here, but if n is not an integer,
what will happen in that case, if n is not an integer in that case, this line will not match with this
line. Either this line will be slightly ahead or it will be slightly backward. That is the problem
with this. How, that is be going to the problem. Because, if n is not an integer, let us say it is 1.5,
so you have 1.5 inventory.
Which you are ordering if n is 1.5, so you will order Q 1 which is 1.5 times of Q 2. And, for that
reason if this is how your Q2 cycles are moving, and we have this other pen to show our Q 1
cycles. So, here Q1 is 1.5 times of Q2. So, you received Q1 and out of that a portion of Q1 is gone,
so, you are left with Q1-Q2. Which is equals to 0.5Q2.So, you are left with only 0.5Q2 here, you
are going here.
Now, by the time you order next round of inventory, so you will go up to 2Q 2 this level. Then,
you will given of that to the installation2, and therefore you will be either having more than
required inventory yet sometime, and sometime you will have less than required inventory. So,
either you will fulfil the demand of this by just 0.5Q. Half the requirement of second stage is
being made, by the available stock at installation 1.
Or, if you are not willing to do that, then you will have excess inventories at your installation1.
So, because of these problems of mismatch of inventory management at 1 and 2. We always
205
keep n as integer , n has to be integer this sometime we call as relaxation of original problem. A
language of mathematics we say that we have done the relaxation of the original problem.
Original problem means by doing the differentiation doing the maxima, minima, and * can be
any fractional value, any continuous value. But, since we require only integer value, so we will
see that how do we follow a process to convert these n* into a proper integer. So, now let us see
that part where we will convert n* into a proper integer, so that this model can be properly used.
Now, if the value of n*.
If n* is less than 1, so your n=1. But, if n* more than 1, in case of n* is more than 1 than how to
select the values of how to round off, the meaning is how to round off, the n. So, that we can get
the proper scientific answer of doing the integer values. And, for that purpose we take this value.
We consider this value, which is the most possible, the nearest highest integer which is less than
this which is less than this is the highest possible integer which is less than this.
The meaning is if the value of n* if n* let us say is 2.57, so in that case n* in this bracket
becomes 2, and then we also have this type of relation. So you conceive this type of relationship
between these three values and with the help of comparison of these three values, you decide
whether we are going to have higher integer values or lower integer values. So, with this you will
come to know.
We will do the comparison with the help of some numerical data in our next class and with that,
we will come to know with that how this model can be used. Now this is the case when we have
done separate optimisation, this is the case where we have done this separate optimisation and in
this separate optimisation we have only considered one factor that Q 1 is dependent on calculation
of Q2, and then we determine the values of n.
And, by doing a proper procedure of rounding off of which we will discuss. We will have Q 1
=nQ2 and that is and of the separate optimisation separate calculation of inventory values. But
you will see we have not taken the benefits of supply chain environment in this particular case.
206
We consider two entities separately, and for any reason there can be a relation of Q 1 = nQ2 and
this by doing this mathematical jugglery we found out this particular process.
Now when I go for simultaneous optimisation I will use the concept of this echelon stock to
determine whether I can use this supply chain environment for my inventory management or not.
The point I am trying to say the total cost, total variable cost of inventory. In case of this
inventory is C that is C1 +C2.
This is the total cost of supply chain in a separate optimisation, separate calculation of inventory
values. Now, for simultaneous optimisation. I will take you to slightly different calculation
which is embedded in the concept of this echelon stocks. So, now we are going for that
simultaneous calculation.
And in this simultaneous calculation, you will see that we will define earlier the holding cost h
was define for the installations. Now, we will define the holding cost for the echelons, so now
the holding cost as we are moving in the supply chain from stage 1 to stage 2. We said the
inventory holding cost h1 and h2, so h2 will be more than h1, because you are doing some kind of
value addition. Now we are defining echelon holding cost which is represented bye.
207
Now, when I am talking of h2, so already I had some value addition up to stage 1 and on that
value addition up to stage 1. I have already paid holding cost h1, so it looks logical that I should
pay at stage 2, the holding cost only for the amount of value addition which is being done from
the stage 1 to stage 2, So this concept will help me in giving the concept of echelon holding cost
and echelon holding cost e1 since this is the starting stage of my supply chain.
So, even is equals to h1, the holding cost h1 is coming because of all the values which are added
at stage1, but the echelon holding cost at installation2 e 2, will be the result of difference of value
you are adding from 2 to 1. And therefore e 2 will be h2 - h1, so this new cost e2 and e1 will help
me in getting a more robust model where I will use supply chain environment for the inventory
management. So, now when I have this concept of e1 and e2.
I will like to rewrite the total cost of inventory in a single expression, I wrote cost of inventory
very well cost of inventory in two separate expressions C2 and C1 and then finally I combined as
C equals to C1 + C2. Now, when I have define this two things I am doing the simultaneous
calculation simultaneous optimisation, I will like to write total C in a single expression and let us
write that total C.
When we write total C, so obviously there will be 2 expressions for the holding cost, and there
will be 2 expressions for the ordering cost, and let us see what happens, and how these types of
results will give much. Now, when we are doing that, you will see that we have total cost
208
Which, is the result of whatever, you are ordering at installation 2, . d by Q 2 and k2 + dk2 upon n
Q2 So, this relation will be from the ordering cost. + The holding cost. Holding cost now will not
be paid for the installation. Rather we will have the holding cost given by the echelon holding
concepts. And, for that purpose Q2 by 2 that is e2 + Q1 by 2 e1. Now, Q1 by 2 can be replaced.
One it is nQ2, so this can be written as n-1 by 2 Q1, e1.
So, this is the or I can write in a slightly arranged manner, that it becomes d / Q 2 , from these 2
expressions. ( k2 + k1) / 2 from these 2 steps, you see you can take sorry, this is Q 2 here. This Q2/2
you can take out, and this remains e2 + (n-1)e1. So, this is the expression, and when you see this
expression you can very well understand.
That this holding cost right now is being written as e 2 and e1 and e1 here. It is actually h2 and h1.
And, it will h2, h1. So, it will like this, and then you will have slightly rearrangement of this
expression, it will be Q2/2. h2 + n h1 – h1. And, this can be rewritten as Q2/2, h2 – h1, will account
for e2. So, these 2 expressions will give me my e2, and h1 is same as e1.
209
So this will become ne1, so my expression if I simply summarise will be if I just for the sake for
clarity, remove these 2 lines. You can combine the expression like this.
It is,
Now, if I see any of my total cost expressions, and you see this expression. So, this d / Q 2 in the
number of orders. I am placing again here, this becomes cost of each order. This expression with
in (k2 + k1)/n . This becomes the cost of placing an order. Similarly, Q 2/2 is the average inventory
level,
And, the expression with in this bracket e2+ ne1. This is actually the holding cost for which I am
taking, so now based on our knowledge of EOQ models which we discussed, now if I rearrange
these to get the value of Q 2. So, the final expression for my Q 2 in case of simultaneous
calculation will be you all can also practise with me, you all can also write with me, that Q 2 will
be now under root to d.
210
Then here comes the cost of ordering that is ( k 2+ k1 )/n by the holding cost. That holding cost is
this e2+ ne1.so, this is the new economic order quantity at Q2 level, and this takes here, because of
the echelon concept this takes here, of the supply chain environment. So, we will see the use of
this formula with the help of a numerical data, in our next class, and I request you to please do
the differentiation of this expression, those who are interested in mathematics.
Do the differentiation of this expression with respect to Q 2, and try to get this formula of Q 2.
Though we have used our past knowledge is of EOQ write in this formula, but you can try a
fresh, so that you can get this formula. Thank you very much.
211
Lecture-18
Multi Echelon Inventory Management (Continued)
So welcome back, and in our last session we were discussing about the simultaneous
optimisation of this problem when we have 2 stations and before that we already discussed about
the separate optimisation separate use of EOQ formula for this 2 station problems.
Now, in case of a simultaneous optimisation, we develop the expressions of Q and on the basis
of that expression.
212
When we had Q1 equals to n Q2, and we had the total cost, which is sum of the cost variable cost
at installation2 as well as the installation1, and this we discussed as d/ Q 2 k2 + d k1 / n Q2 because
Q1 is nQ2 + we discussed about the holding cost, that is Q 2 upon 2 h2 + the holding cost of
installation 1, that is (n-1) Q2 /2 h1. And, now when we rearranged all these things it became
d/ Q2 k2 + k1 /n + we took Q2 /2 out of this expression.
And then when we started rearranging the terms of h2 and (n-1) h1. We saw that it is coming nh1,
and h1 we discuss is equal to e1. That is the echelon holding cost, so 1 one term came as e 1. And,
then h2 – h1, we discuss in the last session is equal to e 2, so that is e2, this we got, and now this is
our (k2 + k1)/n, is the unit ordering cost. And, n 1, and e1 + e2 is our holding cost for average
inventory.
In a supply chain environment, particularly you remember that now we are doing the
simultaneous optimisation. So, this is the effect of supply chain environment. We are trying to
build into our inventory management models. And as a result if I calculate Q 2*, that is the
economic order quantity at stage 2 will be under root 2. This expression that is the ordering cost
k2 + k1 upon n divide by this holding cost that is ne1 + e2.
213
And as a result of this when we are putting this Q 2 in this expression of overall variable cost C,
so this expression give me the value of C and the value of C will be the minimum this value of C
corresponding to the value of Q2. The value of C because of this value of Q 2 so you see that way
we are putting this here and then the value of C which we are obtaining is this, and this is under
root 2d into ordering cost multiplied by the holding cost. So, this is k2 + k1 by n into ne1 + e2.
So this is the minimum holding cost for simultaneous optimisation. Now for this reason when we
have these two expressions readily available with us. The only challenge is to determine the
value of Q1, and Q1 is dependent value of this n. So the next point is what is the value of n here,
for determining the value of n we suggest all our participants to do the differentiation of this
minimum cost with respect to n.
And put that differentiation equal to 0 and calculate the value of n. But since this value under the
right side is square root of some expressions. So, doing the differentiation of this square root
term is not so easy. So as advised in the literature you can go ahead with the square terms of this,
so you take the square of both the sides make it C square, so this under root expression will
become
And, then you take the differentiation and this will not change the result. At as a result of that
when you calculate n from this expression when you calculate the value of n this will be, you I
request all participants to do this on their own, but I writing the expression directly k 1 upon k2
into e2 upon e1, and now you can compare this value of n and the value of n we got earlier in
4that case the value of n was k1 upon k2.
214
When we were doing the separate optimisation where we using the separate EOQ formulas at
that time it was k1 /k2 x h2/h1. So, from that holding cost of our stations, now we are moving into
the echelon holding cost e2/e1 is the echelon holding cost. So, this calculation of n test the
essence of that supply chain environment.
When we are moving the purposes which we are trying to say that the movement of our
calculations from these installation calculations to echelon calculation is actually the capturing
effect of supply chain environment, when we are not considering about the echelons it means we
have not captured the supply chain environment into over discussion. So, with this we can
calculate the values of n also here also in this calculation of n as in previous calculation also.
These values of n can be any fractional value, but we want integer values of n and for that
purpose now let us see that how do we round off these fractional values of n. We discuss in the
last session that if the value of n this calculated value of n is less than 1. If it is less than 1 then
you take it n equals to 1, but if it is more than that then there is process and let me give you a
brief of that with the help of which we can take this discussion forward.
So, now let us have some kind of numerical data with us and with the help of that numerical
data. We will try to see how do we use this whole discussion into practice and what type of
benefit are we going to get with this simultaneous optimisation. So, in small data we will like to
discuss with you and with help of that small data, we will see that how these models can be
actually practiced.
215
And for that purpose we have these two installations, and the cost of setup at these two
installations are given to us. At installation1 the ordering cost or the set up cost, k 1, is 1000
rupees and at the installation2, h1, this cost is 100 rupees. At installation1 the holding cost is 2
rupees per unit per year, and at installation 2, this is 3 rupees. And the annual demand is 600
units. So, this much data is available to us. Now when we have this much data with us, you can
also see the values of h1 and h2.
First, before we start solving this question before we start using this data see the values of h 1 and
h2. h1 is a primary stage, h2 is towards the right hand side of the h1 installation1, and therefore the
values corresponding to h2 are higher than the values corresponding to installation. So that is one
particular thing as a student of the supply chain course, we should understand that what is the
physical significance of these things is.
Now, you see though it is not hard and first but still you see as you are moving the right hand
side, your set up cost decreases if it is purely setup cost it decreases, because when you are
moving towards right side not much tooling is required, because product is already finished and
it is more likely to be repackaged or just the logistics loading unloading types are there. So, the
setup cost is less as you move to right and your holding cost increases.
216
So, this data is very particularly taken to highlight this very aspect of this supply chain, that what
physical changes are taking place as you are coming left to right in the supply chain. So, now
when we have the data with us so I request all my participants that use this data for separate
optimisation as well as for simultaneous optimisation. You need to see that how can we have the
separate and simultaneous optimisation.
And to understand the use of this data for the problem solving purpose, Let me make a table and
in that tabular arrangement, you can understand that what type of data you can fit in and how do
you know that what type of improvement or the modification with different types of calculations
are being done. So, this table can be made like this way and you have direct comparison between
separate and simultaneous.
Now, the first thing is in both these model how do we operate in both these models. The first
thing is the calculation of Q2, that is the first thing. In both these cases the first thing, the starting
step is calculation of Q2. In the case of separate optimisation, the calculation is pretty simple, you
just need to use the old formula, which we have discussed under root 2d k 2 upon h2. So, use this
formula for getting your values of Q2 here.
And, use of these values will give you Q2 = 200. And, then the formula involving the higher
values of Q2 we will see that how do we use that, that will come here. Then, the second thing is
to determine the value of n* and, the value of n* is in this particular case will be k 1 upon k2 into
h2 upon h1.when we are doing the separate optimisation
And that value will come under root 15. Now under root 15 is a fraction and now let us see how
do we round off those fractions into our this tabular arrangement. So what we can do if this n* is
a fraction, so there are three rules for that purpose. If n* is less than 1 than you take n* equals to
1, and if n* is more than 1.
217
So in that case we consider [n*], 1n in this bracket which is the highest possible integer just a
smaller than this n*. So that you can write this expression for an example if this n* is 2.57, so
this will be 2 and this will be 2+1= 3 and now will have this comparison where you take the
ratios of these terms, n*/[ n*].
Now, if in this ratio if left hand side term is less than or equal to right hand side term. Your
round off will be on the lower side and in case of reverse if left hand side if it is this way. Here in
this case n equals to this and otherwise n equals to n* bracket +1. So, by this way of rounding off
you can round off the available values of n*.
Apply same in this particular case, so then you will find that round off value of n is coming to be
4. When you do this type of calculation, you will see that round off value of n is 4. And, when
round off value of n is 4 then you can determine the value of Q 1 = n Q2.Q1 , so it will come 800,
4 into 200 it will come 800. So now in this separate optimisation, you have calculated both these
values Q1 and Q2.
And as a result of Q1 and Q2 you can also calculate total cost of inventory C, and you can use our
earlier expressions which we discuss in our last session of inventory management that what will
be by total cost of inventory. So, that C can be calculated here. We can have the direct
calculation which will tell as C will be 1950, the value of total cost of inventory in this particular
case will be 1950.
So, this is for the separate optimisation. Now when I want to do these simultaneous optimisation.
Because, calculation of Q2 you see here I have done in a very sequential manner, first I
calculated Q2, I determine the value of n and there is no relation of Q 2 and n and then with the
help of the value of n. I determine the Q1 and finally C, but now in the simultaneous optimisation
you can recall just now that the calculation of Q2 involves the value of n.
So here my first step in this case will be this stage. Here it is the first and here it is the first, so
first I will determine the value of n* and it will k 1/ k2. e2/ e1. So what is e2 and e1. e2 will be the
218
difference of e2 is on the right hand side. So e 2 = h2- h1 that is 3-2 that is 1 and e 1 =h1 that is 2,
e2= h2- h1. So 3-2, e2 is 1, h1 = e1. So that e1= h1 that is coming to here.
k1 is 1000 divided by k2 that is 100, so when you solve this it becomes under root 5, so this is
under root 5 and when you solve it under root 5 means when you round off under root 5 you will
see that this will come to 2 and when now this n=2 will be used for determining the value of Q2.
So, now actually the calculation of Q2 in the second case is my third step and when used I this
formula for simultaneous optimisation the value of Q 2 using this value of n will be 379. Value of
Q2 using this expression is 379, and then again Q 1= n Q2, so 379 x 2 =758 is my Q 1. And using
these values of Q1 and Q2. My total cost of inventory comes to be you all can calculate and
directly giving to the values that is 1897. Now you can see now this question is done.
Now, let us have some physical understanding some interpretation of these data, because that is
more important for managers to analyse the data properly. So, this data this information there in
front of us, Now there are certain very revealing factors out of this solution. One which is very
simple to understand that because of simultaneous optimisation, I got the benefit here the cost
was1950 here the variable cost has reduced and it is 1897 only.
So, that is one direct visible benefit tool, but now you see another challenge I am the stage1, and
as stage1 I am the owner of entire supply chain. Now, stage 2 is my retailer and in first case
retailer is keeping inventory of just 200 items. But now I am moving to simultaneous
optimisation the inventory level of retailer increases to 379. Now, as a manager it is very difficult
for me to convince my retailer.
That you increase your inventory level from 200 to 379, and therefore many a times because a
supply chain is fragmented also, on one side in a supply chain we say that we all are connected.
But, we all are owners of a particular entity also. So, it is a fragmented discussion also, so here
you can see the retailer who is the end point in my supply chain that retailer will say that.
219
I am not ready to stock 379 units earlier my stock was only 200 unit why should I stock more. So
developing this understanding in your entire supply chain that know it is beneficial for all of us,
if you stock more so the overall cost of inventory will be less. It is going to help the entire
supply chain to increase it is surplus. Because, ultimately this total cost of inventory is reducing
the profitability of the entire supply chain.
So, developing this idea requires lot of efforts it is much difficult than what we are discussing in
the class, so this model obviously because of this net gain gives profit, but at the same time you
need to have that label of trust that label of understanding between 1 and 2 stages. So, that if this
value is increasing and mostly in almost all the cases, You will find a substantial increase in Q 2
at the simultaneous optimisation process, but the overall cost decreases.
Overall cost of inventory decreases and therefore this model better captures the supply chain
effect and reduces the cost of inventory and therefore it is advisable that we need to build that
trust we need to build that confidence, between stage 1 and stage 2. That what stage 1 or
whoever is owning the supply chain means we know that this is the supply chain of dell.
This is the supply chain of Amazon, this is the supply chain of Walmart, this is the supply chain
of Apple, so there is somebody who owns the supply chain, and then in that supply chain there
are many smaller components, many smaller installations and all those smaller installations if
220
they are not able to understand the philosophy, the way, the owner of the supply chain owns to
run it, it is very difficult to achieve these types of simultaneous optimisation cases.
And, then people will keep on doing their own individual optimisation of inventory, which may
result in a very catastrophic type of situation of bullwhip effect. So, we stop in today’s class at
this point, where we have discussed the simultaneous optimisation with the help of one
numerical problem, that how it can improve our results. In our next class we will discuss this
optimisation for a multi echelon problem.
Where more than 2, it is a very simple case, only 2 stations are there, but in real life we know
there will be many, so we will take some case, where you have more than 2 installations and how
do we operate with the simultaneous optimization of those multiple installation in the supply
chain. Thank you very much.
221
Lecture-19
Multi Echelon Inventory Management for Four Stations
Welcome back, we are discussing about inventory management in the supply chain. And in the
inventory management discussions in last 2 sessions we have discussed about a very simple case
of inventory management in supply chain where we had only 2 installations in the supply chain,
and we used our very conventional EOQ model to handle those 2 installations.
And we saw with the help of an example also in the last class that how when we do the
simultaneous optimisation even in the case of just 2 installations we have some type of cost
saving. The very well cost saving in a supply chain environment. And when we do the individual
optimisation it is not possible. Now the same discussion in this class we will take to multi
echelon systems, where in a supply chain more than 2 installations are possible.
And when more than 2 installations are possible you can have now, a supply chain where we can
think of 4 stations, 1, 2, 3, 4, and as we are moving from this side to this side we know that it can
be a supplier, it can be a manufacturer, the wholesaler, and the retailer. So, these are normally the
4 components which are available in a supply chain, the supplier of raw material, supplier of the
222
components, your other type of part suppliers. Then some kind of manufacturing or assembly of
those components may take place here.
And, these are the wholesalers or the dealer, which are taking products from the manufacturer,
and finally customers get their products at the retailers end. So, this is a popular supply chain,
generic supply chain model. And, we have discussed in our first sessions about this type of
generic supply chain. Now, considering this type of generic supply chain, and using the points
which we have already discussed in earlier 2 sessions.
About 2 installations of inventory management, now here retailer requires Q 4 item, the
wholesalers requires Q3 item. The manufacturer require Q3 item, and supplier has demand of Q 1
item at particular time. Now, it is very well understood that the retailer is getting it supplies from
wholesaler. Wholesaler is getting it supplies from manufacturer, and manufacturer is getting it is
supply from the supplier or the vendor.
Therefore, Q3 is responsible out of Q3 items a part of that will go to retailer in the form of Q 4.
From Q2 items a part of that will go to wholesaler in the form of Q 3. And from Q1 a part of that
will go to manufacturer in the form of this supplier. So, therefore Q 1 needs to be bigger or at best
equal to Q2, and so on for other stages also, Q2 needs to be greater than or equal to Q3, and Q3
needs to greater or equal to be Q4.
So, this type of relationship is to be there, between various quantities which you are procuring at
a particular stage. Because, a particular stage is responsible to supply products to the next stage
in the supply chain. That is one very important thing when we will develop the model for
inventory management. In this type of situation you always need to see that Q 1, Q2, Q3, Q4 or if I
take it to Qn to this level.
This type of relationship that the quantities which you are procuring which you are receiving at
predecessor stage, should be more or equal to the quantities which are required at the receiving
end. And now the other important thing which we have already discuss in the 2 installation
system. The holding cost, because as I moving from 1 to 4, I am doing the value addition, so this
223
is the direction of value addition. So, therefore my holding cost which is proportional to the these
value addition h1, h2, h3, h4.
So my holding cost because of increase in value, because of more value I am adding as I am

moving from left to right. So therefore holding cost will also increase as I moving from this to
this station. Now with this if I go to develop the model for inventory management. Let us see
whether we have sufficient conditions to apply our conventional EOQ model. Now, at this point
I will like to recall what we discuss in the last 2 sessions.
That to apply EOQ model the inventory consumption, and replenishment at a particular stage
must follow, if you recall you can tell me the sawteeth pattern, sawteeth pattern must be
available for applying the EOQ model of inventory management at a particular stage. So, now
will let us see whether the sawteeth pattern is available in this particular case or not. And, then
only the basic EOQ model can be applicable for inventory management.
In this particular case now, let us that at stage number 4, the last stage where we have procuring
because this knowledge we have already from our earlier discussion. That in a 2 stage model we
are procuring inventory at other stages in relation to our last stage, so here the last stage is fourth
one. So, I will procure inventory at other stages, in relation to this fourth stage. So, what I am
trying to say that Q3 should be sum multiplier of Q4, Q3 = ni.Q4.
Q2 should be sum multiplier of Q3, Q2= ni.Q3 and so on Q1 should all be sum multiplier of Q2,
Q1 = ni.Q2. So, all these inventories at other stages are multiplier of their successive stage, and if I
see, if I substitute the value of Q3 here in this case and Q2 here in this case, I can see that I can
represent Q1, Q2, Q3 all three in terms Q4. So in this case also my complete inventory
management will be done with respective quantities which I am receiving at my fourth stage.
Now, at the fourth stage I am requiring this Q 4 items. Now these are the Q4 items, and I receive
Q4 items from stage 3, and I start consuming these Q 4 items. Then after sometimes I will
consume these, and I have adjusted my lead time, I have adjusted my order point. In such a
fashion that as soon as I am touching this 0 level, I get a fresh supply of Q4 items.
224
And, then I will start consuming these Q 4 items, and this way my second cycle will be completed
at fourth stage. Again, I will receive a new supply, and this will go on. So, you see we have a
very clear sawteeth pattern, these vertical lines are representing the replenishment of my stocks.
And these slant lines are representing my consumption. So, at stage 4 you have a very perfect
case for applying the basic EOQ model of inventory management.
Now, let us come to stage3, in a stage3 you require Q 3 items now, what we have done here, a
very specific case now, what we have done, just to make it simpler, I have taken the values of n i,
ni', and ni'', equal to 2. Just for the purpose of our understanding, all these values are 2, 2, 2 each.
So Q3 is actually 2 Q4. So, all these values I have taken equals ni to ni' equals to ni'' equals to 2.
So, these things make me Q3 equals to 2 Q4. Now, I am receiving Q3 supplies here. And, out of
Q3 supplies I will immediately supplied, because I have adjusted the receiving cycle in such a
manner that on day 0, when Q3 is supplied to stage 4, the same day I receive Q 3 supplies at stage
3. And out of Q3 since Q3 is 2Q4, a part of that part of Q3 will go to stage 4. So, what is remaining
with me Q3- Q4, that is with me, and that will go with me till this particular time.
And, you can see that by this time I will be carrying this Q 3 – Q4 stock in my warehouse, and I
will be continuing with this, and then at the second replenishment cycle at stage 4. When again
to Q4 items will be required these Q4 items will be shift at this point, and then I will not have any
inventory for this period. Then again I will receive a new supply of Q3 items.
But, out of that I will immediately supply this is Q 3, this point represents Q3, but out of that again
I will immediately supply that Q3, Q4 portion to the stage 4, and I will be left with Q3- Q4 here
again. And, I will carry this inventory for one more cycle of stage 4 and then this inventory will
come to 0. So, with this way you will see that I have these types of step curves. This is Q3 level,
immediately Q3 comes to Q3- Q4.
I will carry this Q3- Q4, for some time, and then my stock will come to 0 level, and then I will be
having 0 stock, at stage 3 or this much period. This much period is represented by one cycle of
stage 4. And, then again when fresh supply is required at stage 3, I receive Q 3 items here, and out
225
of Q3 items Q3 – Q4 remains in the physical stock at stage 3. Because, Q4 has gone to stage 4, and
I will carry with Q3- Q4 supplies, for one cycle of stage 4.
And, then when the next cycle requires the fresh supply of Q4 items, so Q4 items are available
with me, and I will supply these Q4 items to stage 4. And, the stock will again come to 0. So, I
will remain 0 stock for one more cycle of stage 4. And, then again Q3 supplies will come, and
this type of step pattern, will keep ongoing. You will see that this type of step pattern will keep
on going.
Now, these lines, green line represents the installation stock, last class also we discussed this that
this is the installation stock. So on here also this is installation stock. That what is available at a
particular stage 3 or stage 4. But, we already introduced a concept known as echelon stock in our
last class. So now echelon stock just to revise that echelon stock represents the amount which
you have at that particular stage at stage 3 and at stage 4.
So whatever you have, so if I talk of echelon stock of stage 1, so whatever physical inventory
you have at stage 1 + 2 +3+ 4 at a particular time whatever inventories you have, physical
inventories you have at stage1, 2, 3, 4 that is the echelon stock for stage 1. For stage 2 what
physical inventories you have at stage 2, 3, 4. That determine the echelon stock for stage 2 and
so on for stage 3 and 4.
So, if I talk the echelon stock for stage 2, so you will see that here it is Q 3, and here it is 0
initially, so the Q3 is the total stock, which is available with me, and at a particular time if I talk
of this particular time, if I talk let us say a particular point here. So, at this particular point here,
you see the physical inventory at stage4 is this much. This is the physical inventory available at
stage 4.
And, this is the physical inventory available at this stage. So, I want to see the total inventories of
this + this, this point will come somewhere here. And so on, if I want to see the total inventory at
this particular point, so here it is 0 already. And, this much inventory is available at stage 4. So,
226
this point will come here. At this point you see the total inventory at stage 4, as well as at stage 3
is 0.
So, if I combine all these points of echelon stock I will get a line like this, this is the line which
represents the echelon stock at stage 3. The stock, which is the physical inventory at stage 3 +
stage 4. So, it is the echelon stock, in case of stage 4, since there is no stage, after stage 4, so
whatever inventory is there at stage 4. That is the installation stock as well as the echelon stock.
So, in case of stage 4 this curve is making you installation stock, as well as echelon stock.
So it is same here, but in case of stage 3, these step lines are representing the installation stock.
But, this type of curve where you have a slant line of consumption, this represents the echelon
stock. And, then if you see for at any cycle, you can have this type of curve represented by the
dotted lines. This represents the echelon stock. So, you can see that we can apply now EOQ
model for stage 3 also.
If we consider it is echelon stock we can apply the EOQ model here also. And, now take this
case forward you have stage 2 now, now you take the stage 2, in stage 2 you have Q 2, which is 2
Q3. Q2 is 2 Q3. So, now out of Q2, whenever you procure Q2, so to initiate we have develop the
system in such a manner, that whenever you require Q3 on time t0. The same day you have
procured Q2 also.
So out of Q2 half of Q2 that is Q3 will go to stage 3. So, you are left with Q 2- Q3, and you will
continue with this Q2- Q3 for this much period. The time is second requirement of Q 3 comes from
stage 3, and then it will come to 0, and then again you will remain at 0 level the same discussion
which we have applied for stage 3 will be applicable to stage 2 also.
So, again you can think of this type of echelon stocks at stage 1 also. You can think of these slant
lines which are representing the consumption of echelon inventory, these slant lines represent
the consumption of echelon inventory. Echelon inventory again is the inventory available at
stage 2 + stage 3 + stage 4. So, actually this slope is coming because of this consumption at stage
227
4. At stage 2 and stage 3 you have the horizontal lines. But, the angular lines are available only at
stage 4.
So, because of the angular lines at stage 4, you get the slopes in stage 2 also. And, therefore,
these installation stock which is represented by these step lines. And, the echelon stocks are
represented by these slanted lines. So, this is again showing the case, that this slant lines make us
possible the use of EOQ model, at stage 2 also. Now, come to finally a stage 1 also.
The same explanation can be applied to stage 1 also where you are procuring Q 1 items, in such a
manner that Q1 is equals to 2 Q2. And we have adjusted the procurement cycle of stage1 in such a
manner that whenever there is a demand of Q 2 items. The Q1 items come on the same day. So, on
time t0, you receive Q1 items at stage 1, and out of that immediately Q2 items are shift to stage 2.
So, you are left with Q1 – Q2 items.
And, you will continue with Q1 - Q2 item for one complete cycle of stage 2. And, then you will be
with 0 inventory for another complete cycle, and again by using the concept of echelon stock,
you can have this type of dotted lines to represent the echelon inventory at stage 1. So, now you
see by considering the concept of echelon stock at stage 1, stage 2, stage 3, and obviously in
stage 4. Echelon and installation stocks are same.
So, here we have that because of these echelon stock concepts in these 4 stages. You can very
well apply the concept of EOQ model. So, we will see the model with the help of echelon stocks
the help of echelon stocks with at all these stages, and that will help us in using the basic EOQ
model. Here if you see that we have used the values of ni, ni', ni'', all two.
So, that will help us in very uniformly completing the replenishment cycles. The 2 replenishment
cycles of stage 1, makes one cycle of the stage 2. 2 replenishment cycles of the stage 3 makes
one replenishment cycle of the stage 2. And, 2 replenishment cycles of stage 2 will make one
cycle of the stage 1. And, therefore this will also enable us that I can write all these terms, all
these Q1, Q2, Q3 in terms of my Q4. So, here you see that Q 3 can be represented that we have
already done that.
228
It is 2 Q2, and because 2 Q2 is 2 Q3, so I also write Q2 equals to sorry this is Q4, here it is Q4. Q2 is
equals to 2x2 Q4 that is 4 Q4. And, 2 cycles of Q2 represents one cycle of Q1. So, Q1 is actually 2
Q2. So, here I can also write Q 1 is equals to 8 Q4. So, I have represented all these Q 1, Q2, Q3 in
terms of Q4. Q3 is 2 Q4, Q2 is 4 Q4, and Q1 is 8 Q4. So, therefore the point is that in case of I want
to generalise this discussion.
I can write Qn as my last stage of supply chain. This is the nth stage of my supply chain. And, I
can write all the stages, let us say it is Qi in terms of my values of ni, this is how I can represent
all my stages with the help of the last stage. All my stages with the last stage, so this is very
important that here, when we are going to develop the multi echelon inventory system. In this
multi echelon inventory system we will be determining these 2 things only.
We will determine the value of Q1, the value which is required at the last stage and then for the
subsequent stages. Here we have assumed for the sake of simplicity we have assumed that all n i's
are 2. So, that is why have this type of very uniform representation. But in our other cases we
will determine that whether it is 2 or 4 or 6 or something else.
And accordingly Q1, Q2, Q3 can be determined directly determining the value of Qn only. So, now
one more thing before we close this session, one more thing will like to discuss. Now Q i can be
ni into Q1. Now, Ni as we have discussed earlier also in the case of 2 installations, should be any
integer value. So, it can be 1, 2, 3, 4, 5, 6, 7 anything, and if you remember the example which
we took to understand the case of 2 installations.
In that we took deliberately the value n i as equals to 3. Q1 was 3 Q2, in that case. But, here we
have a limitation and for that limitation we will use a kind of relaxation of this problem. We will
like to use a relaxation of this problem. Now, the limitation is that if for the purpose of this
problem. Here ni is 2, and here ni' is 3, and here ni'' is 2. Let us say these are the values, 2, 3, 2.
In that case it will not be possible for me to make this type of groupings of these cycles. Because,
of odd and even values of ni's I cannot make these type of cycles of my supply chain. That all
229
these cycles are starting and ending, in the same manner. You have the same starting date of all
the cycles, and 2 cycles of this will make 1 cycle of stage 2. And, 2 cycles of stage 2 will make 1
cycle of stage 1.
But, if it is not so, in that case it will be difficult for me to develop these types of relationship 2
cycle will make 1 cycle of this. But 3 cycles of stage 2 will make 1 cycle of stage 2 and then it
will be difficult for me to have same starting and ending date. And, as a result of that either I will
be having more inventories at this stage or I will be running out of stock, for 1 complete cycle if
I want to match this type of analysis.
So, for that purpose we have done a bit of relaxation in this relationship. Where Q i equals to ni Qn
we have used. Instead of that we will be writing ni as 2 to the power mi. And, mi can take any
value, mi can take any integer value 0, 1, 2, 3, 4, 5. So, what will happen because of this
relaxation, because of this additional condition we have imposed, you will get the values of n i,
only in the even values.
If mi is 0, 2 to the power 0, it is 1. Then 2 to the power 1, it is 2. 2 to the power 2 it is 4. 2 to the

power 3 it is 8. So, only even values will be there. 1, 2, 4, 8 like that and, that will make our life
very simpler, and therefore this type of analysis will be possible to handle the supply chain more
effectively. We can go to a bit of mathematics and Roundy was one of the professor who
suggested that if we use this type of relaxation.
There will be slightly higher values of the overall cost, but those values of higher cost will not be
more than 2 %. So, that is very well justified, because of lot of comfort we are going to receive
by doing this type of simplification. So, here we stop in today’s lecture, and we will see the
application of this type of model in our inventory management. So, in the next class we will see
that now we are clear that we are going to use this type of relaxation.
And will have application of EOQ model at all 4 stages. So, now with the help of some data,
with some practical exercise, we will see in our next session. That how do we apply this model in
a multi echelon situation. Thank you very much.
230
Lecture-20
Multi Echelon Inventory Management for Four Stations
(Numerical Example Continued)
So, welcome back in our last session, we were discussing about inventory management of multi
Echelon systems where we have discussed a particular case of 4 installations. And we discussed
that with the help of echelon inventory concept. We can apply at each of these stages, the
concept of basic EOQ model.
Now going further into the discussion, and, having a particular case, now let me have some data
with us and with help of that data we will like to see that how multi echelon model can be
applied to that situation.
Now we have four installations 1, 2, 3, 4 and for each of these installations the setup cost, the k i
is 250, then 6, then 30, and 110. And the holding cost which is h i is 0.50, 0.55, 3.55 and 7.55.
The demand of the supply chain D is 4000. This data is available to us. Now with the help of this
data, one thing which we were discussing in the last session, that is very much visible that h i, that
is the holding cost.
231
As we are coming from 1 to 4, left to right in the supply chain 0.50, 0.55, 3.55, and 7.55. So you
can see these holding cost are continuously increasing. So, therefore, this data is inline with what
we have discussed. Now as we said that to apply EOQ model in this particular case we will
consider the case of Echelon inventory, not the installation inventory. These are the holding cost
given for the installation inventory. So now we will have the concept of e i, that is the concept of
Echelon holding cost.
Not the installation holding cost, and that is in case of installation 1 it will be simply h 1, for
installation 2, it will be h2- h1. For installation3 it will be h3- h2. And for installation4 it will be
h4 - h3. And therefore ei's will be 0.5, 0.55 – 0.50 will be 0.05, 3.55-0.55, will be 3.0 and, 7.55 -
3.55 it means 4.00. So, these are the Echelon holding cost and now using the concept of echelon
holding, and the setup cost.
We will have the formula for calculating the cost of variable cost of inventory, at each stage, so
that is Ci. That is for a particular stage in the supply chain, the cost of inventory will be d that is
ki upon Qi + ei Qi by 2. This is the cost of inventory, variable cost of inventory at particular stage.
Now when you use this formula and apply the concept of basic mathematics for getting the
maximum, minima differentiated.
Apply the maxima, minima and then calculate the value Q i , because that is the only variable
here. So Qi will be under root 2d, k i upon ei. I request all the participants to please practise this
and get this formula on your own. I am directly giving this formula here. Now in this formula,
there is a term which is very important for us in further calculation. This is this k i upon ei, the
value of Qi will depend on the ratio of k upon e.
232
The ratio of k upon e will determine the value of Q. Now as we discuss in the last session, at
stage 4, the quantities is which we are getting are coming from 3. At 3 quantities are coming
from 2, and at 2 quantities are coming from 1. So you need have highest value for Q 1 , then Q2,
then Q3, and the least values will be for stage 4. So, that is the way. Now this ratio will play very
important role in that case.
ki/ei, if the ratio of ki/ei, and the ratio of the next term that k i+1/ ei+1. If you see this comparison k/e
for 2 stages ith stage, and the subsequent stage the successive stage in that. In this case if k i/ei is
more than ki+1/ei+1, what will happen Qi will be more than the next step, that is desirable or the
last possibility is it is equal to this. So, in this case Q i will be equal to Qi+1.that is also possible in
cases.
That if Q, whatever is coming at stage 3, the same you are transferring to stage 4. Can you recall
any supply chain situation where it happens, that whatever you are receiving the same you are
transferring to the next stage. We can think of situation like cross docking. In cross docking this
type of situation may emerge that whatever you are receiving at particular stage.
Because, cross docking we know we do not keep inventories. Whatever we are receiving the
same just by repackaging the size of packets may vary depending upon the requirement. So,
cross docking is a kind of aggregation. So, just to take you off from this discussion. In case of
cross docking, what happens this is a place, where cross docking is taking place, cross docking.
So, you are getting supplies from the manufacturers, so bigger packets are available A, B, C and,
then we just make much smaller packet of A1 + B1 + C1, A1 + B1, A2+ C2, B2 + A2 +C3 of variety
of combinations. Depending upon the requirement of retailers, so we will just take the bigger
supplies from the manufacturer, which are more homogeneous, which are uniform, and as per the
requirement of the retailer, we will mix the supplies of different wholesalers to get the order of a
particular retailer.
So here whatever is coming, because stocking has the vary point of cross docking is that you do
not keep the physical inventories. So total input is equal to total output, on a any particular day
233
total input is equal to total output, no inventories are there. So, that is the situation where, you
can think ki upon ei can be equal to divided ki+1 upon ei+1.
Because in that case whatever is coming at particular stage will be given to the next stage. So, in
such situations this type of calculation may come, but in most of a situation which we have
discussed in our earlier examples, we will keep some kind of physical inventory. But, now a days
it is of much significance, that none of the partner.
None of the partner wants to keep the inventory, and for that purpose the systems of cross
docking are becoming more and more relevant. And at the same time we also have the influence
of just in time. And in just in time of systems we normally do not want to keep inventories. And
therefore whether you talk of JIT, or whether you talk of cross docking. In both these situations
whatever is coming the same is going.
No one wants to keep the physical inventory for a particular period at its stage. So in current
example, in current situations it is quite possible that we have in practical cases more and more
those types of examples where Q1 equals to Q2, Q2 equals to Q3, Q3 equals to Q4 and so on. But if
I talk of a traditional type of a system particularly in case like India, because many a times we
see we keep inventory to avoid fluctuations of demand and supply.
But in that case inventory becomes a slight typical issue where we need to maintain a particular
level of inventory, So that we can achieve the requirement of the customer and we can achieve a
particular level of service, but at the same time because of new and new types of challenges
coming to us. We do not know how the demand will behave in the future, and therefore we do
not want to keep higher inventories also.
Like you see in our earlier case the example which we discuss in the last session at stage 1 the
inventories where 8 times then the stage 4. A stage 4 was keeping just Q 4 at any time. At stage1
was keeping 8 times of Q4 in our last example. So now the stage1 is blocked for 8 replenishment
234
cycles of stage4. So now a stage1 the point I am trying to say the stage1 will be able to consume
this much is stock of Q1 item.
In 8 replenishment cycles of a stage1, because in each cycle the quantities which are received by
a stage4 or 1 by 8 times of Q 1. So, it may happen that after two replenishment cycles or three
replenishment cycles, demand all of a sudden changes and therefore the whole system of
inventory at 2, 3, 4 will be at loss. So, there is a challenge for the management also that how to
handle these uncertainties.
And, therefore we always look for some kind of optimum values and therefore because more and
more uncertainties are coming. You will find these types of cross docking examples, you will
find that people believe in JIT, people believe in flexibility. Here yourself can understand that I
am waiting for 8 cycles of a stage4, 8 times the order is procured at this stage 4 than a single
order is procured by a stage 1.
So these stages 1, 2, 3, 4, 5, 6, 7 and 8 these 8 cycles of stage 4 represents the one cycle of a
stage1. Now stage 4 after the third cycle customer says no we do not want this product, we want
a different type of product which should have this much feature. But this stage1 is still having
this much inventory, so because now customers are not willing to purchase this product the
whole inventory in the supply chain not only a stage1.
But at stage2 also there will be some inventory at stage3 also there is some inventory that
complete inventory is at loss, and therefore this is a big challenge in front of us that how to
address these issues and probably more real time analysis will help us to reduce this type of
relationship. We need to have more relationship that we can minimise this period.
And you see this is a case where only 4 such stages were there if it is the even longer supply. In
case of India where our large number of population is living in rural areas more than 60% of
population in rural areas and in rural areas the supply chains are typically very long typically we
have very long supply chains in the rural areas and therefore this value may further increase it is
8 times of your Qn.
235
But if I talk of a rural supply chain the values may further increase and then this problem will be
even more severe in that case. So therefore it is very much required that we need to have some
better models and we need to understand that what is this significance of this particular aspect
that you cannot have very high values of this n factor here, because more higher values.
You have that much rigidity in your supply chain it will be difficult for your previous stages to
change their products, because they have already stock that much item at their stages. So we
need to have smaller values we love a smaller values of this factor, So that our supply chain is
more lean, there is less inventory at each stage.
And that what we are going to discuss in this particular case that the inventory should be as less
as possible and the value of n is one important factor in that. Now coming further with respect to
this particular problem. Now here we see we have understood the meaning of k/e, now let us
calculate k/e for these stages.
So, here it is 250 upon 0.5. So the value of this comes to be 500, then 6 upon 0.05, this is 120,
then 30 upon 3 that is 10, and then, 110 divided by 4, so this is 26.5. Now you see the values 500
no problem we cannot say anything about here, but the next value is less than this first stage, so
236
this is okay. The next value is 10 which is much less than 120, absolutely okay. The next value is
26.5 not possible this cannot be there.
This value is not suitable because this is more than 10 and I cannot accept because, if I take it so
the Q4 will be more than Q3 and that is unacceptable, Q4 more than Q3 is unacceptable. So this is
a problem, now to sort out this problem we will do that we will merge a station 3 and 4, we will
merge these two installations 3 and 4. The meaning of merging is that you will have same value
of Q3 and Q4 finally.
But for the sake of solving this problem we will merge a stage 3 and 4, and we will recalculate
the values of k/e after merging this. So when we merging this you have a stage 1, 2 and with is
you have also merge this stage 4, so you have new values of k i's, ki will be 250, 6 and now we
have merge stage 3 and 4, so therefore this k i will be 30+110 that is 140, and similarly you will
merge the values of ei's also.
So, ei will be 0.50, 0.05 and this 3 + 4 will become 7, and now you again calculate k/e. So, this is
250 by 0.50 same 500, 6 upon 0.05 the same 120, and 140 upon 7 this is 20. So, now you see 120
and 20, so you have a pattern that the ratios of k/e in this particular case are following the
required condition, and now we can apply this model the results, because now then we have
reduced the 4 installation problem into a 3 installation problem.
And for that purpose the results will be you need to calculate Q1 and Q2, but whatever we will
calculate for this stage, the meaning is that Q3 equals to Q4 , that is the meaning of the result there
will be Q1 and Q2, and Q1 and Q2, will be the multiplier of obviously Q 3 or Q4 which ever you
say, but Q3 and Q4 will be same or other terms you can say Q3 is equals to here the value of ni.Q4.
This is equals to 1 this is how you can understand, and then you can go for calculation of Q 1 and
Q2. So now let us start solving the problem using this final table and for that purpose, first we
will determine the starting solution to initiate the problem. We will determine the values of Q i's
with respect to these different 3 stations, so you this formula simple formula.
237
And using this simple formula you can calculate the values of Q 1 which will be under root 2d is
already given to us that is 4000 into ki that is 250 upon 0.50, so on for Q 2 and for Q3 or Q4, so
you will have these calculations and these calculations will give you the values of this will come
as 2000 the value of Q2 when you apply k/e as 120, so this will come as 980 and for stage3 and
4 this will be 400. So this is the initiation of the problem.
This is the initiation of the problem, the first solution which we get that is this Q 1 is 2000, Q2 is
980 and Q3 Q4 is 400, but now you can see on your own once you have calculated these values
and the meaning is Q3 is also 400 and Q4 is also 400. This means Q3 is equals to Q4, now 980 is
not direct multiplier of 400, and 2000 is also not direct multiplier of 980. So the condition of that
Qi is equals to ni .Qn is not satisfied.
This condition is not satisfied here, because you are not able to achieve an integer value of n i that
is first thing and second relaxations which we discussed that it has to be 2 to the power m i that is
even not satisfied, because none of them is even to closer this condition, so it means whatever we
have done this is the phase1 of the solution process or you can say this is the initialisation of the
problem.
That you got some initial value, now my request to you that my request to all participants is that
you calculate the total cost you need to calculate the total cost on the basis of this, so you need to
238
have this Ci and if you do the total cost C that is the sigma of Ci equals to 1 to n, so you calculate
this total cost. This is you can say this is the lower bound of our solution process.
The meaning of lower bound is that our solution cannot have a higher cost, lower bound means
this is the worst solution and the meaning of a solution in our cases that this is the highest cost,
this is the maximum cost you will incur in managing the inventory in this particular case, but
now in our second phase we will try to improve this value of C lower bound.
The total inventory cost, lower bound will like to improve this lower bound to a better value and
for that purpose, we want to use this initial value. Now we want to achieve this type of
relationship. The objective is to achieve this type of relationship and when we have this type of
relationship probably we will have a lower values of total cost and those lower values of total
cost will be improved solutions.
We will give you the improved solution and we will see for that purpose what to do, now for that
purpose to initiate our phase2 process, we will see first that this Q i which is ni into Qn in this
particular case we have to determine the values of ni and ni's are if I am at a stage2, so there will
be n2. n3. Q4 if I am at a stage1, so there will be n 1. n2. n3.Q4 and if I am at a stage 3, so Q 3 equals
to n3. Q4.
So, we have this type of arrangement. Now we can simplify this calculation by understanding by
this particular multiplying factor, we can simplify our calculation instead of writing n 3, n2 into n3,
n1. n2. n3. I can simply write one factor pi, and pi is nothing but the multiplier of these things as
per the stage. pi value can be pi if I am talking of stage 3 it will include only n 3, if I am talking
stage 2 it will include n2 into n3.
If I am talking stage1 it will be n 1. n2. n3, so in general pi will be nothing, but the ni.....n,n-1 this
will be pi. ni into ni-1, and upto n,n-1. And, with the help of this we can generalise the relationship
as Qi equals to pi. Qn, so this becomes our required relationship that Q i equals to pi. Qn and now
we will say but how to determine the values of p i's and where to get the optimum values of p i's
for this solving purpose
239
And, with the help of that we will be able to determine that yes, now the solution is being
achieved, now let us stop in this class here and in our next class we will see the phase 2 of the
solution process where we will be determining the values of pi's for different stages and then we
will see how to stock in that particular condition and how to complete this solution Thank you
very much.
240
Lecture-21
Multi Echelon Inventory Management for Four Stations (Numerical Example
Continued)
So welcome back in our last session, we were discussing about multi Echelon inventory
management and we discussed a problem where we have 4 installation in our supply chain
and we have the cost of setup for each of those installations, we have the cost of holding at
each of these installation and then we discuss the concept of Echelon holding cost and this
table be prepared in the last session.
And we also discuss about the important ratios for that purpose that ratio is of K/e. We
calculated the ratio of the setup cost and Echelon holding cost and we saw that the
consistency of these ratios are not there at stage 3 and 4. Because we want that the K/e should
be lower for a successive stage, because value of K/e determines the quantity you are going
to order at a particular stage.
Now since at stage 3 and 4 if you recall in the last session we had the value of K/e is 26.5 at
stage 4 and at that stage 3 it was 10. So what is the meaning that Q 4 was more than Q3 and we
have already discussed that we will withdraw, we will take the amount of Q 4 from Q3. So it is
never possible to give more amount from the less, you can give from a bigger lot you can
take a small allowed, that is understandable.
241
But from a smaller lot you want to withdraw a bigger lot it is not possible and therefore for
the modelling purpose we combined these two stations in our last session, station 3 and
station 4 and then we rewrote this table and in that class itself we got these values that the
value of holding cost at installation one and installation 2 remain as it is, but we added the
values of holding cost for installation 3 and installation 4 and it became 140.
Similarly the values of a Echelon holding cost for remain same for installation one and
installation 2 but for installation 3 it was 3 and installation 4 it was 4. So the total Echelon
holding cost for this combined installation became 7 and now we calculated this K/e again
and then we found that there is a proper consistency. The meaning of consistency again
simple K/e should be largest for installation one.
Then it should be less for installation 2 and then it should be further less for installation 3, 4
combined states. If this type of continuity is there, if this type of relationship is there for
ratios at 1, 2 and next stage, then we say the K/e is consistent, if it is not so then again we
need to combine those installation and then proceed further, without having this consistency
this is a very important issue that first we need to see whether data is consistent or not.
Without this consistency we cannot proceed further. Now when this consistency is there this
problem will be handled in two phases, this problem will be handled in two phases, in the
first phase of the solution we started in our last session the phase 1 of the problem we will
start calculations of values of Q i's, for these three different installation, now you simply
consider that instead of 4 installation.
We have three installations in this problem and in the last when we will be finalizing the
solution at that stage we will see that how this solution of 3 installation will be converted into
a solution of 4 installation. So at the moment we will only talk of 3 installation 1, 2, 3. Now
we are going to solve for these 3 installations. So you will calculate Q 1, Q2 and Q3. And these
Qi's will be calculated using our basic EOQ formula.
And that basic EOQ =√(2dKi/ei) and use this formula, so Ki in the case of installation 1 will
be K1/e1, here it will be K2/e2 and here it will be K3/e3. And when you calculate these values in
Phase 1 the answer will come 400 then 980 and then 2000. These are the values you will get
242
by putting d= 4000 and K upon e will change depending upon 1, 2, 3 and these are the values
of for inventory stocking at different levels.
Now you see that we have already discussed at the time of model building that we want to
have a particular type of relationship between Q1, Q2 and Q3. And we wanted that Q2 should
be explained in terms of Q3. Q1 can be explained in terms of Q2, and here you see and already
we know when we discussed about only 2 installation problem that n 1 and n2 we want integer
values. Now simply just by seeing here Q 2 is 980, Q3 is 400. So 980 is not a integer multiplier
of 400.
Similarly Q1 is 2000 and Q2 is 980. Q1 is not an integer multiplier of 980. So this condition is
not fulfilled here, the solution which we got is not fulfilling the required condition and this
condition is only for the two installation case. In case of a multi Echelon system we already
discussed in our model building that we cannot have just n1 and n2, we want to have a further
relaxation of this problem.
And that for the relation we did by introducing this term 2 m , n1 is replaced by 2m , n2 is
1 1
replaced by 2m . So that you have only even multipliers of inventories for the previous stages.
2
You cannot have any multiplier; if you remember in one of the session when we were
discussing about 2 installation problems at that time we had given .
So that was possible in a 2 installation problem, but here we have already discussed during
the period of model building that you cannot have any values of n 1, n2. So we want only
integer values for n1 and n2 and for that purpose so that the very purpose of integer value was
that you can have same date of starting and finishing of replenishment cycles at the
successive stages.
So that was the reason and for that purpose we took a further relaxation of 2 m and 2m . Now
1 2
this data is at all not suiting this type of relaxation so we want to go to phase 2 of this
problem and for phase 2 of this problem you can understand that right now I am explaining
Q1 in terms of Q2 and Q2 in terms of Q3.
243
So for explaining Q1 in terms of Q3 , I will So . So now here I
have explain Q1 in terms of Q3. And the reason of this explanation is simple that I want to
explain the quantities required at each of these installations just in terms of my last
installation inventory Q3. So Q2 can be directly explain with the help of Q 3 that is multiplier
of n2 and Q1 can also be explained in terms of Q3.
And if there are many more stages you can see in the same way we can explain all quantities,
all Qi's in terms of Qn. If I consider this as Q5 and these are different values of n and then it
becomes Qn. So I have this type of general phenomena that to explain to explain the inventory
required at a particular stage I will have i this 2 can be replaced by i+1 and then up to n of N-
1.
So with this idea I can have and this product I can represent as p i which is the multiplier pi is
nothing but the multiplier of various values of multiplication factor to get the inventory of
successive stages. This becomes the inventory at any stage we want to keep
Now there is a small question that what will be the value of p n, the value of pn will be 1, pn
will always be equal to one. So please remember because in that case we are talking of the
last stage and last stage the inventory itself Q n is equals to Qn. So at that stage pn will be
simply one. So now in context of this we will revise this problem. So that you can apply this
formula in this particular case.
244
So now we move to phase 2 of this problem and phase 2 of this problem will require a bit of
calculation which we will derive from the Phase 1 only and in the meantime I request to all of
you to please consider the calculation of total cost of inventory what we are getting in Phase
1. In Phase 1 we are getting some values of Q1, Q2 , Q3 and on the basis of these values ,you
will have cost of inventory associated with that.
So this is the column of queues and you take this column of cost of inventory. So you
calculate cost of inventory with respect to these calculated values and then when you
summarise the values which you will be getting these are 1000 units, these are 49 rupees and
these are 2800. So the total value which will come of the cost in the first phase is about 3849.
This is the total cost which will come to you, now let us see the idea is can we further reduce,
so this we already discussed is known as C lower bound. We cannot have total cost was then
this. This is the highest total cost of inventory our supply chain will incur. So idea is now to
improve this cost means minimize the cost whenever in terms of cost improvement we say so
the meaning of cost improvement is to minimise the cost.
So by doing some kind of mathematics in these quantities of Q 1, Q2 , Q3 can be minimise the

cost, so we will like to see how to go ahead for that. So now we start our phase 2 of the
problem solving. So now in phase 2 to initiate the problem we want to revise these values of
queues, so initiate the problem I will directly take the value of Q3 here.
And this will make my Q as 400, but as I just explained now I want this Q 2 to be some
multiplier of this Q3 term. I want Q3 to be Q2 to be either 400 or I want 800 or I want 1600 or
I want 3200. So I want a multiplier even multiplier of 400, not this 980, but it should be
somewhere close to this 980. The idea is somewhere close to 980 which is possible here, so
for that purpose what we are going to do we will see that what is the most suitable value.
What is the most suitable value of this Q 2 and accordingly we will get the value of our ‘m’ or
you can say ‘n’ which will come here n2 or m2 whatever you say and for that purpose we have
this Q2 equal to 980, this is 400 and this side also you put 400. Now use this symbols here
and here I will like to write 2m, and here 2m+1. Now 2m will be selected in such a manner that it
is just slightly less.
245
If I write 21, so it means 21 x400=800, if I write 20 this becomes one. Then it is 400 only, so
that is very less rather you have immediate less value as 800. So here I can take m=1 and so
it is 21 x400=800less than 980 and this side it becomes m+1, so it becomes 22 x400
So this is 800 which is less than 980 and this is 2 4 x400=1600. Now I have two options
whether to take 2m or 2m+1. I will apply the same concept which we use for deciding the
integer values, which we decided to take what is the integer value of n. So same concept I
will take the ratios of these quantities and the ratios will say that 1 ratio is 980 / 800.
The other ratio is 1600 /980, now when you calculate these 2 ratios you will see that 980 /
800. is less than 1600 /980, this is less than 16, the left hand side is less, it is somewhere
around 1.1 and this is about 1.7, so the left hand side is less than right hand side, so whenever
left hand side is less than right hand side the value of m here m = 1.
If it would have been otherwise if left hand side would have been more than right hand side
in that case m will be m+1.So, Q 2 =21 x400. Now we will apply this same issue, the same
process for deciding the value of Q3
And here in this case you will have multiplier of 800 and in the centre you will have 2000. So
if I take again the value of m = 1, 800 x 2 that becomes 1600, which if I increase m=2, So
4x800 becomes 3200 which is not less than 2000. So this m = 1 is ok, if I take m =0, so then
it becomes 1x800, that is 800.
So which is less and because next higher value is 1600 is available which is less than 2000.
So m = 1 is the right value to be used here, it is less than 2000 and here m = 2, 4x800 =3200.
So here 1600 is less than 2000 which is less than 3200 and again we will apply the same
procedure that 2000/1600 and on the other side 3200/2000. And you will see that in this case
also the left hand side ratio is smaller than right hand side.
You can do on your calculators and you will see that left hand side ratio is less than right
hand side ratios. So these are the new values of Q 1, Q2 , Q3. Q3=400, Q2=800, and Q1=1600 .
And these values follow that issue of Q 1=4 Q3. Q2=2 Q3 and the issue of consistency is very
well handled here.
246
Now whether we have any advantage with this type of issue that is known by the calculation
of cost with these values of Qs. With these values of Qs you are supposed to calculate fresh
cost, you need to calculate the new cost and the new cost are 1025,50 and 2800 and total cost
comes 3875, which is slightly higher because of the relaxation of the problem.
So you have the best cost 3849, but this total cost 3875 is higher, but this type of arrangement
400, 800, 1600 will this gives you a lot of logistics comfort. But now the question is that
whether this solution 1600, 800 and 400 is the final solution or we can have further
improvement in this and for that purpose what we need to do that we will do one more level
of iteration.
Now in phase 2 we will do one more iteration and in that iteration the idea is that we will like
to improve the values of Q1, Q2 , Q3. So that this 3875 can further reduce you had this lower
bound, the best cost, but now you are trying to use the model for improvement so that you
have the logistics comfort. But because of relaxation which you have used the cost has
increased.
Now we want to see can we reduce this cost, can we do something with this 3875 and the
idea is that here we have calculated the values of n 1 =2, n2 =2. Now we will in the second
iteration again calculate the values of Q1, Q2 , Q3. And when we are calculating Q1, Q2 , Q3 we
will also determine the values of n1 and n2.
247
There is no change in the value of n1 and n2 so, no change in final answer. But if there is a
change in the values of n1 and n2 we may need to go for a third iteration also. The rule is that
we want to have stabilization in the values of n1 and n2.
As long as the values do not stabilize there is a fluctuation in iterations to iteration between
one value of n1 and another value of n2 between one values of n2 to another value of n2 we
need to keep on doing the iteration. But if in this case the values remain same in that case it
will be the final answer. So again the process will start by using the value of Q 3. Now since
we already have the values of n1 and n2 here.
So we will use these values of n1 and n2 for determining the value of Q3 here. Now for that
purpose we need to go back to this original formula of determining the inventory value EOQ
and using that formula we will determine the Q 3 here and using that formula you will see that
we are going to give you directly the formula and on the basis of that formula you can apply
the calculation.
Now since you have the values of one initial values of n1 and n2 you can see that our total cost
of inventory ci represented as . Now this formula can be changed in terms of

QN and pi the concept which is discussed and this will become as total cost of inventory
248
Now when this is the total cost of inventory just by seeing you can observe that our new setup
cost is Ki/pi and our new Echelon holding cost is ei x pi. so using this our formula for
calculation of QN will be it is the same formula this, but since here the setup cost and Echelon
holding cost are simply K/e.
So this K and e will be replaced by here this is the setup cost K i/pi and this Echelon holding
cost that is ei x pi and since I am calculating for the last stage, so you need to take summation
of this i=1 to n.
And now let us see the particular application of this formula in our this problem, so you will
see that here I want to know the value of Q3.
Now using this you will calculate the value Q3 and for p1,p2 you require values of n1 and n2.
p2 is simply n2 and p1 is n1 x n2. So you use this formula and then when you calculate Q3 will
come using this formula we have just expanded this general formula for a specific case and
when we substitute all these values Q3 will come 425. So now apply the same funda and you
will see 800 coming in between from here and 2m ,425 this side 2m+1 ,425 on the right hand
side and then you take 21.
249
So this becomes 850 which is higher than 800 not possible, so you take simply 425 here.
Because then you take m=0, it is 1, 1 x 425 it is 425, 800 and on this side it becomes 1, so 1 x
x 425, 850. And then when you calculate the values will become n 2 equals to because it is
done when you compare the ratios so you will take 2 m+1 here and m+1 the value of n1 n2 will
be 2 and the Q2 will be 850.
And then use take simply before the ratio then you will see it will become 1700 and annual
will become 2 and then you calculate the total cost C in this case that will come 3867. Now
you see that from this stage to this is stage the values of n are not changing, here n 1 remains
2, it is also 2. N2 was 2 there, it is 2 here. So all these values remains same so there is no
need to go beyond this stage, this is your final answer 425, 850, 1700.
And the total cost is 3867 which is slightly reduced from 3875 because this is a better
solution and this is the final solution. In practice meaning of this is that you have Q 1=1700,
Q2=850, Q3=425 and Q4 =425. Because for the modelling purpose because of the constraint
we have combined stage 3 and stage 4, but actually these are two different stages, so this is
the final solution.
And this way we can use the concept of multi Echelon inventory management, the concept of
EOQ for optimising the inventory at each state of your supply chain. So with this example
now I think the concept of handling the multi Echelon inventory is clear to us. Thank you
very much.
250
Lecture-22
Network Design in the Supply chain
So welcome back, so far we have discussed two important things with respect to decision
making, one was related to forecasting and another was related to inventory management.
Now we move to third important issue of decision making in the supply chain and that is with
respect to network design. If you recall in the beginning of this course we have discussed
about one of the very important driver of supply chain that is facilities.
Now where to locate facility?, what will be the size of the facility?, what will be the role of
that facility?. All these are some of the important questions, which we like to answer in the
network design. Now as we are moving from a local market to globalise market this
particular issue has become even more important and we will see that what are the important
issues we need to keep in mind when we take decisions with respect to network design.
In this particular session and few more sessions after this we will also discuss some of the
mathematical techniques, some of the analysis processes which we will use for taking a
decision with respect to location of our facilities in the network.
251
Now these are some of the important points which we are going to discuss in this particular
session. Now as I mentioned that we are going to discuss about the facility, the facility role
that is first important thing that what will be the role of the facility? our facility can be a retail
facility, our facility can be a wholesale facility, our facility can be a manufacturing facilities.
So these are the different types of roles facilities play in a supply chain.
So whenever I am going to have a new facility so whether it is a manufacturing, whether it is

a retail, whether it is a whole sale or what type of facility it is, that rule I need to be very
clear. So I take decision with respect to role of the facility, that is one decision. You have
thought of whenever Indian Oil, whenever Bharat Petroleum, whenever Hindustan
Petroleum, they take 2, 3 types of facilities, one facility is the refinery.
That is one type of facility in the network, then other type of facility is depo where you have
seen if you travel on Railways or some Highways, so close to those Highways and Railways
you will find some big depose of these oil companies where you see large tanks and
petroleum products are stored there and then you have retail outlets through where you get
your scooters, cars, motorcycles fuel and these are the retail facility.
So in the network of in the complete system of a petroleum company you have refineries, you
have oil depose and you have retail outlets. Similarly you can think of any other company
there will be different types of facilities you can think of Army, you can think of public
distribution system. In Public Distribution system which is very popular in our country for
providing the food grain to below poverty line people right now.
You see there are local ration shops where you have seen queues of poor people those who
are looking for food grains, sugar or kerosene oil. Then you have depose local depose, local
depose are normally governed by district food authorities and then you have the central
depose which are part of FCI Food Corporation of India, so there are big bear houses of Food
Corporation of India then a smaller warehouses of district food corporations.
And then you have local ration shops where normally the customers go and take the ration so
these are the different types of facilities which are there so you need to take a decision about
the role of facilities in your network decisions, then the location where should I locate my
facility, that is another important decision area we take decisions with respect to network
252
design. When I want to locate a facility the role of a facility is the customer interaction, the
retail facilities.
So obviously a retail facility has to be located close to the customer, but even in that case you
have an example like Walmart, the retail facility, but it is not very close to customer it is not
in the crowded market of a city, because of the model of Walmart where they want to fulfill
the requirement of the customer at the minimum possible cost, so they locate at their facilities
on the outskirts of the cities or maybe some semi urban areas.
So that the cost of land, cost of facility is lower and therefore they can transfer that benefit to
their customers, but on the other side we have discussed in our earlier sessions about 7, 11
also where they locate their facility in one of the most crowded areas of their cities, so that
customer can go and have the convenience of buying things with minimum efforts, so
location is one important things then if you want to have a manufacturing location.
So it depends whether you want to have manufacturing location close to customer. For
example if I talk of PET bottles which are used for containing water etc. in your refrigerator
or dustbins made of plastic buckets, made of plastic. For all these type of facilities you will
have lot of manufacturing facilities which are closer to your customers, because the
transportation of these products is a very compressive activity.
On the other side if I talk of automobiles, cars, scooters, motorcycles you will have limited
number of facilities for manufacturing these products and these maybe centralised throughout
the country. So depending upon the type of product you can have the location of your facility,
so that is another important area in the case of network decisions. Then the third is capacity
allocation, what should be the size of that facility? that is also very important.
I cannot have arbitrarily any size of the facility, I take a very logical rational decisions when I
want to decide about the capacity the size of this facility, what should be my size of the retail
outlet?, what should be the size of my depo?, what should be the size of my plant? Etc. So all
these are the decisions we take with respect to capacity or with respect to size of the facility.
And then market and supply allocations, which facility will supply to which market? or
which supplier will supply products to which facility?.
253
So both these decisions are also very important. So that you may have heard that in a
particular city there are 2 retailers of a product, 3 retailers of a product and then they will eat
the market share of each other and there will be a kind of local competition between 2, 3
retailers of the same product. Now to some extend this is not a very good network design
from the supply chain point of you.
You need to give sufficient market share to your individual facility, you need to allocate a
particular size of market to a particular facility, so that it is lucrative enough for that facility
owner to serve that product. Otherwise it will be difficult for the facility to survive in the
market, so we will take these four types of decisions when we are talking of facility in case of
our network design.
Now what are the factors, now let us see what are the factors which are influencing the
decisions of our network design, the first is strategic factors. The strategic factors are very
important in selection of a particular size, location etc. You have examples of India now
because of the very stable and strategic location we have so many multinational companies
coming to this country and making their manufacturing business.
Now you go to some other countries because India is a very stable country because it offers a
very huge market, so it is quite understandable that companies are interested in coming to
this country, but there are certain countries because of the strategic advantage like the
development of lot of warehouses, development of business in countries like Turkey is a
254
result of strategic factor. Because turkeys location is very much strategic, it can take care of
Europe, it can take care of Asia.
So lot of companies are interested in making their warehouses, in using the facilities of
Turkey for transportation of their goods. So it is important that strategic factors affect some
of your decisions with respect to location of the facility. strategic factors primarily affect the
location issue it primarily affect the location issue. In our country also you will find there are
certain state borders, the development of areas around Delhi particularly if I talk of Noida,
Greater Noida, then Gurgaon, then Faridabad.
All these areas have some kind of strategic advantage, because you are not in Delhi, so that
type of issues resolved and then you are close to Delhi. So you take the strategic advantage of
some type of benefits which you can get because of proximity of national capital. Then
technological resources are also important in deciding the network decissions. Now
development of lot of engineering facilities in a particular country will obviously help you in
deciding the manufacturing location in that country.
Particularly if I talk of countries like Japan, countries like Germany, where because of
technological excellence most of the manufacturing activities were concentrated and most of
the automobile companies, most of the electronics companies were interested to take the
advantage of those technological factors and this again affects the kind of location. But at the
same time technological factors affect the capacity allocation to a facility.
Because some technologies cannot be adopted, cannot be applied on a particular scale, you
need either higher scale or lower scale and this constraint of technology can also become a
particular reason for particular capacity allocation to a facility. So if I want to operate on a
particular economy of scale so I need to have a particular size, so technological factors are
also important. Macroeconomic factors are on other important factors in the queue which are
affecting my decision for network design.
I need to have a stable economic market, economic condition of that area, that country
particularly if I talk. Now if economic activities are not stable then probably I will not like to
establish my facilities in that area. If you recall all in last year's time in 2016 November
255
government of India decided to withdraw the legal status of notes of 500 rupees and 1000
rupees and all of the sudden the Indian macroeconomic market becomes volatile.
And as a result of that lot of uncertainty was there in this macroeconomics situation of Indian
market and at that time it is difficult for a company to locate its facilities particularly
manufacturing facilities in the Indian area. So therefore we look for stable macroeconomic
conditions, predictable macroeconomic conditions for deciding the location of a facility.
Then political situation we want stable political situation, earlier we used to have favourable
political situation.
But now slowly and slowly we are moving from favourable political situation to stable
political situation. Because now more and more focus is there on industrialisation throughout
the globe. So there is no question that Political situation is not favourable, but unfortunately
because of so many reasons you can say in most of the countries political environment many
a times remain volatile.
And as a result of that the effect of this political uncertainty makeup on business and
therefore we want stability in the political environment. If stability is there it will help
business because business will have a predictable line of action. If a political uncertainty is
there you will not have a predictable line of action and therefore you will not like to establish
a facility, a bigger facility in that area.
You see India is very fortunate that we have lot of political uncertainty, political stability in
this country and this is one very important major factor in attracting the global investment to
this country and most of the Global Giants they look for establishing their facilities in Indian
market in Indian subcontinent because of political stability. On the other hand you see our
neighbours and different types of other markets where political certainity is not there.
So investors are not looking to establish their facilities in those markets. Then infrastructure,
infrastructure also plays very important role in deciding the facility and particularly facility at
the manufacturing level at the warehouse label, these two facilities are very closely related
with the availability of the infrastructure. If appropriate infrastructure is not available these
facilities are difficult to establish.
256
You can see the example of some of the states of India, Uttarakhand, Jharkhand,
Chhattisgarh. These are slightly new states and what is most recently we got Telangana in
this list. Now the states like Uttarakhand develop the infrastructure for Industrial
Development right from day 1 and as a result of that lot of new industries, lot of new
facilities by industries were developed in these areas because of the availability of good
infrastructure.
This includes good network of rail and road. These include good air connectivity, this
includes good power availability, this includes good water availability. So these are the things
which are required for surviving business and if you can provide a decent kind of
infrastructure, so business will come or the facilities can be located into those areas. But if
infrastructure is not there, if road is not there you require that products will come to you using
rail as a mode of transportation.
But if there is no rail network available in that area, so how will you locate a facility in that
area. So it is very important that whatever type of infrastructure requirements you want to
have for your business, these requirements must be met by the local people, local authorities
and then the business can be located in that area. Then competitiveness, competitiveness can
be related with two aspects. The one aspect of competitiveness is that what competitiveness
area offers to you?
The facility where you are locating, the location can offer some kind of competitiveness. So
when I say the area or the local offers competitiveness, it is very much similar to strategic
type of thing. So we locate certain facilities because of its offering to the competitiveness and
the second is the competition available in the area, what type of competition is available in a
particular location?, that is also a very important factor in deciding the location of the facility.
If and in competition also there can be a case of positive competition or negative competition,
like we all must have seen that in all cities of our country India we have a particular Jewellers
market, now in that Jewellers market you have all the shops of jewellery, adjacent to each
other, you have this market where everyone is competitor of each other, but still a new shop
which is opened in an isolated location not in that Jewellers market.
257
It is very difficult for that shop to survive, it is not that competition is not there and because
of that the shop must go the fact is just reverse of that, why because this market offers a
positive competition, because all the customers will come to this market only and therefore
there is a positive relation, there is a positive relation and with that positive relation the
growth of all the facility owners take place.
And as a result of that any a new comer, any new comer in that Jewellers market will get the
benefit of that positive competition, but you always cannot have, you know there are certain
clusters in our country and ministry of micro small and medium enterprises is working on
that cluster mechanism, that you have a market of locks, manufacturing of locks in a
particular city known as Aligarh in UP.
You have a particular sarees known as Banarasi sarees in Banaras, you have a particular
product which is famous for a particular location. So these things offer you positive
competition, so if you want to have the similar kind of product, if you want to have a factory
of making the auto components. So Faridabad which is close to NCR is one of the most
prominent place for that, because you find thousands of auto component manufacturers in
that place.
So the local ecosystem will help you in developing in using those various facilities which are
customised for auto component manufacturers. So is the example of Pune, so is the example
of Rudrapur in Uttarakhand that and so is the example of Ludhiana in Punjab. That these are
some of the places where we have developed the clusters of auto component manufacturers
and similarly you will find many places, you have in Himachal Pradesh.
The industrial area known as Baddi where you will find lot of pharmaceutical companies are
coming up and there is a ecosystem local ecosystem of Pharma products manufacturers. So
positive competition is also an essential element for developing your infrastructure,
developing your facility related decisions and then we already discussed under the
infrastructure head, logistics and facility cost.
Though in place like NCR, Gurgaon, Faridabad, Noida we have highly developed
infrastructure. So ideally if I want to start a new business I want to start a new facility for
manufacturing something, these are the places in my first choice, but because of excessive
258
logistics and facility cause. These areas the cost is very high, already they have achieved a
very high level of success and as a result of that it is very difficult for me to meet the
challenges of very high cost.
So that also becomes a very prominent issue in my decisions of facility location that what is
the cost of facility, what is the cost of logistics in that particular place. If I want to open a
retail counter in a place like Chandni Chowk in New Delhi or in Connaught place in New
Delhi the cost is so high that it is almost impossible to think of a bigger showroom in these
places.
So what I do I will shift to some low cost areas where the cost of logistics and facilities are
relatively much less than these places. So these are some of the factors and I will take
decision of the combined consideration of all these factor. I take all the factors into my
combined consideration and then I take a decision that where to locate the facility?, what will
be the size of the facility? and what will be the role of the facility.
Now moving further if you see this particular graph it is some kind of representation of cost
of location of a facility and the response time. Now on this y-axis we represent the cost, on
the x-axis is the response time. Now where you see this curve, here you have local fast foods
and then you are moving with these type of makes regional local work in process, central fast
good, central work in process and then you are coming to custom production with raw
material at suppliers.
259
Now here your response time is very very low, customers do not want to have very high
response time and here the when the response time is very low obviously the cost becomes
high and custom products with raw materials at suppliers. Now when we are moving to this
way that you can provide the custom products and customer is ready to wait for sometime
and with this raw material is procured from the suppliers on the basis of your specification
products are produced.
So here your response time is very high, you have a high response time means the time taken
to supply the product is high and since time taken to supply the product is high you can
expect the low cost, because customer is now ready to wait and when customer is not ready to
wait customer immediately wants the finished good so in that case the cost will be high
because you have to fulfill the response immediately.
And accordingly when you need to keep the finished goods at the local level the last point of
delivery from where the customer is picking the product, customer is ready to pay extra
premium on that, but when customer is ready to wait, it means you can take depending upon
this particular curve you can see that what type of customers are there for you? and on the
basis of that you can decide the location of your facilities.
If your customers or you are providing the finished goods to your customers immediately,
you need to be close to your customer and therefore you will be in a crowded market where
customers often visit and this type of curve will be there and if customer can wait, so there is
no need to locate a facility close to customer, you can be far away from the customer, you can
be far away from the costlier places because customer is giving order.
And then you are procuring raw material from the supplier and providing products with the
customer, so here because customer is ready to wait you are on the extreme side where
though this is how you see a straight line is coming local finish goods that makes then some
kind of regional finish good. So from a particular reason there is a type of local area from
where you can provide, you have one situation where locally you are providing the product.
260
And in some other case you have a region and from here you are providing the finish product
to the customers, so this is reason, this is local. So in this case when I am providing the
product from the regional cases my cost slightly decreases because I can take the advantage
of economy of scale but and in our further discussion also you will see that as my number of
facilities will increase I am moving actually from regional facilities to local facilities.
When I will provide with the local facility to my different customers for these 5 customers I
can conceive 3 local facilities like this and one here. So with these 3 local facilities I can
fulfill the requirement of my customer in a much better way, so as my number of facilities
will increase you see as I am increasing the number of facilities on this X axis, the response
time means my ability to fulfill the requirement of my customers also decrease.
I can fulfill my customer on a very fast bases, very fast basis so the response time also
decreases with the increase the number of facilities, but there will be a cost associated with
increased number for facility. So in our next session we will see the relation between the
number of facilities the cause related to those facilities and then how to achieve and optimum
number of facilities, so that you have a lower response time and at the same time you have
optimum level of total cost of facilities. Thank you very much.
261
Lecture-23
Network Design of Global Supply Chain
So welcome back in our last session we were discussing about network decisions we
discussed about 4 important type of decisions which we take. We take decision with respect
to role of the facilities, we take decision with respect to location of the facilities, we take
decisions with respect to their sizes and then we take decisions with respect to market and
supply allocation.
We discussed about various factors which are responsible for taking decisions for network
decisions and then in our last session we also discuss that how response time and number of
facilities are related.
We discuss that if we have more number of facilities the response time will decrease, as we
are coming from left to right on the x-axis we are increasing the number of facilities, so our
response time is also start decreasing.
262
So we already discussed this in our last session. No going further if you see this type of figure
that where you have this particular market which is plant, then here you have distribution
centres, these yellow points. These are representing the distribution center and then all these
dots are representing the customer. So you have these points representing the plant these
yellow point representing the distribution center.
And then these marks small dots are representing the customer, so this is a type of situation
where you have so many customers, so many distribution centres and few plants. Now here
you see we can say that we require one week order response time, in this particular case
because you have only 3 plants 1, 2, 3, 4 plants and you have very limited number of
suppliers with respect to distribution centre.
So here I talk like if I talk with this type of a distribution centre or this type of distribution
centre here your required at least one week response time, so in this particular case the
response time is very high because of limited availability of the distribution centres in this
overall global market.
263
Now let us go to this type of situation, here you have this distribution centre this distribution
centre. Now within this market you have 2 distribution center and here in this particular case
because 2 cycles are intersecting, 2 yellow cycles are intersecting. Now you see here in this
particular case for these people those customers which are coming in the intersecting cycle,
here we have two distribution centre.
This distribution center can also fulfil the requirement of customers coming in this
intersection cycle and this distribution center can also fulfill the requirement and as a result of
that the response time has reduced from 1 week to 5 days. So that is the result of 2
distribution centres for the customers which are in the intersecting cycles.
Now you have more such distribution centre and now you are reducing the cycles of
distribution center and therefore you will find that in some of the cases you have more than
264
one cycle which are able to fulfill the requirement, because you have 5 distribution centre for
this market and talking of only this market right now, so for this market this US market you
have 1, 2, 3, 4, 5 distribution center and there are at many places customers can be served by
more than one distribution center. And what is happening that we have reduced the response
time from 5 days to 3 days.
So as we have said that as our facilities if you go back to this slide so as our number of
facilities are increasing we have one facility so at that time 7 days was the response time we
had 2 facilities, 2 distribution centre, so response time decreases to 5 days. Now we have 5
distribution Center to response time decreases to three days. So that is how we are moving for
more number of distribution centre, less response time means we are able to fulfill the
requirement of the customer in a faster manner more with less time.
265
And now you see you have how many 13 distribution centre, you can see the 13 yellow
circles here and these 13 yellow circles will further reduce your response time and here we
see that the response time is the next day, you order today next day you get the supplies. So
because in many cases here you will see large number of intersecting circles and as a result of
that customers can get benefited by reduced response time. So more circles, more distribution
centres and your response time for that decreases.
So again strength the same thing that more facilities and response time will decrease. Now
going further you have now how many you can count these yellow circles. You can count
these yellow circles and the result is already displayed here that is these are 26 distribution
centre. These centres and your order can be received same day, the same day order or in some
cases like in this case where you do not have any intersecting circles.
266
So probably in these cases if I talk of this customer, so probably because of the distance from
this distribution Center to this customer order may be supplied next day in this case those 3
customers are very close but there are two customers which are on the periphery of the circle
for them order may be supplied next day. There are few customers which are not part of any
of the circle for these customers order may come after one day only.
So as we are improving our number of facilities we can see in this diagram that the response
time has further reduced you are able to get the product in the same day, so from one
distribution center when we were having only one single yellow circle at that time our
response time was 7 days, one week and now we have 26 distribution centres and the
response type is either the same day or next day. So as we are increasing the number of
facilities it is very simple to understand that the response time will keep on decreasing.
Now you see in this particular case what are the different types of cost which are associated
with facilities. Now as we have this cost verses number of facility graph, so your with facility
as you are increasing the number of facilities, your inventory cost increases.
267
Because now if you see you have various distribution centres 1, 2, 3, 4, one manufacturer, all
these are the DCs, then you have large number of retailers maybe 1 to 50 and as you are
increasing these numbers from 4 to 6, from 50 to 60 you need it at each location at each
facilities some amount of inventory, some amount of products to be kept at each of these
locations. So when you are having more facilities in your network your inventory cost is
bound to increase. You have higher inventory cost, so that is one simple lesson that more
facilities more inventory cost.
The second is transportation, now in the cases of transportation you see the cost starts
decreasing as we are increasing the number of facilities the cost starts decreasing to some
point here and then all of a sudden it starts increasing, why it happens, why it happens that
way that cost is decreasing to a point and then starts increasing.
268
The reason is that transportation cost can be divided into two parts if this is a facility in
supply chain. So there are some materials which are coming to the facility and there are some
materials which are leaving the facility in a supply chain. So materials which are coming to a
facility the cost of transportation associated with the inbound material is known as inbound
transportation cost.
This is inbound transportation cost and the cost associated with the materials leaving the
facility that is known as outbound transportation cost, it is very simple to understand that
normally the size of consignment which we receive at a particular facility is larger, is bigger
than the size of consignment leaving the facility. The size of consignment leaving the facility
is normally is smaller than the size of consignment coming to facility.
We know that if I am a manufacturer I receive raw material in large quantities and which is
very homogeneous also and the finish product because these finished products will go to
many dealers, many distribution centres, so the size of outbound transportation, outbound
consignment will be smaller. The point which I am trying to say here you get inbound
logistics, here you get sufficient economy of scale, but here that economy of scale will not be
there.
Though we try to keep some amount of economy of scale even in the outbound transportation
also. When we are making more and more facilities we are trying to go closer to the
customer, so what will happen we may lose after a particular level of facilities after a
269
particular number of facilities we may lose all of a sudden economy of scale even in inbound
transportation cost because more facilities so the size of inbound consignment will also
reduce with the assumption that market is not growing.
If market is constant and I am making more and more facilities so in that case it is quite
possible that with increasing number of facilities my economy of scale my benefit of
economics will not be there even in the inbound transportation and if that happens because of
more number of facilities, so transportation cost starts increasing after a particular number of
facility.
Otherwise as long as I am coming to this site transportation cost will decrease because I will
be going closer to my customer and I will continuously decreased my outbound
transportation cost, I am going closer to the customer, so this outbound transportation cost
which is much higher than the inbound transportation cost. This cost is higher than this
inbound transportation cost. So making more number of facilities will help me to reduce the
outbound transportation cost.
As long as I am able to maintain economy of scale in the inbound transportation cost but
what I said that after a particular number of facilities I may lose economy of scale in the
inbound transportation cost also and at that time the benefits of reduced outbound
transportation cost will be nullified or rather I need to pay more if that happens if I lose the
economy of scale in my inbound transportation cost.
And therefore I can have this type of increase very sharp increase in the transportation cost of
the product after a particular number of facility, so that is the discussion about the second
important cost with respect to number of facilities.
270
The third cost of the facility cost itself, this is the facility cost, so as I am making more
number of facilities I need to develop the required infrastructure, I need to put my capital into
that and as a result of that with more number of facilities I have increasing the cost of facility.
Up to some level you see I am moving in a very linear fashion up to here, but beyond this
point there will be a very sharp increase exponential increase in the facility cost.
I will ask my students to think on this aspect that why the cost of facility increases in this
exponential manner, after a particular level. After a particular number of facilities. And
probably in our next session I will like to around reply I will give you this question right now
that why you have this straight line up to a particular point of number of facilities and beyond
this the shape of the curve changes at it becomes more exponential in nature with respect to
facility cost. And you please think through and in our next class we will like to discuss the
answer of this every question.
271
Then if I combined my various types of cost, so this type of curve emerges where this curve
at the top represents the total cost which is the sum of facilities, inventory, transportation and
labour cost, more number of facilities so with increase in facilities you have a step increase in
the labour cost also, you have transportation cost decreasing and then increasing, you have
inventory cost which increases with number of facilities and the cost of facilities itself.
And as a result of that you have this total cost curve where total cost is decreasing up to a
point and then it starts increasing because of increase in the transportation cost and then the
role of services and facilities is also very important with your increased number of facilities
why are you increasing the number of the facilities?. The question is why are you increasing
the number of facilities?, we increase facilities because we want to serve our customer in a
more satisfying manner.
Or you can say that we want to achieve higher level of services by providing more number of
facilities, more number of facilities going closer to the customer and going closer to the
customer will bring better service level so as you are seeing with this curve that your service
level is increasing with number of facilities as you are increasing the number of facilities
your service level is also improving. So this is a combined discussion of variety of cost and
the service level which you get with the increase number of facilities.
Now with this discussion we cannot understand that what should be the ideal number of
facilities. Now if you see the total cost curve, the lowest point of this total cost curve
somewhere here around this point and here if I get this level of service and if this level of
272
service is within the my promise limits to my customer then I will like to keep this as the
number of facilities, these number of facilities I will like to keep and you see this is the point
of intersection of my facility cost and the transportation cost also.
So this is the point where I get the minimum cost of facilities and this is the level of service.
Now if I want to improve my service level this is only possible when I will incur additional
cost of facilities. Additional cost of facilities will be required for improved service level for
higher response rates and in case of a product where we want very high level of
responsiveness we can go with higher level of service and accordingly cost of offering that
service will also increase.
So those products probably of emergency nature where we require very high level of service
in that case cost many a times may not matter and therefore you will provide high level of
services by going to the higher level of caused by increasing the number of facilities like in
the case of humanitarian supply chains, the disaster supply chains where you are providing
the facilities for better rehabilitation for immediate relief of the victims.
In those cases cost issues are secondary, the ability to provide relief immediately is more
important and in such type of situations we will like to have more number of facilities in
those disaster prone areas in those risky areas where you feel that disasters may come and for
that purpose to provide higher level of services you keep your ambulances, your nursing
homes, your other dispensaries, your other facilities, always readily available.
And therefore your cost is also high but in those situations we require very high level of
response and therefore it is well justified but in many other cases the higher level of response
of service may not be required and therefore it is not essential to increase the cost of your
offering by increasing the number of facilities. So therefore in most of the cases it is always
advisable to understand this graph very carefully for the practitioners those were attending
this class.
It is always advisable that please understands this break up and tries to optimise the number
of facilities. Otherwise you will increase the number of facility but you will not have
proportionate increase in service level, because you see the service level curve becomes
273
almost is stagnant after a particular type and therefore understanding of this graph is very
very important.
Now moving further let us quickly see once we are deciding about the location of the facility,
so how do we select the site in a global environment, what are the different phases involved,
you see this is a 4 phase process, phase 1, 2, 3, 4. The process starts with the supply chain
strategy, this is the first stage of our site allocation, site selection that what is my supply chain
strategy.
Because all the facility location decisions must be in line with your supply chain strategy, so
so understanding the strategy is very very important, I discussed in my last class about
Walmart and 7-11. The Walmart strategy is to provide low cost product, the 7-11 strategy is
to provide the very high level of responsiveness. If I am an emergency type of hospital, so my
condition is different and if I am a hospital for routine medicines, so in that case my condition
is slightly different.
So this depends upon me whether I am a routine type of hospital or I am there to handle the
emergency cases. So as per that my first important thing is supply chain strategy. Now supply
chain strategy is we have already discussed that for every quick brush up will like to discuss
that it comes from the competitive strategy of the organisation, what is the competitive
strategy of the organisation that derives your supply chain strategy.
274
And then the other factors which contribute in proper understanding of the supply chain
strategy or proper you can say written framework of the supply chain strategy one is the
internal constraints and the second is the Global competition. These are the two more factors
which are very important in deciding the supply chain strategy. So totally there are three
inputs. Three inputs for my supply chain strategy one is the competitive strategy of the
organisation.
Second is the global competition what is going in the global level with respect to your
industry and the third is the internal constraints related to Capital, the growth strategy, the
existing network etc. These things also play important role because if I am let say Tata
Motors and I am launching Tata Nano a new vehicle in my portfolio, so I cannot have
entirely new network of distribution for Tata Nano.
Because I already have a distribution network for my other products and I will like to use I
will like to have some kind of strategy with my existing network. On the other hand we have
the example of recent example of Maruti Nexa, where Maruti to promote its exclusive type of
product is not using the existing Network and Maruti has created a separate network of
distributorship named as Nexa.
So it depends upon what type of growth strategy what type of existing network in case of
Maruti though we all know Maruti is known for its largest network among the automobiles,
but still for its exclusive products if they want to have some kind of different positioning of
some of their products they gone for the Maruti Nexa. So it depends that what type of growth
strategy?, what type of capital requirement you have and then on the basis of that you decide
supply chain strategy.
The second phase is the regional facility configuration. Now on the basis of that you have
certain regional consideration and regional considerations come from the tariffs and tax
incentives like when new states of Uttarakhand, Chhattisgarh and Jharkhand came into
existence in India. Uttarakhand government, Chhattisgarh, Jharkhand, so on gave of tax
incentives to new organisations.
And as a result of that many new companies, many new MNCs they open their facilities in
these new states. So some type of tax incentives, what type of tax incentives are available in a
275
regional area?, what is the regional demand?, whether we have any regional demand or not,
most of these states were not having any kind of regional demand. But because of very heavy
tax incentives many facilities came in this areas.
Then certain political stability type of issues are also very important and because industry can
only survive facilities can only service where you have political stability. Then another issues
related to production technologies whether we have appropriate technologies available. This
issue is more important when I look at the global level. So because in that case the regional
issue is a particular country.
If I talk in case of India a regional issues related to a particular state, so in that case what type
of technology support will be available in country like India or any other country where I am
thinking to establish my facility and what is the competitive environment do we have a
positive competitive environment in that regional area or not, so that is also very important
and with these things I draw some conclusion about my regional configuration.
I divide the global market, global arena in certain regions and then I do this type of analysis
for all those regions and on the basis of that I may consider one or more than one reasons
which are favourable for my decision making and once I have done that third issues related to
site, within that region what are the possible sites where I can locate my facility like if I take
regional issue related to Uttarakhand , if Uttarakhand is one region.
So desirable sites can be Udham Singh Nagar, desirable sites can be Sitarganj, desirable site
can be Haridwar and out of that I will look that what is the available infrastructure in each of
these sites. Now I will go to a very specific issue related to power availability related to
manpower availability related to railroad infrastructure related to water availability etc. and
the production methods is still needs.
And response times means what type of skills are available in that particular site and what
will be the response time how I will manage the response time from this particular site. So
depending upon my market and supply allocation I will take the decision with respect to
response time and on the basis of this finally I will like to decide I will come to a particular
location where I will see the issues related to logistics costs, issues related to factor cost of
labour material and some site specific issues.
276
So when these four stages are passed I can take decisions with respect to location of a
particular facility in a particular region at a particular site. We will see, we are stopping our
this session at this end but in our next session we will see that some of the mathematical
model, some of the mathematical analysis which can help us in selecting the regional facility
and desirable sites on the basis of various inputs which we are required and some
mathematical models are there.
And we will request you that if you can practice some Excel solver kind of thing, if you can
install Excel solver in your computer and we will do some questions here and with the help of
those questions you can also practice at your end and then you can see that how in real-time
the modelling related to networks can be done, because the decisions related to stage one of
the supply chain where we are talking of manufacturing related decisions.
These are permanent type of decisions, but decisions related to next stages retailers,
distribution centres are more dynamic into nature and where we require more real time
analysis. So thank you very much we stop at this time and will look for your participation
with Excel solver in our next session. Thank you very much.
277
Lecture-24
Alternative Channels of Distribution
So welcome back in our last session we started discussions about network design decisions
and we discuss the role of facilities in our supply chain decisions and facilities we have
already discussed are one of the important drivers of supply chain cost with the help of
facilities you can achieve responsiveness as well as efficiency. If I start making large number
of facilities in my supply chain configuration, so I will achieve high level of responsiveness.
If I consolidate facilities on the other side I make fewer facilities maybe with respect to retail,
maybe with respect to distribution centre or maybe with respect to manufacturing facilities.
All these consolidation will give me economics of scale. And in our last two sessions we
have discussed that how with respect to facilities different types of cost maybe cost of
inventory, maybe cost of transportation and the cost of facilities itself will change.
And therefore we need to decide one optimum number of facilities in our supply chain
configuration. This optimum number of facilities will give us the minimum possible cost of
facilities in a supply chain configuration, but sometime we want higher responsiveness that
optimum level of services optimum level of facilities will give you a particular level of
response in the supply chain.
But there are certain cases where we require even higher responsiveness and when we want
that high level of responsiveness we need to go beyond that optimum number of services,
optimum number of facilities and therefore depending upon yesterday we discuss that your
supply chain strategy is the first important area from where we will start discussions about
the configuration of your supply chain.
And supply chain strategy where we have already discussed in the earlier session that it has
one broad spectrum. On the one side of the spectrum you have highly efficient supply chains
and on the other side of that spectrum you have highly responsive supply chain. So that
278
spectrum you have various intermediate locations and these intermediate locations will have
variety of combinations of responsiveness and efficiency.
So the number of facilities will decide what type of combination you want to achieve of
efficiency and responsiveness. Normally for a General type of decision, for without going
much into the strategic aspect, we want somewhat responsive, somewhat efficient type of
supply chain, we want to keep our self at the centre of that supply chain strategy spectrum
from be responsive site to the efficiency side we want to keep our self at the centre.
Now if we want to keep ourself at the centre where we get a particular level of
responsiveness with minimum number of facilities, so if I see in that slide the conventional
supply chain network are like the above slide where you have various raw material suppliers
these are the vendor distribution center and this raw material suppliers suppliers of different
types of components which we required in making our products.
Some of the vendor distributions are supplying simply the raw material, some of them are
supplying the semi finished or half finished type of products or some of the various
components, so different vendor and different center will supply the different types of
products which are required in making these product. You take an example of a car or
automobile product.
In that case I get products like hot rolled Steel, then cold rolled steel, or making the body of
my car or bike, then I get tires from some of the tire manufacturer, I get various components
279
of the engines, various components of the air conditioner, various components of my other
electrical systems of my vehicle from other vendor. So different vendors will provide me
different types of product.
Now these products will come to this manufacturing area where you have this manufacturing
facility from where you can see the flow of manufacturing this is known as work in process,
so from material distribution Center to component manufacturing component manufacturing
to custom component distribution Center and final assembly. So this is the system we have at
the manufacturers' level. So these are the different types of vendors and then you have the
assembly system of those components, those semi finished product at the manufacturing
level.
When final assembly or final product is ready then we finished product or stored in a
warehouse within the plant. Normally those who have some kind of exposure of
manufacturing activities may be aware that in each of the manufacturing facilities you have a
local warehouse, local warehouse within the plant remises. So this final assembly product
weekly product go to this plant warehouse.
Now from plant warehouse you send to the your finished goods distribution centre, In a
conventional systems if my plant is located in let say Gurgaon area. So I will have some
finished goods distribution center, one may be near Mumbai which capture the demand of
Gujarat, Maharashtra, Madhya Pradesh, those area and may be one near Chennai or
Bangalore to take care of the demand of 7 area of the country.
And from my own warehouse because I am located in Gurgaon, so I will take care from that
the entire northen region of the country. So this is the conventional system that my finished
goods will go to my distribution center. The company owned distribution center. Now from
my finished goods distribution center these products will go the different customer
distribution centers.
Now I will give normally my company, my agency to some customer site and from these
points my customer finished goods will go to the distribution centres of the customer end. So
these are the customer distribution centers. So depending upon like if my finished goods
280
distribution centre is in Chennai, so I may have customer distribution centers in Bangalore, in
Hyderabad, in Trivandrum and those places to take care of entire southern part of country.
If one of my finish good centers is in Mumbai, the customer distribution center can be in
Ahmadabad another can be in Nagpur, another can be in Pune. So that I can take care of the
western part of the country and this is how you come to the customer distribution center and
then finally you come to the retail outlets, customers retail outlet from where finally customer
will purchase the individual customer will purchase the product.
So some bigger customer stores a direct procures material from the company’s finished goods
distribution center and the smaller customers will purchase from the customer distribution
center. So both these type of network are possible from this point to this point. You can use
direct distribution or you can use over level of distribution, we have already discussed in our
earlier sessions one of the system of distribution where you have different types of
distribution systems.
Where depending upon how many intermediaries you use to you reach the final customer that
depend up on variety of factors and some of them we can discuss here again like this is a
manufacturer, so you can put from here to here this entire set up as manufacturer area, so this
is manufacturer and this is customer, so from manufacturer to customer we have various
intermediaries. In this case the first case manufacturer is directly communicating with the
customer.
281
There is no intermediary between manufacturer and customer, so this is known as direct
distribution. No middle values involve, no intermediary is there. In the second case you have
one retailer between manufacturer and customer. Now manufacturer will supply products to
retailer and retailer will finally give to the customer. This one intermediary is involved, this is
known as one level of distribution.
So in this case if you see if finished goods distribution centre, if supplies to the customer
store with the help of this customer distribution centre, this is an example of one level
distribution, but in our case when fitness distribution centre is directly supplying to customer
store, this is the direct distribution. So both these type of distributions we have shown in this
conventional network.
Then you have other type of distribution network also, where you have 2 intermediaries, 1
wholesaler and another retailer. You have 2 intermediaries here, from manufacturer to
customer, so this becomes 2 level distributions and in some cases you have a bigger
wholesaler, a bigger retailer one and smaller retailer 2. That is also possible this is 3 level
distribution.
Because you have three intermediaries involved here from manufacturer to customer. Now in
0 level direct distribution, one level, 2 level, 3 level, what is the relevance of this for our
network design decisions, it is very important to understand that in which situation,which
type of distribution network is to be used. Now we all can understand that direct distribution
where no intermediary is involved between manufacturers to customer.
So can we recall can we think of any product which we purchase directly from the
manufacturer, probably it is very difficult for us to think of such products which as a
individual customer we are purchasing directly from the manufacturer, but think of some of
the products which I can name like Uttarakhand Jal Vidyut Nigam Limited, Tere hydro,
NTPC and NHPC all these are the power companies.
They require turbines to develop a power project. Now they purchase turbines directly from
BHEL, so BHEL is manufacturer and NTPC, NHPC, THDC all these are the customers. So
there is no intermediary all these customers directly purchase product from the manufacturer.
And in this particular case it is direct distribution.
282
Now the very simple reason that there is no retailer involved in this types of project because
the cost of the product is so high that it is almost impossible to have any intermediary here,
because these intermediaries, this retailer, wholesaler, retailer to, all these intermediaries they
will also take some of the part of the profit. They are only working because they are
interested to take some kind of profit share.
Now when the unit value of product is very high in this case the cost of the product will
further increase because of the involvement of the intermediary here and that extra burden of
cost will come to the pocket of the end customer, so it is not advisable or it is not possible to
incorporate any kind of middle man in these types of very unit products.
High unit value products are normally directly distributed, as an individual customer we do
not purchase those high unit value products. For us the high unit value product maybe a car,
maybe a house, these are the high unit value products which as an individual we purchase and
probably in a car when you purchase a car you have this one level of distribution Maruti than
the distributor of Maruti and you are the customer.
So it is a relatively high unit for us, but not so high unit value product that you directly go to
Maruti and purchase it. So therefore 1 level of distribution is required, but in case of house
when you purchase a property you will find use of both these type of distribution system,
sometime developer himself will contact the customer and you will go and purchase the
product, but sometime you will see developer will also have some kind of agents.
And now days lot of agents are available through e-portal and that is one level of distribution,
that developer is directly distributing the sale product and developer is using a retailer also
for distributing the product. Because house is a sufficiently high unit value product. In high
unit value products we can use direct distribution as well as well develop distribution and
now you come to these type of distribution systems when you have a wholesaler and retailer
and in some cases you have more than 2 distribution intermediaries.
Now these type of distribution systems are normally seen when the unit value of product is
very less and you have the examples of fast moving consumer goods FMCG where you find
this type of distribution network is very prevalent, you take the products like cosmetics, you
283
take the food items, you take the other type of FMCG. In that you find that wholesaler is
there, a big retailer maybe there and a retailer in the rural area purchasing products from the
retailer available in the urban area and supply products to the rural consumer.
So you find many intermediaries when the unit value of the product decrease. Because
burden because the cost which customer is going to pay is not going to increase substantially.
So as the unit value decreases you can offer more number of intermediaries to reach to large
number of customers in your market. Another important thing which is there in selection of a
particular type of distribution network that is the nature of the product.
Nature with respect to perishability, but the perishability we all know that how long a product
can survive, how long you can derive the used values from the product. Now each product
there are few product where issue of perishability does not come but there are large number
of products where issue of perishability is also a very important phenomena in designing
distribution network for them.
And as a supply chain manager we should be aware that what are those perishability issues,
which are important in designing with distribution network. Now for an example whenever
we talk of perishability the first few examples which come to our mind these are related to
food products, the dairy products, the milk products, but these are products where the degree
of perishability is relatively higher.
And if you talk of chairs, you talk of tables, you talk of computers, you talk of camera, here
the degree of perishability is relative very low. Now when the degree of perishability is
relatively low you can without any difficulty have a long distribution network. Because the
value of the product is not going to decrease, but when you have a product like curd, you
have a product like ice cream, you have product like vegetables.
In that case as the time will pass the value of the product will decrease, so you want a smaller
distribution network in case of these products where the degree of perishability is high and
therefore you can see that in case of food products, milk products, sweets, all these products
where degree of perishability is relatively very high, you can store some products may be for
2 days, for 3 days, for 4 days.
284
Some products you need to consume within 24 hours. So in all those types of products where
perishability is a very important issue you need to have either direct distribution or 1 level of
distribution. You can see in your area of wherever you live the milkman is the manufacturer,
responsible for getting milk from cows and buffaloes and he is directly distributing products
to the customer, maybe within few hours of taking milk from those cows and buffaloes.
So because degree of perishability is very very high, if you do not treat that milk within few
hours then milk will lose its value and therefore it has to be distributed directly in most of the
cases. Though we have examples of Amul, Parag, Mother Dairy etc. where we have long
distribution network but that long distribution network, please remember is only possible as
you have increased the shelf life of the product.
And now you have when you are increasing the shelf life of the product you are decreasing
the degree of perishability and therefore the product can survive for long durations and you
can therefore choose a longer distribution network. Now coming back to perishability issue
there are these are the normal examples vegetables, fruits, milk products, sweets etc. where
degree of perishability is relatively very high.
But there are products where you can think of the degree of perishability is even higher and if
you put slight stress on you, you can realise that degree of perishability is extremely high in
case of services. Services are those products which are intangible, services normally we
discuss these things in a class of marketing management but because we are discussing this
issue of distribution network.
So services are those products with degree perishability is extremely high and these are
intangible, you cannot store them, you need to consume services as soon as your producing
them and therefore in case of services you can only have the direct distribution, you cannot
have any other level of distribution. This is only suitable for services type of products. You
are delivering a lecture to the class when you have not this type of arrangement where you
are getting my lectures on online board.
But if I teach class with students are having one to one contact, in that case if you are slightly
out of the class your mind is out of the class that portion of the lecture is gone away. A doctor
can only write a prescription when a patient is there, without a patient doctor cannot write a
285
prescription, you can get a haircut only when a customer is there, without a customer haircut
is not possible and therefore all the services products can only be possible when a customer is
there.
And therefore because you cannot do inventory of services, and when you cannot do
inventory of services you can only directly distribute them, customer has to service provider
to avail service, in case of because the discussion of supply chain management is equally
applicable to the services organisation.
Services organisation like a bank, like a hotel, like restaurant, like a gym. All these are like a
hospital, all these are the examples of services organisation. So it is very important to discuss
the supply chain analytics for the services organisation also and particularly in the case of
India it is even more important because you all may be aware 60 to 65% of contribution of
GDP is coming from our service sector, the contribution of manufacturing is stagnant is
around 16 to 17%.
So the major share of economy is being controlled by the service centre, so we need to see
that how services supply chain network is developed and therefore it is very very important
that you should understand that for case of services organisation you need to develop the
direct distribution. Now the only issue in case of services distribution network you need to
see a particular case that whether you are going to customer to provide a service or customer
is coming to you to avail service.
In one case a service provider is going to a customer for giving the services. In another case a
customer is coming to service provider to avail the services like we go to hospital to take the
services of the doctors.
But in some cases it is also possible that doctor may visit a patient, doctor may go to a call.
So in that case service provider is coming to customer and now days because of various
issues you will mostly see that customers are going to the service provider, but in some
exclusive services you will see that service provider is coming to customers, you have a
system of private tuitions.
286
In that case service provider is coming to the customer but when you join classes of FIdji,
when you join classes of Vidya, when you join Akash. These are the examples when
customer is going to service provider, so you need to design what type of arrangement you
want to have and that will again come from your supply chain strategy that what type of
strategy you want to have.
And accordingly you will see what type of distribution system of services will be are there,
whether manufacturer is going to services going to customer or customer is coming to
manufacturer. And in rest of the system because you have a system of store, you have a
system of inventory and therefore you can use these intermediaries. These intermediaries can
only be used when you have inventory system, otherwise you cannot.
Now with respect to the perishability there is one more type of product that is very common
all of us use that, that is newspaper. Now newspaper is another example very typical example
where you will find a very long distribution network, degree of perishability is even higher
with respect to newspaper and you talk of food products. So degree of perishability of
newspaper is more than degree perishability of food.
But in case of newspaper you have a very long distribution network. But the distribution
network is designed in such a manner that ownership transfer from one state to other from
this stage to this in a very fast and therefore within few hours the newspaper is printed here at
about 12 midnight and at the morning 6 o'clock or 7 o'clock news paper is in the hands on the
customer and you have various intermediaries, you have the press.
Then you have the logistics service provider, the vehicles then you have the local agency,
news Agency where the newspapers come from the manufacturer the press and then you have
the local hawkers taking papers from that agency and distributed to the customers house. So
this is a long distribution network but because you are able to deliver it on a very fast basis,
you can deliver it in a quick time.
So these are the two important issues, one is perishability and another is the unit value of the
product on the basis of which we take decision whether to use one level, 2 level, 3 level or
direct distribution. But nowadays we do not use a single type of distribution network rather
287
we use more than one type of distribution network simultaneously. As we have discussed in
this particular case that we are using one level and direct distribution.
And so on you will see that now days even all these four type of distribution networks are
used simultaneously and that is known as dual distribution. So most of the companies are
going for this type of dual distribution where you are directly distributing to the customer
through e-portal and then you have your exclusive showrooms and then you can go for this
wholesaler retailer type of system also.
So because we want to reach to large number of customers, we want to achieve very high
level of responsiveness, so we want to use all types of distribution network parallelly. So we
stop here in this class and in our next class we will discuss some of the models to optimise
the location of these intermediaries in our distribution network. Thank you very much.
288
Lecture-25
Location Decisions in Supply chain
So welcome and we are already discussing about decisions for the development of a network
in our supply chain environment and we discussed in our last session a very conventional
type of distribution network you have distribution centres of the vendor, then you have a
complete setup of manufacturing and you have a warehouse within the plant, then you have
distribution centre of the finished goods, owned by the organisation itself.
Then you have customer distribution centre and the customer stores. And there was a very
serial network between all these facilities was there. Nowadays if you see the changes which
are taking place over a period of time in the supply chain systems particularly with respect to
development of facilities and their rules.
The current slide gives you an idea that in our latest systems we have in a global kind of
supply chain environment you have one finished goods distribution centre at the national
level, so this is a central facility within a particular country, then you have various regional
finished goods distribution centre, so maybe if you take an example of India so within Central
India somewhere maybe around Nagpur, Bhopal in that area you have a national finished
goods distribution Center.
289
And then if you divide India in four major regions North, East, West and South for that you
have four Regional goods distribution centre, finished goods and then you have local
distribution centre in maybe State wise you can think of that you have 1 in UP, one in MP, 1
in Maharashtra, 1 in Gujarat, 1 in Tamil Nadu, one in Kerala like that. And here you see this
distribution centres are working on the principle of cross docking.
We have already discussed in detail in our earlier sessions that what is the meaning of cross
dock?. The meaning of cross dock just to give you a fresh idea that where we do not keep
physical inventory. These inventories are coming either directly from the national finished
goods center or from the regional centre and these inventory are repackaged, redistributed at
the local distribution centres which are working on the principle of cross docking.
And then from these local distribution centres these products are going to customer
distribution centre or directly to the customer store. So you are either using one label of
distribution from this point to this point, where you involve this customer distribution centre
or you can do direct distribution from this point to this point where you are by passing, you
are not using the customer distribution center.
And directly these products are coming from this cross docking location to the customers
store, so these are the changes which are taking place in our modern supply chain network
which are working on the multi Echelon finished goods network. So these are multi Echelon
either you can say in our earlier case you have only one finished good facility, 1, 2, 3 all these
finish good facilities were known as simply finished goods of the company site.
But nowadays instead of having a single finished goods facility you are having 1, 2 and 3
installation for the finish good at the company site. So therefore this term multi Echelon has
evolved and multi Echelon finished goods inventory is coming, before that whatever we have
discussed that before this National finished goods distribution Centre you have the factory
where you are having a warehouse within the plant.
And before that on the left side of that you have various vendor distribution centre, so that
system is already there but in this finish good side the right hand side of the supply chain we
have now developed a multi Echelon system by these three stages which we have included in
290
the distribution system of the products. The very point why should we have the multi Echelon
finished goods network.
I give this question to you that you need to think a lot but why do we have earlier in the
conventional system we were happy with a single type of if you see that we were having
simply this finished goods distribution centre we were not naming this as a national, regional
or local finished goods distribution center and now this finished goods distribution Centre or
now changing into national facilities, regional facilities and local facilities.
And then this is the customer site which is already over there in the conventional distribution
network also, so these changes are taking place because we want to fulfill the customers
requirement with minimum possible time, so if we are serving a country like India with one
single distribution centre, finished good distribution centre the response time will be very
high, so therefore we want to reach to very close and even maintain that economy of scale.
And therefore with national we are going to regional and we are going finally to the local
distribution centre, so this is a type of arrangement we are following in modern supply chain
network. Now as discussed in the previous section, now we will discuss some of the methods
for developing the location of these facilities.
291
We discussing that we need to have 3 particular aspects, one is related to role of the facilities,
the second is related to size and third is related to location where to locate the facility and
now for answering this particular question we have some of the common methods and I
request all my participants that they should learn the uses of excel and we will solve this
model with the help of excel and we will see that how this model is being solved.
So the issue of location can be handled by this gravity method, it is a very simple method like
gravity method is to give you the background of the method, if we have this space where we
want to locate our facilities and now these are the probable sites where we can locate our
facilities, so and these are the locations of my customers, so in the gravity method with
respect to all these sites which are represented by circles I will calculate the cost for moving
cost of logistics from facilities to my different customers care facility.
Now whichever facility, whichever location gives me the minimum cost that will become my
answer. So that we will discuss in this gravity method for location. Now before that one more
issue we will like to answer that is the role of the facilities, well these discussed can be of 3
types manufacturing, warehouses and retail. But there is one more way to look for the roles,
that role is what type of strategic advantage you want to have.
And this strategic advantage can be with respect to the type of role you expect from your
facility strategic advantage with respect to defensive or offensive. Now what do you mean by
defensive and offensive role in case of a supply chain distribution network. Now defensive
292
role is that where I want to develop a supply chain based on conventional thoughts which we
just discussed in the last session, that this type of conventional network is there.
And I want to use the same network for my product also so that is very defensive thoughts
and I just want to reach to the maximum possible customer as per my supply chain strategy
and I do not want to experiment the point which is important, I do not want to experiment in
the case of my defensive role of facilities, so whatever my competitors are using, whatever
type of distribution network my competitors are using I will also go with the same type of
distribution network in a defensive supply chain system.
In case of offensive supply chain system now it is a different ball game. I will not go with this
conventional type of distribution network, I will not go with this type of routine finished
goods distribution centre, routine customer distribution centre, routine customer store etc. I
want to use my network for some kind of advantage and that is the offensive strategy of the
distribution and in that case I may use some very innovative locations.
So that I can get the strategic advantage. My other competitors are using A type of market for
distributing the product but I will use B type of market for distributing the product, so it's a
very different type it is always not sure that offensive type of approach will give you success,
but at least you are not going on the conventional way of distributing the product, that is the
idea behind this offensive distribution system.
The conventional distribution system gives you defensive approach, competitors are doing
that way, I will also do that way, so that I will at least match my competitors performance.
Offensive can take you over to your competitors or maybe you maybe at the totally lost, so
offensive strategy is a kind of innovative strategy in case of distribution where you are
expecting that your facilities will have a role which can provide you a strategic advantage.
That is why the offensive way of distribution. In our literature you will see some time when a
manufacturer, when supply chain manager decides for offensive type of strategy and if it
becomes successful then all other competitors will start following the same strategy and over
a period of time that offensive strategy will become defensive strategy, so you have examples
where you started using ATMs.
293
So ATMs once upon a time was part of offensive strategy because normally you go to a bank,
the physical branch of the bank and to withdraw or to deposit your money, but when ATM
started coming so you can go to a ATM at any time to withdraw your money and then later
on to deposit your money also. So this became an example of offensive distribution but over
a period of time now it is almost impossible to think any bank without ATMs.
So now ATM is a very much part of defensive distribution system of the banking products.
So all success initially we started with some E-portals in India for E-Commerce. At that time
customers was hesitant whether this E-Commerce, E-portals will be successful or not. But
over a period of time now we see almost every company, so initially it was offensive
distribution.
We were very comfortable with bacon motor distribution system, you need to open a
showroom, you need to open a shop etc. but this type of distribution using electronic mode
was almost new focus and products like shoes, sweaters, jackets, mobile phones, laptops, all
coming through this electronic mode of distribution. So it was initially the offensive way of
distribution.
But nowadays it is almost impossible for a company to survive without this E-mode of
distribution, electronic mode of distribution. Therefore it is very necessary that for initial
advantage you think on the basis of offensive way of distribution and later on it can become a
defensive way of distribution. But we also have one example where offensive way of
distribution did no need to defensive of distribution.
You may know a name of company known as 3M. 3M is a very popular name and which is
world leader in the market of adhesives. Now 3M once upon a time lost the product known as
phenyl, phenyl all know is used for cleaning your floors etc. and phenyl normally we all
know is available through hardware shops, is available through general stores, maybe in some
cases medical stores also.
So these are the places normally we all know or the shopping malls also, these are the places
normally where you get the phenyl as a product, but 3M because they were thinking of a
offensive way of distribution, so 3M used a very different ways of distributing the product,
294
3M thought that let us use beauty parlours for distributing this phenyl, now the beauty parlour
for distributing the phenyl is never seen before, so 3M use this beauty parlours.
Now in case of beauty parlour the idea was that normal phenyl for the cleaning purpose is
used by females in the families and females visit beauty parlour and they wanted to promote
their product as it is safe for skin, it is safe for nails, it is safe for nail polish etc. etc. So they
thought that latest promo, latest position this product as a beauty product for the cleaning
purpose, but unfortunately this distribution idea of using the beauty parlours, beauty shops for
distribution of phenyl did not work.
The reason is also very simple I think we all can understand that those who are visiting
beauty parlours are not using phenyl for their household purposes, they can afford maid etc.
and maid made do not visit the beauty parlour, so the idea did not click finally, but it became
every classic example in literature that how to think of offensive way of distribution where
you have totally innovative idea of distribution. So this is with respect to role of distribution
facilities.
Now another important area is location, so we are now going to discuss the area that is
gravity method of location to decide about the location how to decide that which facility
which site should be selected. Now for that purpose the model which we are going to use can
be this type of model where you have x and y the coordinates of the warehouse. So these are
the coordinates of these type of facilities which are the warehouse. So, I want to locate the
295
location of my warehouse, so these are represented by x , y, then these x n and yn, these are the
coordinates of delivery locations n.
So if you see or a graphical access, so this x , y are my delivery locations the coordinates of
these delivery locations are xn and yn and I have to select a site for my warehouse the
coordinates of this are x , y and d n is the distance to delivery location n, so d n is the distance
of this warehouse site to this place, this is dn, so for all these delivery locations I can have dn
with respect to that.
And Fn is the annual load which I want to deliver to a particular deliver location n, so that is
the load with respect to a particular delivery location. Now the formula for dn ,
Now I want to determine actually be coordinates of my warehouse x and y, and for that
purpose I will see that how I will determine, so this is the formula for determining the
location of x and y. Here you see this is being
296
Now once I determine this x and y then with the help of this example you can come that how
can we use for the purpose of determining the location of our warehouse, we have these
different sources and these are the market so I have to take a decision where to establish my
source facility, so these are Pune, Gurgaon, and Jamshedpur. These three areas are the
sourcing facility, this is for some kind of automobile problem.
And the markets are available in Ahmedabad, Cochin, Hyderabad, Chennai and Bangalore.
The transportation cost that is rupees per ton per kilometre is 90, 95, 85, from these supply
sources and to these different markets sources. Now quantity in terms from the supply
sources are available 400, 200 and 600 and in the demand for the different markets are this is
the date. Now on this graphical source the coordinates of supply sources can be 700,000.
297
So if I see this is, if I want to have India on this map, so the coordinates of Pune, coordinates
of Gurgaon, coordinates of Jamshedpur, so you can get the idea of the coordinates, now will
go to the calculation, the model with we just discussed, this is how now you can also start
using the Excel on your computer and this is how we will develop the input data for our
gravity location model.
You can use almost the same table this is the portion of the table is spreadsheet which I will
like to show you that from column B to column F. This is how you will input the data, these
are the sources, these are the markets you also please try simultaneously with me Pune,
Gurgaon, Jamshedpur, Ahmedabad, Cochin, Hyderabad, Chennai, Bangalore, so you use this
data, the transportation cost data, quantity, which are available at these place.
And which are required at this place, this is the source, these are the destination and the
coordinates, so this is the input data in fact this is exactly the last table which was given to us
in the simple format and I have just copied same table in this Excel spreadsheet and I think
you also start using from B to F and from row number 2 to 11. So that you can follow while
doing the calculation at your hand with the help of formulas which we will be developing and
you can see that how these results are coming there.
298
Now for that purpose the formula which we need to use, you see this spreadsheet from B2 to
B11 we already have, now we need to develop some formula which we will build in this
spreadsheet and what will be those formula that we can see here, in cell number G4, so which
is that G4, G4 is this cell and in this we want to determine this distance dn
The formula for dn is written this G4 cell and that is
299
E4 I am giving as absolute value and B15 the dollar symbol here, these are for the relativity,
because as I am moving from this fourth row to fifth, sixth, seventh, eighth, ninth, tenth,
eleventh, I need to change the values of xn.
The values of xn and yn are constant, these are B15 and B16, so I need not to change these
values but I need to change E values, E4 and F4, so E4 and F4 will change and B15 and B16
will not change because of this dollar symbol. So these are fixed and then I will copy this
formula which I will be developing initially for G4 I will be coping this formula from G 4 to
G11, so I will copy this formula from G4 to G11.
So far this entire column I have copied the same formula, so when I am copied the same
formula with this type of equation so in each of these cells B15 and B16 will remain common
so formula will change as per the different values of cells.
And since the initial values of X and Y ,I have kept 00, so these are the calculated values with
the help of that formula. Now coming to second formula that is for cell number B18. In B18
we are putting the total cost, so total cost as you know is the sum product of these three
values, if you remember the formula which we discuss in the slide of initial slide of
gravitational model.
So this formula where you have multiplication of G4, D4, C4,D4 is the total supply available,
C4, is the transportation cost, so C4, D4, G4, so this is one product, similarly C5, D5, and G5
this is another product and so on you will have up to C11, D11, and G11. The sum product of
all these is written like this, in Excel it is written like that from G4 to G11, we will like to
multiply D4 to D11 and also C4-C11.
So G4 will be multiplied with D4, will be multiplied with the C4, G5 with D5, with C5 and
so on G11 with D11 and C11 and the sum product of that is coming in the B18 cell. Now I
want to optimise and for that purpose I will use, this spreadsheet portion is getting after you
have the location of x and y.
300
If I go with this data the location of x and y is at the centre, at the origin because x and y is
given 0,0. Now my job is to reduce this total cost as much as possible and for that purpose I
will invoke the solver and solver will say that if you invoke the solver, so this type of solver
parameter will come, this type of dialogue box will come and here my target cell is B18, this
you will write here B18. So this will be your target cell.
I want to minimise this, so equal to these options will come, so I will select this box
minimise, in some cases will go for the maximize also but here we want to minimise the total
cost so we'll go for the minimum value of the total cost, now minimum value of total cost will
be achieved by changing the values of B15 and B16, so by changing the cell , these are the
decision variables, if you have the knowledge of linear programming.
So we can say that B15 and B16 represent my decision variable, so these are the changing
cells and with the help of this now I will start solving this.
301
And once we start solving this the values will be 700 and 1500 approximately, because once
you have done this then you put the cursor on the solve, click this button and this solver will
give you this type of optimise result, where are now you see your values of all these 1220,
538, 854 are changed , these are 500 1118, 806 and so on and now the values of x and y are
700 and 1500,.
And the total transportation cost is now reduce tremendously from this value which was
477873268, now it is 193549144 and the location of warehouse is 700 and 1500, so you can
see this 700, 1500 is very close to this Hyderabad, 700, 1500 is almost Hyderabad, so you
will decide that the warehouse should be at Hyderabad. So we are stopping at this point and
will see further that what other models are available for getting our optimise location of the
facility. Thank you very much.
302
Lecture-26
Network Optimization Models
So welcome back and we are already discussing about network optimisation models., in our
last session we discuss the use of gravity location method where we decide coordinates of
various supply and demand locations and we want to locate a warehouse, so we basically
focus that we will locate warehouse from where the overall transportation cost is minimum
and using the concept of graphical method where we have calculation of distances on a
graphical plane.
And to minimise the total cost on the basis of that and in that we also saw simultaneously the
use of Excel for particularly Excel solver that how to use Excel solver for such type of
decision making with that data where one manufacturer of automobile company products
wanted to discuss that where should I locate my warehouse and we had some data and we
solve that.
Now moving further we want to have some interesting issue related to a allocation of the
facilities. In last class we were discussing about important issues one was role, one was size
and one was location. So we discussed already in detail about role and locations in our last
session. Now today we will discuss about this size, that what should be the capacity
allocation of my facility.
303
And our gravity location method, the method which we discussed in the last session that was
basically dealing with the locational problem, now today in this session we will discuss about
the size of the facilities, the size of our capacity or the manufacturing facility, so that is what
we are going to discuss.
Now the allocation of capacity to a particular facility, the key cause are there is a fixed cost
of the facility, so whether you use that facility for 2 products or 200 products, there is to be
fixed cost of the facility that is there, then the transportation cost you are using that facility to
transport some product from one end to another end, so that transportation cost will be there,
the production cost obviously.
304
This is a variable type of cost as many number of products you will produce that was
production cost you will incur, then inventory cost and the coordination cost, you need to co-
ordinate with your different other facilities in the network. So depending upon your location
and the size of the facilities this coordination cost will also come. Now the questions which
are there in front of you that you have 4, 5 locations in your radar and you need to decide
which plans to establish.
You want to establish a, b, c, d, e, these 5 location are there and out of these 5 locations you
can make 10 units here, 20 here, 5 units here, you feel this plant is not suitable, this location
is not suitable, so you will not decide to make here and capacity of 10 units for plant E, so
this is what we want to configure the network. We need to decide which plant established and
then what should be the size of each of these plants.
So both the issues we will like to answer in case of this network optimisation models. So for
this purpose you have this type of demand allocation model. Now in demand allocation
model you see that you have various facilities you will like to see which market to be served
from which facility like these are the production facilities and these are the consuming area
and this is a warehouse.
So you need to see that from one to one, this is one, this is 2, 3, so you are serving your area
number 3 from the second facility and maybe from the warehouse also, so that decision of
this type that how your different facilities will provide products to the consumer end, the
305
market that is very important and which supply sources are used by a plant, if this is a plant
for a moment please consider that this is a plant.
Now for this plant what type of vendor distribution system 1, 2, 3 are serving this plant. So
one issue is with respect to market another in with respect to plant that which distribution of
vendor will supply to which plant and which plant will supply to which market. So this is the
configuration of your entire supply chain network. Now for that purpose you are simply
consider this is site i and this is customer j.
So quantities which shift from one source i is source and j is destination, so x ij represents
quantity shipped from i source to j destination, so we want to minimise the total cost, so c ij,
represents the cost of transportation from my ith location to ith destination and x ij is the
number of units move from ith location to jth destination.
So, total cost of logistics involved in moving x number of units from ith source to ith
destination is cij x xij and we want to minimise this and we consider that there are n number of
sources and m number of destination. So therefore i varies from 1 to n and j varies from 1 to
m.
This objective function is subjected to 7 constraints, the constraints are that demand of all the
destinations are completely met, so if I have 1, 2, 3 destinations, so destination 1 requires 10
products, destination 2 requires 5 products, destination 3 requires 15 product. So my total
demand is 15 + 5 + 10 = 30.
306
So this is Dj represents that 30, so total demand is fulfilled and the total demand is fulfilled
from various units which are coming from other ith sources, so x ij giving to from 1 to n to
different js, so that is one important condition that the demand of each destination is fulfilled,
in first request 10 and so I am able to fulfill the demand of 10 from all ith sources.
If second require 5, so I am able to fulfill the demand of second destination from all the
sources, if third requires 15, so I should be able to fulfill the demand of all 15 from different
sources. It is quite possible once you develop this model and solve this model that if I have
sources and destination, so if I have x11 =0, x21=5 and x31=5, so this makes the demand of my
destination number 1 which is 10.
So it is possible that sum of the sources not supply to some of the destination but for the sake
of modelling we write in such a fashion that all ith sources will supply to a particular
destination and demand of the destination is completely better. Now there is a question to
think about is that why we are fulfilling the complete demand in some cases it is also possible
that you have
You have less than equal to sign here that you cannot reply, the meaning of that is you cannot
supply more than the demand of a particular destination in that case, if a destination required
10 units here so you can supply 8 units, 9 units, 7 units, but you cannot supply more than 10
units, that will signify, here what we are assuming in this model that we are fulfilling the
100% demand.
We will develop our model, we will consider our model in such a fashion that we will be able
to fulfill the 100% demand of a particular destination, so in some cases you can take less than
equal Dj, but in no case you can take that this is left hand side is greater than equal to right
hand side, you know market, no market will accept that you supply more than its demand.
So therefore this greater than equal sign in favour of left hand side is never possible. Now
coming to the second constraint, the second constraint is about the supply, a particular supply
307
from a particular source is not supply what it is producing, it cannot supply more than its
capacity, so if you are supply from ith source and Ith source has a capacity of K units, so the
total supply from ith source to various destinations.
To all the destination from j=1 to m should be less than total availability at ith source, so the
second constrain represents this and obviously this xij represents the constant of non
negativity that the value of xij cannot be negative. So these are the important condition of the
model that you are fulfilling, you have the mathematical representation there, but the
meaning of the mathematical representation is that we want to fulfill the 100% demand of
each destination.
And a source cannot supply more than its capacity, so these two things are written in the form
of these type of mathematical expression, source cannot supply more than its capacity, we
want to fulfill the 100% demand of each destination. Now moving further you see now we
have another situation, we have multiple sourcing locations like this particular case where
you have A,B,C,D,E.
And in this case you have multiple sources locations with different type of capacity. So here
you see we want to minimise in that case in the last model we did not consider this fixed cost,
we discussed about the fixed facility cost, so in that model we are only allocating the demand
and we are only bothered about the transportation cost, but we did not consider this facility
cost in this particular model.
308
Now here in this particular case we are also considering the fixed cost of facility and that is
what we are seeing here this step in this case
is the same what we discussed in the previous slide , that is the cost of transportation, but out
of these locations A, B, C, D, E at some location I want to have a plant and at some location I
do not want to have a plant, so where I want to have a plant that is represented by yi.
I introduce a new decision variable y i, and yi signifies whether I want to have a plant at a
particular location or not, so now yi takes the binary variable in our model, it can take only
two values 1 or 0, the value of one signifies that there is a plant at a particular location and
zero means we are not interested in opening the plant at a particular location.
And fi is the fixed cost of opening the plant, so the fixed cost only apply when yi will be 1 and
if yi is 0 so this factor will become 0. So if you are not interested in opening a plant at a
particular location, so there will not be any fixed cost associated at that location. So in this
model you are having more comprehensive D where you are considering the transportation
cost as well as the fixed cost.
And as usual this is the first constrain that you are interested in fulfilling the 100% demand of
a destination, the second constrain it determined that you are interested, you can only supply,
what you are producing at a particular facility, at a particular plant. Now the important thing
is that in the earlier case it was only Ki is this xij from j=1 to m should be less than Ki.
But nowadays you have introduced yi also because it is not necessary that there is a plant or
not, so to introduce that factor that this Ki will be there only when the plant is there, a plant is
not there you cannot supply anything from that location, to take care of that issue because if
D there is no plant so you cannot supply anything from this location D.
309
So therefore to take care of this aspect we have multiply right hand side of this constant, if
you see the earlier slide it is
But now in this case it is
so that this Ki will only be operative if a plant is there. Otherwise it will not and then you
have the last factor that is
the values of yi can only take place 0 and 1.
So you can have we value of y i only the sum of that should be less than 1, therefore you
cannot think of if two different types of the meaning is that at a particular location here if I
want to conceive two types of plant one is of smaller capacity, another is of some large
capacity. So I will only open one particular plant at this location, so take care of this aspect
this third constrain is being added that yi should be less than Ki.
310
And after this we can think of developing a model so let's go for some kind of data, so this is
a data available to us which we will use for developing the spreadsheet model, now here in
this data I again request you all that you can open your Excel spreadsheet and you can copy
this data from this table and you can use the same columns and rows, so that whatever
formula we develop you can exactly follow same formula.
Now, we have certain supply sources, these are available in USA, Brazil, Denmark, UAE,
Vietnam and there are certain demand points. These are same USA, Brazil, Denmark, UAE
and Vietnam. Now the demand of these destinations, for USA the demand is 12 units, for
Brazil it is 6, for Denmark it is 14, for UAE it is 16 and for Vietnam it is 7. Now at each of
these locations USA to Vietnam, you can have two types of plants, low capacity plant and
high capacity plant.
Low capacity plant means of 10 units and high capacity plant means of 20 unit, the fixed cost
for opening a low capacity at different location from USA to Vietnam that is available from
G4 to G8, the cost fixed cost of opening a high capacity plant from USA to Vietnam is given
in I4 to I8, it is different at different location. Now coming to this data from D4 to F8, this
data be B4 to F8, this data is the cost of transportation and production.
If I am producing 1000 units, so cost of production and supplying 1000 units from USA to
USA is 50 rupees, then from Brazil to USA it is 70 rupees, from Denmark to USA it is 120
rupees, this includes cost of production as well as transportation and UAE to USA does 90,
Vietnam to USA it is 150. So this data is available from all combinations in B4 to F8,, so this
is the data at USA to Vietnam you can open either in low capacity plant or a high capacity
plant.
And fixed cost is given demand of various destinations are available to us, now we have to
decide whether to open plant at USA, Brazil, Denmark, UAE, Vietnam, so which of these
locations where will like to open the plant and what will be the size of those plants, whether
these are the low capacity plants or high capacity plants and idea is same, we want to
minimise the overall cost of production, transportation and facility cost. So now let us see I
hope you have copied this data in your Excel sheets.
311
So this data once the same data which we copy in the Excel sheet, now we make a decision
variable table in the same spreadsheet, just below the original data we develop a decision
variable table, so you can also use for making this decision variable in the same format. Here
actually what you want to determine, we want to determine 2 things, one where to open the
plant, second what will be the size of that plant.
And third when we are configuring our network that from which supply stores to which
destination, how many quantities, how many units will be served, so these are three important
questions we will like to answer. So these 3 are our decision variables, whatever we want to
determine that becomes our decision variable. So this is how you will make your table.
So these are almost similar kind of table as you are seeing the upper part of the spread sheet
USA, Brazil, Denmark, UAE, Vietnam, these are the supply sources, these are the destination
and to initiate the problem, to initiate the solution because I do not know where I will open
from where how much I will shift to a particular destination, all these are marked as 0s, just
to initiate the solution process.
312
Now we will develop formulas, the objective is to minimise the overall cost of fixed as well
as variable, so now the formula for objective function will be because this is the representing
the weather a plant will be there or not and these are representing the number of units to be
shipped from my location to a destination, so one cost will come that whatever you are
supplying from USA to USA.
Whatever you are supplying from USA to Brazil that units you will multiply with the cost of
transportation, so this cell B4 needs to be multiplied by B40, that is and similarly for all these
cells they need to be multiplied respectively in this portion of the table. So this is one type of
cost which is coming because of production and transportation. The second cost will come
because of the developing a facility.
Because of the opening of a facility, so this is if you open a facility here means G column i.e,
low cost, so in those places one will come and if you open a high capacity i.e, 4 column in
those places one will come, for the low capacity i.e, G14,..,G18 these to be multiplied
respectively by values of G4,..,G8 and the values of this column here from H14 to H18 need
to be multiplied by the values of I4 to I8 respectively.
So that is what and then you need to take the sum of all these products, so we have written
that from B14 to F18 we are multiplying by B4 to F8, so that is one some product, that is for
production and transportation cost. Then G14 to G18 will be multiplied by G4 to G8 that is
cost of opening fixed cost of opening a low capacity plant and H14 to H18 will be multiplied
313
by I4 to I8 that is the fixed cost of opening a high capacity plant. So that is the overall cost
and this is my objective function which I want to minimise.
So this is the place where I will put the formula for my objective function in cell B31. Now as
we discussed that this will come with some constraints also what we discussed in the
beginning. Now for that developing the constraints in this spreadsheet you see we have given
sum rows and columns for writing the constants also.
The constants will be with respect to demand and with respect to supply, so in case of
demand we will see the unmet demands are there and unmet demand means in these places
you have 12, 16, 14, 16, 7, so all these places you will have the unmet demand. That in the
beginning of this calculation no demand is fulfilled and finally we will see that all these
unmet demand should be fulfilled and initially the capacity of these places are considered to
be 0,0,0.
So we see that where to develop the capacity, so that this unmet demand is fulfilled. Now in
this session we are stopping at this point and in our next session we will see that how to build
that formula into this spreadsheet, so that we can invoke the solver and we can get the result
and we will also see how to analyse these results. So thank you very much.
314
Lecture-27
Using Excel Solver for Network Optimization
Welcome back, in our last session we were discussing about network optimisation problems
and in that network optimisation problem we discussed about one method of this decision that
is gravitational method and by using the coordinates on a graphical plane we thought that
where should we minimise the distribution cost as well as the distance and on the basis of that
with the help of a numerical example also we discuss this problem.
And we saw that out of various market sources and the supply sources where to locate a
particular warehouse. Now in that session also we started about discussion of problem which
is about a global supply chain environment and we have taken some data to explain it on the
global environment, but the same can be explained for a national level data also. Now in this
case we have 2, 3 major issues to be decided about the configuration of supply chain network.
One is there are certain possible supply locations, so we have listed USA, Brazil, Denmark,
Vietnam, UAE, these are possible supply locations and in each of these location you have
option either to have a low capacity plant or a high capacity plant. Now you need to decide 3
things, whether to have a plant at a particular location or not to have a plant that is one,
second is whether to have a low capacity plant there or to have a high capacity plant there.
315
So these are the three important issues, one the location of plant and second the size of the
plant low or high and then another issue is related to configuration of your supply chain
model. In that there are certain demand locations USA, Brazil, Denmark, UAE, Vietnam,
same 5 locations are the demand locations also and in each of these demand location these
supplied location will fulfill the demand.
So which combination of supply will go to the particular combination of destination, so we

need to determine this combination also that from here how many units will be transported to
Denmark, from Brazil how many to Denmark, from Denmark, how many to Brazil or how
many to USA or how many to Vietnam. So depending upon the capacity which we have
decided that we want to have a plant of 10 or we want to have a plant of 20 we will decide
this allocation of capacities two different demand destinations.
So now with this type of problem as I told you in the last session we will build this Excel
sheet and I request all my participants to copy this data on your Excel spreadsheet almost in
the similar fashion. Once you copy this data then if you have some exposure of linear
programming, if you know the linear programming we require similar kind of issue in this
problem solving where we will see there are certain decision variables.
So there are three decision variables, one is whether to have a look capacity plant, whether to
have a high capacity plant or how many units to be supplied from a particular source to a
particular destination, so these are three types of variables, these are three types of questions
we want to determine and for that purpose we call them decision variables, these three
questions are three types of decisions variable.
So this portion of a spreadsheet from A12 to H18, this version of spreadsheet is basically the
allocation of those decision variables and you can see this table and here initially I have kept
everything 0,0,0,0 here, all these values 0,0 are given by us. 0,0 means is to begin with low
quantity is supplied from a particular source to a particular destination, we are taking binary
decision variables as we discussed in the last class for these plants.
If a plant is open, 1 represents the open, 0 represent the plant is closed. So right now initially
there is no plant, there is no open plant and all plants are closed and therefore in the next
issue we will see that how many units we are producing and how many units are going from a
316
particular source to a particular destination. So now once we have decided about the various
decision variables and we have a allocated particular cells for this decision variables.
Now moving further we will see that we have incorporated certain formulas in some of the
cells and the formula is like this, first we will talk of this formula that we are going to build in
cell number B31 and this formula is our objective function. Now objective function is
actually the total cost of this exercise, now if these plants are open (G column), so the cost of
these plants will be getting from the multiplication of G14,..,G18 with the respective cell of
this column G4 to G8.
If high capacity plant is open the cost will be getting from multiplication of H14 to H18 with
the respective multiplication of values from I4 to I8, then the cost of moving and producing
particular number of units at a particular source to a particular destination, so that cost will be
respective multiplication from cell number B14 to cell number F18 with cell number B4 to
cell number F8.
And that is what we are going to write in the objective function, so (B14 to F18 will be
multiplied by B4 to F8), that is the cost of transportation and production+(G14 to G18 will be
multiplied by G4 to G8) that is the cost of opening a low capacity facility+(H14 to H18 will
be multiplied by I4 to I8), that is the cost of opening a high capacity facility. Now this
becomes your objective function.
317
You put this formula of objective function in this B31, please remember always in the Excel
all the formula which you are going to build it starts with this equal to sign, Now in our last
session we discuss briefly about this constraints also.
So here we will discuss in slight detail about these constraints, when we are developing the
model in the last session if you remember there are 2 types of constraints one is we want to
fulfil 100% demand of each of my demand destinations, so my demand destinations are USA,
Brazil, Denmark, UAE and Vietnam. These are my demand destination.
And I want to fulfil this demands which are 12, 6, 14, 16, 7 respectively, so I want to
completely fulfil these demands, and another thing is that I cannot supply to each of this
destinations more than their demand, so if the demand of USA is 12, I cannot supply 13 there,
I will try to fulfil the complete demand of this center, the second is the supply sources, these
supply sources USA to Vietnam, right now you have all the values as 0,0,0,0,0, the reason I
will just explain you.
But we will not to have excess supply, excess capacity of these supply sources. A supply
source if it is of 10 units or 20 units, if it is a 10 unit supply source it cannot supply 11 units
from that, if it is a 20 unit supply source it cannot supply 21 units from that, so it can only
supply 18, 19 or up to 20 in case of a 20 unit supply source, it can supply to 8, 9, 10 units in
case of a 10 unit supply source.
318
So what is the excess capacity available right now we are considering in the beginning of this
problem that there is no excess capacity and therefore the demand at all the destination is
unmet in the initial stage. Now just to see the constraints, these are the 2 constraints, one is
the constraints coming because of supply and another because of demand.
This B28 is this cell, so this is the demand constraints, now I want to fulfill this demand 12
here. Now this demand is the result of whatever is available whatever is the ( total demand of
USA that is 12) – (what all is supplied to USA from different sources ) so USA is getting
from USA to USA, from Brazil to USA, from Denmark to USA, from UAE to USA, from
Vietnam to USA.
So these are the units which USA is going to get this is the sum of B14 to B18, if I subtract
the sum from B9, so that is my unmet demand, what I am trying to say that from B9 where I
have written the demand of USA minus from USA is receiving from the B14 to B18 that is
what USA is receiving, so B14 to F14 this becomes the value of unmet demand that is this
B28.
B28 is equals to this, B9 which is presently 12 and in the beginning all these are 0,0,0, so the
total is this sum is zero, so the value at B28 is coming to 12 and the same logic applies for
Brazil, Denmark, UAE and Vietnam. So I request all my participants of this course that for
these values 6, 14, 16, 7 what you need to do the use of need to copy this formula [B9-
SUM(B14:F14)]. From B28 to F28 and automatically all these values will appear 6, 14, 16, 7.
319
The excess capacity USA is supplying to USA Brazil, Denmark, UAE and Vietnam. USA
supplying to all these different location and this formula, if I see when you are supplying
from USA to USA, Brazil, Denmark, UAE and Vietnam.
In that case but total supply from USA SUM(B14:F14), this is the total receiving at USA,
Now this B14 to F14 which USA is receiving supplying
B22=G14*H4+H14*J4-SUM(B14:F14)
So this formula if you write in cell number B22 this will give you the value of excess
capacity and then you copy this formula from B22 to B26.
Since initially you are having all 0,0 values so therefore these excess capacity are coming to
be zero, so that is why you need not to write these values of zeros and 12, 14, 16, 7, the
formula will automatically give you these values that how this zeros are appearing here and
how these 12, 14, 16, 7 are coming and then once you have this so at then you also write this
objective function formula in this B31 cell.
After that now we go to start the solver activity know we go to start the solver activity so I
hope all of you have done this building of formula and then copying those formula cells
either to B22 to B26 or from B28 to F28, then once you have developed this entire formula
into this spreadsheet then we invoke be solver, solver we start and in that case now my target
cell is this B31 so here once you have this solver dialogue box so this B31 you write here this
is your target cell.
320
As we discussed in the case of gravity model here also you have option of maximum
minimum etc. so we want to minimise the cost, so put this symbol on minimization. Now
what you will do by changing the value because this minimization is possible by changing the
values of decision variable, so your decision variables B14 to H18.
So from B14 to H18 you will change the values of these decision variables and as a result of
that in this box you write the equation for the constraints and these equations for constraints
are one is the values of B14 to H18 all these decision variables are nonnegative they can be
either 0 or any positive value. So the first constraints represents that these are non negative
decision variables.
You need to use this add button here for writing the constraints, B22 to B26 you can have any
positive value of excess capacity, if this positive value of excess capacity is not there if this
constraint sign is reversed the meaning of that can be you can supply more than available
capacity at a particular supply source.
So that is not possible, that is not in reality so we need to put this constraint that from this
source from B22 to B26 you can have any positive value available, no negative value is
possible or if a supply source is completely supply whatever is available here then it will be
at best with zero, so there is no excess capacity whatever is available you are supplying
completely to the to the destination.
Then as we discussed in the beginning that from B28 to F28 which are representing the
unmet demand at different destinations since we have assumed we have modelled in such a
fashion that you are going to fulfill the 100% demand of each of these situations, so therefore
from B28 to F28 we are using this equality sign here, these cells are equal to zero.
In some cases if you feel that you can leave some of the unmet demand, if model suggest that
way that even if you are not going to fulfill the 100% demand of some of the destination does
not matter. In that case you can have B28 to F28 greater than equal to zero but in that case
there will be a big challenge, because we want to minimise the model, so what will happen
minimization will lead maximum there if you put that that there can be a positive value of
B28 to F28.
321
So do you think what the potential is risk with respect your model and it is very interesting
question that you should understand the potential risk with respect to the model. Because you
are allowing positive values here, so positive value can be any positive value it can be as high
as 12 it can be as high as 16 for different different location and when these unmet demand are
there.
So the physical significance is that you are not transporting, you are not producing anything
and that case it will be very interesting that the total cost of this model will be zero as soon as
you change the sign of equality to left hand side greater than equal to right hand side and I
request all the participants because it is now very easy for us to do that you just I am writing
here equality sign.
But you immediately try the greater than equal to sign here and then this G14 to H18 these
two columns G14 to H18, these two columns are with binary sign and the binary sign that
only 0 and 1, these two values are possible so these are the four constraints we have put here
and after putting this for constraints we click the solve on button and once you solve this
button.
And before I show you the next slide you can see the because you are put greater than equal
to sign here and if you put solve quick then in that case you will see that the total cost of this
model will come zero and the excess capacity is will remain as it is and unmet demand will
remain as it is, because we want to minimise the cost and zero is the lowest possible cost for
any model it is a dream type of situation.
So nobody is going to produce anything, nobody is going to transport anything and you are
going to have zero cost of your model, and you know that is not the objective of this model.
Therefore we are building this type of constraint that left hand side is completely satisfied
and there is no unmet demand, yes it is possible in some situation that you do not leave this
right hand side totally open.
You can feel, you can make some kind of percentage that you need to fulfill at least 80%
demand, you can try some of the variants of this question that I can say that you need to
fulfill 80% demand of each of the destination for different different destinations you can say
322
that you need to pull 80% demand of each of the destination or it is also possible that you can
have more number of constraints.
And you may say that for USA market and Denmark market you have to fulfill 100%
demand, for Vietnam market you should fulfill 80% of the demand or even more than it is
very good, for Brazil market again 80%, UAE market 75%, so for different different markets
you must set different criteria and all those criteria need to be built into this constraints, so
this is a simple generic model but more customised more specific model can be built by
writing the constants for that in this constant dialogue box.
Now going further once you have started your solver and the calculations are going on it will
take some fraction of seconds to complete the calculation and then this type of model will
emerge, this is the solution of the model where you see let us first see where are we going to
open the plant, so where we are not going to open any low capacity plants, for all low
capacity plants it is 0,0,0,0,0 here.
So there is no low capacity plant, you have three once in high capacity plant one for Brazil,
one for UAE and another for Vietnam. So you are going to have 3 high capacity plants in
these locations, you have some values in these two cells, one for USA and another for
Denmark, so these are 10-13, these are because of internal calculation errors in excel
So these are always you can say equal to zero non-significant, so we consider them also 0, so
in fact you will have 3 high capacity plants one at Brazil, one at UAE, one at Vietnam. So
323
that is about the size and location of your plants. Now the second question we want to answer
that from which facility how much you will ship to a particular location, now you see in USA
you are not going to have any type of plant neither small nor bigger.
So you are not going to ship any unit from USA, so USA to all other destinations it is 00000,
you are going to have a bigger plant of 20 units at Brazil, so from Brazil your shipping 12
units to USA, you keep 6 unit of that in Brazil itself because the demand of the Brazil is 6, so
Brazil will fulfill all its demand from the local capacity, so 6 will come to Brazil, and 2 will
go to Denmark.
So 12 + 18 +220, so the Brazil capacity is no exhausted and Brazil will not supply anything
to UAE and Vietnam markets. Now again in Denmark there is no plant, so you are not going
to say anything from Denmark to any of the destination, you have one bigger plant in UAE,
now the demand of UAE is you see interestingly only 8 units of its own plant will be
consumed in UAE and remaining units out of that 12 will go to Denmark.
And that is how 12 + 8 the UAE capacity will be exhausted and in case of Vietnam again you
are going to have a bigger plant, Vietnam one requirement is 7 units, so out of 20, 7 will be
consumed at Vietnam itself and remaining 8 will go to UAE and 7 + 8 makes 15 so Vietnam
will be left with 5 units of excess capacity that you are seen in the cell number B26 and in all
other cell USA and Denmark these are 10-12.
So almost equal to zero and for Brazil and UAE also you have zero values, so infact you can
say USA, Brazil, Denmark and UAE have no excess capacity and you have a excess capacity
of 5 units at Vietnam and then you see the unmet demand in case of this B28 here you are
showing some kind of 10-12 of value where it again means non-significant it is zero.
And for all other Brazil, Denmark, UAE and Vietnam these are zero and demands the
meaning is that with this model now you see you have three plants your 5 units are excess
and the cost is coming 23470 rupees, so this model is a very perfect type of model where you
decided where to locate your manufacturing facility, what will be the size of that were
factoring facility and how are you going to configure your supply chain network which
facility will supply to which market.
324
And the customisation of this model as we discussed can also be done depending upon our
requirement that if we want to have in this particular case you are making 5 units additional
at Vietnam because you wanted to fulfill 100% demand of all these decision, now you see if
it is possible that I do not fulfill 100% demand of some of the location so maybe at Vietnam I
can work with a smaller facility.
And in that case I may remain with some of the unmet demand in some of the location and
that will reduce my cost of supply chain but it depends upon my supply chain strategy
whether I want to completely fulfill the or I can bear with slightly lower level of service. So I
think we can try various versions some of the version which I can suggest you that you
should go with the different type of constraint combinations here in this case.
One constant combination you note that we can have you can have any kind of unmet
demand available. In that case there was no unmet demand left but the another constraint can
be you have some unmet demand left. In that case you will change this sign of equality in to
greater than equal to just see the results, the result will be zero cost here. This is case number
one.
Now case number 2 you try that the demand at Vietnam did not to be fulfilled 100%, you can
have 80% demand of Vietnam to be fulfilled and in the same case UAE only 80% of the
demand need to be fulfilled. So you see that you need to apply 2 additional constraints from
B28 to F28 it is not that in case of B28 to D28, it will be equal to zero and you need to write
two additional constraints one for UAE that is for E28, and another F28.
325
So you see that what will be those two constraints for UAE the constraints will be that.
E28=0 , E28≥.80(E9). Similarly if I want 70% of the Vietnams demand should be fulfilled, so
it means F28 ≥0.75 F9.
So you need to customise this as per the situation and then this solver activity will help you in
getting the results very quickly and you can try different type of conditions depending upon
the supply chain strategy and then get the results whichever result gives you the minimum
total cost that is what we want to have in our supply chain network. Thank you very much.
326
Lecture-28
Uncertainty in Network Design
Welcome back, in our last session we discuss the different types of issues in developing a
supply chain network, developing a supply chain network is meaning the location of
facilities, the role of facilities and the sizes of facilities. But all those decisions which we took
in our earlier session are with respect to a very certain type of environment. But we all know
the environment is not certain.
And particularly in case of India in recent past we all have seen that there are lot of
uncertainty in the external environment and those uncertainties must affect the supply chain
decissions, if you do not take into account those uncertainties it is very difficult to develop a
very responsive type of supply chain, responsive the meaning here is the supply chain which
can handle the challenges of the existing environment.
So for that purpose it is quite necessary that we need to evolve our models, we need to evolve
our calculations, so that our decisions are with respect to uncertainty of the environment, so
now in this session we will focus that what are the different types of uncertainties and how to
design a network, how to design a supply chain with respect to those uncertainty in the
external environment.
327
so let us see the outline of this lecture and here will see the impact of uncertainty on the
network design, and when we talk of impact of uncertainty on the network design, so first
thing which we are looking since last two, three lectures that supply chain decision include
investments in number and sizes of the plants than your transportation facilities the number of
warehouses you want to have.
And probably after warehouses the role of customer comes, so if I am Maruti Udyog Limited,
if I am Hero Honda, if I am LG, if I am Samsung. So I look up to that stage where I need to
provide the finished goods in my warehouse, so these are the decision which I am looking
with respect to network design and where uncertainty can play a very important role. Now
another important thing with respect to network design that you cannot change these
decisions.
328
The very important point here is that these are long term decisions, these are very long term
decision and it is easily not possible to change these decisions in the short term, you are
developing a facility like let say a manufacturing facility, but developing a manufacturing
facility a plant for making motorbikes will take 2 years or 3 years of time.
You are developing a facility like refinery, it may take 4 years of time, now when it is taking
such a long time in just erecting the facility, so you can very well understand that it is not
possible that you should change these decisions in the short term, so you have to be very
thorough, very practical, pragmatic, so that whatever decision you take these are long term
decisions.
So that is the second important that whatever decisions we are taking with respect to
uncertainty, these decisions cannot be changed in the short term, these are long term decision.
Then another important point is there will be a good deal of uncertainty in demand, demand
increases, decreases, you talk of demand of these air filter which we wear on our like mask,
last year in the month of October and November in NCR area where was a severe problem of
smoke and demand of mask the air filter increases all of a sudden to every high level.
But do you think can we develop a new manufacturing facility all of a sudden to fulfill the
that excessive demand of air filters which was high uncertain nobody has predicted no
weather forecast was there that this type of demand may come, but all of a sudden there was
something in the external environment, the happening of Deepavali, the crop cutting at
Punjab and Haryana and all that things happen simultaneously, there was no wind in the year.
So that it can take away all that pollutants from the air and it was like night in the day and as
a result demand of those production increases, so there are fluctuations with respect demand,
with respective prices, you have another example same in India the issue of demonetization
came in the November last year and as a result of that prices of so many products particularly
if I talk of gold, the prices of gold started crushing down.
The gold was considered to be one of the very important type of investment in India but
because of the issue of demonetisation the prices of gold starts decreasing, so the uncertainty
is there with respect to prices also, you have a increasing trend of prices, but all of a sudden
329
something happens and prices may decrease or something may happen prices make
increased, in India many times we hear stories about escalating prices of audience, escalating
prices of pulses.
And that is also the part of uncertainties with respect to prices, then exchange rates the dollar
versus rupee, pound verses rupee, Yen verses rupee, so that type of uncertainties are also
there and then the uncertainties with the competitive market over the lifetime of a supply
chain, that how the competitive landscape of the market will change, these uncertainty is will
also be there.
We know few years back Nokia was one of the popular market, one of the popular brand of
mobile phone, so in country like India if you ask anybody which mobile you want to have,
Nokia was a popular answer, but now what is this situation Nokia is almost nowhere, the
Samsung and then lot of Chinese brands and lot of local Indian brands are also making a
good space, a good presence in the mobile market of India.
So competitive landscape of the market of a particular product in the supply chain over its
time is also very very uncertain. Once upon a time Bajaj scooters was the Pioneer two
wheeler of this country and now days even scooter product is nowhere in the market, forget
about Bajaj, Bajaj is also shifted from scooter to motorbike, so there are plenty of such
examples where over the lifetime of a product, the competitive situation changes.
And initially we know that Sony cameras used to compete with the Kodak cameras, Kodak
cameras used to compete with Canon cameras, but nowadays what is happening Sony
cameras are competing with the camera of Apple which is inbuilt in the mobile phone, so that
is a new type of competition which is coming up. So it is not necessary that competition may
come from your own product over the lifetime of a product, competition may come from a
different category of product.
One newspaper is not competing with another newspaper but one newspaper may compete
with another sports, news paper may compete with the televisions, newspaper be compete
with social media, so you know that different types of things are happening in the
environment and therefore this uncertainty with respect to competitive market over the
lifetime is also a very important type of uncertainty.
330
And you need to be ready to face the challenges coming from the competition in your market,
therefore what we say that building flexibility in to supply chain operations allows the supply
chain to deal with uncertainty in a manner that will maximize the supply chain surplus, the
answer for handling the uncertainty is to build to infuse the element of flexibility in your
supply chain, with flexibility in your supply chain you can probably handle all these issues.
But what is the flexibility, what is the optimum level of flexibility, how much flexibility we
should built into our supply chain, a system which is completely flexible cannot survive a
system which is completely rigid also cannot survive, so you need to have a balance of
rigidity and flexibility, so in this particular session and then few more sessions after this we
will discuss that how much flexibility at what stage of the supply chain that flexibility should
be included.
And the impact of uncertainty on the network design is now concluded that for that purpose
we need to have flexible supply chain, without flexible supply chain you cannot handle
uncertainty of the network design.
Now for that purpose there are different types of modelling approaches to develop the
concept of uncertainty in the supply chain network. One of the approach is discounted cash
flow analysis, now supply chain decisions as discussed just now all very long term decision,
when we are talking of developing of a facility, in that case it is certainly a very every long
term decision.
331
You cannot developed a new depot, you cannot develop a new warehouse in one night, you
can take some decisions which are for the short term, you can start outsourcing all of a
sudden, you can stop outsourcing because demand is not there, so that type of decisions are
possible within very limited time, but developing a facility like warehouse, developing a
facility like manufacturing is not a short term decision.
And therefore supply chain decisions are for long time, so they should be evaluated as a
sequence of cash flows over that period. Now when I am developing a manufacturing facility
and I feel that the life of this facility is 20 years or 25 year. I am developing a warehouse I
consider the life of warehouse is 30 years. In 30 years time how much cash flow will be
generated on the basis of that we will take decision.
So this is one very simple method where over the alternative which are available to us. We
will calculate that which alternative is giving how much cash flow to us, and depending upon
that cash flow we will take the decision that whichever alternative gives us maximum cash
flow, maximum positive cash flow that is the choice for our network decision, so now what is
the discounted cash flow analysis.
So this DCF discounted cash flow analysis evaluates the present value of future cash flows
and allows managers to compare different cash flows in terms of their financial value. The
meaning is that if you have some money today in your pocket that is more valuable than what
you are going to get tomorrow, the last point says this based on the time value of money this
is a very important concept in the classes of finance and accounts.
You must have heard about time value of money even in the classes of economics also you
have heard about time value of money, in the classes of project management also you would
have heard about time value of money. So will not go into detail of time value of money
analysis but time value of money says that a rupee today is worth more than a rupee
tomorrow, so if you have 1 rupee today and 1 rupee can buy 1 Kg of rice for you.
After one year again you have 1 rupee in your pocket but at that time you will not be able to
buy 1 Kg of rice with that rupee, you may be able to buy 900 grams, 800 grams of rice from
that 1 rupee. So it is now the purchasing power of that rupee has decreased by 100 grams or
332
200 grams over 1 year span. So that is the time value of money that rupee which is today in
your pocket it is more valuable than what you are going to get tomorrow.
So based on that we will calculate in this discounted cash flow analysis we will calculate the
present value of various future cash flows over a period of time which we are going to get
and then we will see that what is the total present cash flow value of various alternatives, so
now going further the formula for such calculation if we see so in that K, this K is the rate of
return, normally in our day to day discussion we say that K is nothing but rate of interest.
In our language of supply chain or in the language of finance and accounts we say K is rate of
return, so this becomes rediscounting factor 1/K.
333
The meaning of discounting factor is if I am today having this is 2017 1 January, so I will get
on 1 January 2018 Rs. 100, so I want to know what will be the value of this 100 rupees which
are going to get on first January 2018 on this first January 2017 and as I am seeing the value
of today money is more than what I am going to get tomorrow, so what is the meaning of
discounted factor that today's 90 rupees, todays 90 rupees let us say I take simply a statement
just to explain you that today's 90 rupees maybe 100 rupees of tomorrow.
so I need to calculate that if I am going to get 100 rupees tomorrow what is the value of that
100 rupee in today's context, so that is the use of discounting factor so, present value of that
money= 100x(1/1+k) if it is let say rate of return is 10% so 100x(1/1+0.1), so 100/1.1
The second formula for total net present value, I am going to get not only on 2018 the project
on my warehouse my factory is for 10 years, so 2018, 19, 20 and so on, for next 10 years I
am going to get some cash flows, it is 100 rupees then 1,000, than 10000, then 50000 and
75000 so on like that.
So I will like to calculate the present value on first January 2017 of all those future streams of
cash flows and for that purpose here I come with this formula
Where Co that is seen is the cash flow which I am getting today in many times this Co is
negative because today I am not getting any positive cash flow, the cash flow which I am
receiving is represented as a positive cash flow and cash flow which is going from my pocket
is represented as negative cash flow.
So the first year because I am putting my money into that facility development so in most of
the cases we will see that Co is negative and then this second term in this expression
represents the sum of all the future cash flows and their present value, so CT is like for 2018
we discussed Ct is 100 rupees, t becomes 1, and i.e, 100x(1/1+k), if I am getting in first
January 2019 let us say 1000 rupees.
So in that case I want to calculate its present value, so it will be 100x(1/1+k) 2 because now it
is 2 years back, so if there are total T periods for which I am going to get the cash flows so
this is how you will get T sum of those future cash flows with their present value, so this
becomes the total net present value in this case of discounted cash flows.
334
So this we will compare this cash flow often this total net present value we will compare for
different alternative, different supply chain design options and on the basis of that the option
which gives me the highest net present value, the highest NPV that will give me the highest
financial return, because of financial return of the supply chain surplus is the objective which
we discuss in the very first session that we want to maximize the supply chain surplus.
So we will do this calculation so that my supply chain surplus is increased and with the help
of this example of one company known as ABC logistics, we will see the use of this NPV
method and in this case the example is that how much space to lease in the next 3 years this is
the objective, but the demand is 100000 units it requires 1000 square feet of space for every
1000 units of demand.
The revenue is 1.22 dollar per unit of demand, now what is the alternative which are
available, alternatives are decision is whether to find a 3 year lease or obtained warehouse
space on the spot market, the 3 years lease cost is 1 dollar per square feet and spot market
cost is 1.2 dollar, 1 dollar 20 cents per square feet, so spot market and slightly costlier and the
rate of return is 10% which is also written as 0.1.
335
So this is the data available to us and now we will use this NPV method, now for leasing
warehouse space on the spot market, one option is to consider the spot market and is other
option is to have 3 year lease, now taking the calculation of the spot market in the calculation
of spot market the expected annual profit is one lakh units.
And the data for revenue from one unit is 1.22, so this is the profit and this is the cost of the
space 1.2 square feet so you because 1000 square feet is useful for thousand units of demand
so if you have 100000 units you require 100000 square feet, so total area is this much so you
have the annual profit of 2,000 dollars. Now cash flow for 3 years will be 2,000 dollar each
year.
When you are going for the spot market no list, so 2000 in the first year, 2000 in the second
year and 2000 in the third year, for each of 3 years you are getting 2000 and for first year the
rate of return is 1 + K i.e, 1.1 then for the second as I explained there it because 1.1 2 so it is
1.21. The calculation yields the total net present value of 3 years cash flow is 5,471 dollars.
That is with case of spot market, now before I go for the next slide I request my students that
they should practice the same calculation with the lease option and I am waiting for a
moment that you complete your calculation and then you can match your results with our
results that when we are going for using of warehouse.
336
Expected annual profit=100,000x$1.22-100,000x$1.00=$22,000. Because of that other cash
flows 22000, so our total net present value of cash is $60,182 and you can see NPV(signing
lease) is $5471 earlier, so there is a difference of around 54711 which is significantly higher.
So now you have a very clear option that you will decide to sign the lease, you will not go to
the spot market because in case of sport market our profit has just 5471 dollars and hear the
present value of our profits is 60,182 which is much higher, can I ask you one question that
this is a case of again lot of certainity in the market.
But it is quite possible that you will need to rethink your decision because of uncertainties in
demand and cost. Now in case you have signed this is lease for 3 years so you need to pay
this much rent because you have taken the space, so whether you are able to sell this much or
not the respective of that you have to pay this much amount to the other party to the provider
of the space.
And in that case your profit may your net present value be significant go down, it is also
possible here for 3 years we have taken a flat rate of return 10% it may also change with time
and some predictability you can do that next year it will be 10% rate of return and after that it
will be 15% and then it will be 16% etc, etc. so in that case also you can have different
denominators here.
If you have different rate of return, so according to those different rate of returns you may
have different denominators because here with same rate of return for all the period we have
same value of k in denominator for all the years. Now what I was saying that in spot market
though here it looks that our profits are very very less, our net present value is just 5471
rupees in case of least arrangement it is 60182.
But it quite possible that you are not able to use this much space, you see your data that you
require 1000 square feet of space for 1000 units. Now if in a particular year on a day to day
basis after second year you realise that the demand for next year maybe just 10000 units so
you will take only 10000 square feet for 10000 units.
And in that case you can save lot of money which you are giving to your space provider but
in case of these lease arrangement you have no such option, so therefore uncertainty in
337
demand and cause many times forces not to go for long term relationship, not to go for long
term this type of lease relation, though here with the data because we have not infused the
element of uncertainty in this example yet.
We are only considering that you have all rate of return, you have a fixed demand, you have
fixed cost everything is fixed and in that case obviously with this simple calculation we can
conclude that this lease arrangement is made is a suitable arrangement, but when we go for
uncertainty there will be two types of representation of uncertainty that we will discuss in our
next session.
So here without any kind of uncertainty our discounted cash flow model is a very suitable
model to take decision about different type of supply chain alternatives. In our next session
we will see that how we are going to inbuilt the element of uncertainty in the network
decisions. Thank you very much.
338
Lecture-29
Network Design in Uncertain Environment and Flexibility
Welcome back and in this course of supply chain analytics in this session we are going to
discuss the network design in uncertain environment. Nowadays we know lot of type of
uncertainties are there in the external environment and time and again we have discussed the
importance that we need to have a supply chain which is adaptive to these changes. So in this
session with respect to network design that how can we make those types of network which
can handle uncertainties of the external environment.
So the outline of the sessions are we will be discussing this in next two, three sessions and the
impact of uncertainty in the network design, then we have different types of project options
and for that purpose we will see the role of cash flow analysis for selection of a particular
type of projects for the network design with the help some uncertainties we will see that how
to make the mathematical model for making the optimum price.
339
And to start the discussion of uncertainty in a network let us see what are the different type of
impacts uncertainty can create in a network design. Now supply chain design decisions there
are different type of design decisions and some of the decisions which are short term which
are intermediate term and which are long term. When we are talking of network design these
are long term decisions.
And these decisions include investment in number of and the size of the plants it is related to
number of transportation facilities like your trucks, lories, trailers etc. then the number of
warehouses. All these all these decisions where huge amount of capital is required, so
therefore issues related to cash flow analysis become important in handling this long term
decisions because of the amount involved in this things.
This decisions cannot be easily changed in the short term, you are making a warehouse in
Punjab and now another day it is almost impossible for you to shift that warehouse from
Punjab to Gujarat, so these decisions are very long term decisions and you need to be very
very careful in selecting a particular location, the same thing is with the size also, you need to
consider the demand for long period of time.
And accordingly you need to see that what should be the optimum size of those facilities,
warehouses, your plants, your trucks etc, and therefore many time you may think as a
solution strategy that we do not build these resources at we try to outsource many of such
resources because if you put your money it will be very difficult and in this season itself we
will see that how to minimise our investment in these issues.
340
So that you have flexibility in dealing the uncertainty of the environment, so we cannot
change this decision in the short term, then there will be a good deal of uncertainty in
demand, demand is highly uncertain nowadays and it is very very difficult for any kind of
mathematical model to determine when a Black Swan will come and therefore demands are
highly unpredictable.
We use good forecasting, we have already discussed in our earlier session about various
forecasting techniques where we try to model different type of characteristics of our demand
data but even then you will have some situation because like for example disasters you have
situations of disasters where all of a sudden demand of a particular product may increase, in
last year during the time of Deepavali we have discussed that because of pollution in NCR
and the demand of mask.
So that you can breathe easily with that increases to a very high level and that was highly
unpredictable there was no mathematical support to that type of increase in demand,
similarly in many other area of country different type of phenomena take place and
accordingly demand of some of the product may increased to a very high level
demonetization happened and as a result of demonetisation demand of various products
decreased to a very low level.
There is a plant of one of the largest two wheeler manufacturer in Uttarakhand and that plant
remain closed for more than one month because there was no cash available in the market and
as a result of that demand of that product went to a very low levels, so these are uncertainty
with respect to demand and these uncertainties are continuously increasing and we all know
about this VUCA and because of VUCA we have more uncertainties in present environment
with respect to demand.
Then you have prices also because of a lot of globalisation, because of a lot of international
issues we have seen that there was a time when the prices of crew increase to as high as 140
dollar per barrel and it decrease to as low as 20 dollars per barrel. Now this huge shift of price
include created a very volatile price sensitivity to the commodity prices also, so you are also
very much uncertain about the prices related issues also.
341
And certainly these prices related issue will create impact on the supply chain. Because you
have developed the supply chain to a particular level of final price end price, but you see
because all of a sudden the raw material price has increased, all of a sudden a new competitor
has come to the market we have a very good example in this particular case, there was a very
sturdy mobile service market in India.
All of a sudden Reliance jio came into the market and Reliance jio started offering 4G
services at a very low price and initially without any charge it was always free and you can
use as much as like and now they have also extended the services to 31st of March. Now at
this type of competition when Airtel was there, where Vodafone was there, BSNL was there,
Idea was there and all of a sudden Reliance jio came and started giving services without any
charge.
So this type of price uncertainty create huge amount of pressure on other competitors and
nowadays this uncertainty is again increasing because many competitors, many companies
many service provider ,many marketers they try to complete on the basis of low prices, so
price uncertainty is also impacting our supply chain decisions to a great extent, then exchange
rates international environment is creating lot of fluctuations in the exchange rates.
The dollar verses rupee price, dollar verses pound price, pound verses rupee price, euro
verses rupee price, all these fluctuation are happening almost on the daily basis and because
of these uncertainty our Global trade and not only Global trade because we are living in a
globalised economy our internal trade is also getting affected because some time it is cheaper
to import from Europe, sometime it is cheaper to import from UK, sometime it is cheaper to
import from USA.
So now when you have different countries giving you different type of cheaper products, so
you cannot have one single source of supply for your supply chain, you need to develop your
suppliers in USA, you need to develop suppliers in UK, you need to develop suppliers in
Europe and maybe in Southeast Asian countries also, because exchange rate will create a
pressure on you to always look for cheaper source for your raw material, cheaper sources for
your other kind of some components etc.
342
So and it also creates a pressure on you because if your currency becomes dearer, if you are
currency becomes strong, import from your country to other country means your exports will
be costlier, so you may lose your market in that case, so there are uncertainties with respect to
exchange rate and these uncertainty is may help, may create pressure you can find more
cheaper sources of raw material or your exports may become costlier.
So both these things are possible and therefore we need to have our supply chains adaptive to
these changes also. Then the competitive market over the lifetime of a supply chain network,
the competitive markets are also changing, new market are emerging and old market are
dying, you have a very interesting situation when one digital camera was competing with
another digital camera, but nowadays digital camera is competing with a mobile phone.
Because mobile phone is also offering a very good camera options and therefore you have
new competitive markets, the uses of mobile phone they are the competitive market for the
digital camera, so these type of changes are also happening in the supply chain, so new types
of markets are emerging, we earlier had the news papers printed on papers but nowadays all
those news feeds are available in our mobile phone.
So now the news paper market is competing with the mobile markets, the application market
and therefore new types of competitive market are also coming in the lifetime of a product
and you have to be competitive you need to understand the impact of these things on your
supply chain, so you can get uncertainty from demand, you can get uncertainty for the prices
of your raw materials, prices of the labour, prices of the machines, prices of the market.
You can get uncertainty from the exchange rates you can get uncertainty from the different
types of competitive markets, another interesting example of the competitive market in some
of the cities we see that Local boys are doing deliveries for pizza or other food items and now
companies have these type of policy that they will deliver within 20 minutes of your order 30
minutes of your order and these delivery boys are very much expert in knowing the shortest
route to deliver a product to a particular destination.
Now most of the E-Commerce companies are trying to hire this delivery boys so that they can
also use their expertise in using the shortest route to deliver the product and therefore this is
giving a good amount of uncertainty to those retail food joints those were involved in the
343
delivery to the door step of the customers, so these types of uncertainty who can think of that
a delivery boy delivering the pizza can be a very important source of uncertainty.
Because he can be hired by some other E-Commerce portal who is also involved in delivering
the products to the customer, so these type of uncertainties are always happening and we need
to be very much adaptable to these types of challenges. So what is the solution therefore we
need to build flexibility in to supply chain operations. Flexibility is said that one key answer
for handling the uncertainty is in your supply chain.
Your flexibility in the supply chain operations allows the supply chain to deal with the
uncertainty in a manner which can maximize our profit, uncertainties are bound to increase,
so you cannot control these uncertainties because these are coming from the external
environment, now our effort should only be directed to align our supply chain, our supply
chain operations to match these uncertainties and for that purpose flexibility is considered to
be the keyword.
This flexibility is considered to be the keywords that if we can build flexibility in the supply
chain, this will help us to handle the impact of uncertainty in this supply chain. Now building
on this idea that we need flexibility in the supply chain, if I see the traditional role of supply
chain, we have gone through so many sessions of supply chain, but if I want to summarise the
supply chain.
344
So the traditional role of supply chain is we want high speed of delivery, the product can be
delivered to the customer with minimum time and low cost. The cost of delivery should be as
low as possible, so we want higher speed and lower cost, that is the supply chain for most of
the companies. The meaning of supply chain management is these two words, we want higher
speed of delivery and lower cost of delivery.
However, there is a contradiction in this many of us may feel that as we increase the speed of
the delivery as we want faster mode of delivery, if I want to deliver a letter through rail the
cost will be less and if I am delivering that letter using aircraft as my mode of transportation
obviously the cost will be more. So I need to find some kind of optimum balance between
these two things that I can achieve a moderate level of speed and a reasonable cost to achieve
that level of speed.
We have already discussed this under the responsiveness and efficiency spectrum where how
to achieve the balance between speed and cost was discussed in detail. So it is just to
understand that these are the two important drivers for supply chain for a company highest
speed and low cost. Now it is high speed and low cost, these things worked perfectly fine in a
supply chain which is operating under steady state conditions.
Where you have less uncertainties where you have every fixed of environment, in those
situations this high speed and low cost can work well, steady condition means that conditions
are not changing with respect to time, though conditions may change with respect to location,
these conditions may be different in Japanese market, these conditions may be different in
European market, these conditions may be different in Indian market.
So these changes are possible, so steady means with respect to time, those conditions which
were there in India 3 years back the same conditions are there today also and same conditions
will remain after 4 or 5 years, those conditions were there in the European market, these
conditions are there today also at after 4 years also the conditions will remain as it is. So
steady condition there are two meanings, one is with respect to time, one is with respect to
location, here I mean is with respect to time.
So at a particular location with respect to time, the conditions remain same whatever level of
demand is there it is almost constant or it is having a particular type of trend which you can
345
easily model, if there is a seasonality in the demand you can easily model that seasonality in
the demand. So whatever type of characteristic is there in your demand data you can easily
make a pattern of that characteristic.
Whatever is the fluctuation in the price you can already predict those fluctuations and
accordingly you can take the decision, so steady conditions are as per the time at a particular
location and in this case the entire supply chain is focused on Economy of scale that is to
achieve the low cost, so you can plan what should be the optimum size of your warehouse,
you can plan what should be the low size in a truck.
All these things what should be the optimum batch size of production, what should be the
optimum quantity you should procure at a particular time POQ, all these things will help you
to achieve economies of scale which is directly linked with the issue of low cost, then
delivering quick supply for the least amount of money, delivering quick supply means you
know how to change products from one place to another place because you have a very
routine pattern of demand.
So you can achieve the high speed also, so this economies of scale and delivering quick
supply these things will help you to achieve both these things higher speeds and low cost in a
steady state condition where much fluctuations are not there. But what we are discussing this
is not possible because if changes are sudden in the demand then it is difficult to achieve high
speed and low cost, because to meet this changes if demand increases all of a sudden.
So you need to go backward into your supply chain, so that you can fulfill the additional
demand you may subcontract, you made give overtime to your workers or otherwise some
backlogs will be there, if backlogs are there you will incur extra cost, even in the case of
overtime, even in the case of sub contract your cost of offering the product will increase, so
when the changes are sudden in the demand, in the prices, in the exchange rates.
All the sudden changes will be a challenge to achieve higher speed and low cost, so therefore
it is important to understand the role of flexibility and flexibility is as we discussed is one of
the key answer to handle the problem of uncertainty in your network design, now what do we
mean by flexibility in a supply chain it is very important to understand that what is the
meaning of flexibility.
346
You will go to Google and you will just type flexibility in supply chain thousands of answers
will come in front of you defining flexibility in supply chain, now to explain flexibility in
supply chain I take the help of one of the article published by Lee in Harvard Business
Review in year 2004.
Lee published this article in year 2004 and on the basis of that this article we have these three
important meaning, three important dimensions of flexibility in the supply chain, these are
supply chain should be adaptable, these alignment is the another important dimension and
agility is the third important dimensions for achieving flexibility in the supply chain. Now let
us discuss meaning of these three things one by one.
Adaptable means the supply chains should be able to adapt to the changes, you need to make
changes in the real time and here comes the role of our supply chain analytics, the art of
analytics where we are doing the predictive modelling, these things will help us in making the
supply chain adaptable at the real time, you are not taking decisions based on some data of
1847, you make decisions based on real time, what is happening today.
Unfortunately IT is a very important enabler which is helping us in making our supply chains
more adaptable with the real time happening, things like POS point of sales data these things
will help us a lot in future about making real time analysis, so adaptable supply chains are 1
important dimensions of flexibility in the supply chain, what is the meaning of adaptability
347
we adjust supply chains designs to meet structural shifts in market, modify supply network
strategies products and technologies.
We are able to make changes in our supply chain with respect to structural shift in the market
modify supply network strategies and products and technologies, now give example how you
can achieve the adaptable supply chain. In India we used to have brick and mortar stores for
most of the products, most of the products we go to the market, we have cash in our hand or
cheque in our hand and we go we select a product pay the price of the product and purchase
it.
This is what the system used to follow, but last few years we all have seen that how different
E-Commerce portals have come to the market and because of that many of us have started the
convenience, the better shopping experience, shopping at your time and the variety, the
discounts so many different factors are there which are helping in improving the performance
of these E-Commerce portals.
So now the market structure is shifted, we are continuously talking about how to eliminate
the middle man, so the role of middle man we do not want in the supply chain and in this
changing scenario when the structures are shifting, now these middle men have also started
offering the special services through E-commerce and this is a good example that how you
can be adapting yourself to the changing requirement.
Because more and more literature you will find more and more discussion you will find and
in almost all the discussion they personally I believe that middle men are very important
element in a supply chain but the literature the discussions they are always against middle
man, they say that middle men takes away most of the profit of the supply chain and
particularly when we talk of food products, when we talk of Handicrafts, when we talk of
products coming from micro small enterprises.
In these cases we are highly anti to the middleman and therefore if a supply chain which is
very much dependent on middle man and the structure is changing and when organised retail
is coming when organised activity are happening so those middleman need to adapt
themselves to the changing market structures. Otherwise, they all will be out of the market
because most of the discussion is happening about how to eliminate the middle.
348
So this becomes a very good example that if you are adaptable to these changes you will
survive, if you are not adaptable to these changes we will not be able to survive in this
changing structure of the market. You need to modify your supply network strategies, we
have discussed the example of Dell in earlier classes and it is important to discuss the same
example in this class also.
That when Dell found that customers are not ready to wait for the product because the orders
are not very specific, almost similar type of orders were coming from different customers so,
Dell immediately change their supply network strategy, then it started the retail supply of the
product, earlier it was only through special orders, but after that Dell started retail distribution
of the product.
So Dell became adaptable to changing situation of the market, so you need to be adaptable to
the changing market structure, you need to be adaptable to the changing supply network and
all those things which we require that how you take changes positively into your supply
chain, so this is the first important aspect of flexibility in the supply chain. So we stop here in
the session, we will continue with the same topic and will see other elements of flexibility in
our next session. Thank you very much.
349
Lecture-30
Flexibility in Supply Chain
Welcome back friends, we were discussing in the last session about different types of
uncertainties and how these uncertainties can impact the network design decision and in that
uncertainty in the network design decisions we discuss that how flexibility in supply chain
can help us in handling those uncertainties in the network design decisions. Because these
network design decisions are long term decisions and huge amount of capital is also required
in putting a particular type of network.
You make warehouses, you make factories, you purchase trucks, so all these are the decisions
which you cannot change in the short term, so therefore you should be very careful in making
a network which can take care of uncertainties. We discussed about different types of
uncertainties in the last session that you can have uncertainty with respect to demand, you can
have uncertainty for prices, you have uncertainty of exchange rates, you have uncertainty of
different types of market structure.
And then the other type of uncertainty which is coming that is uncertainty of the business
models. Now a days different new types of business models are coming day by day and this
business models also create impact on supply chain network. Earlier we use to understand
business models very clearly that you are giving me a product I am offering you money in
lieu of that, but now days it is happening that Google is offering as a service of Gmail.
We all use WhatsApp, we all use Facebook and we do not pay anything to Google, we do not
pay anything to WhatsApp, we do not say anything to Facebook, but still we all are using
their service. Though there is a business model, there is a profit stream in the business of
Google, in the business of WhatsApp, in the business of Facebook, but this is stream is not so
visible to us.
Some of us may say that enticement are the major source of revenue for Facebook, some of
us will say that they have over data and they send the data to various product companies and
350
that is the source of revenue. So we all are guessing, we all are trying to speculate that how
do these companies make huge amount of profit, but these things suddenly point for us that
there are lot of uncertainties with respect to business models also. So you need to handle all
these uncertainties in your network decisions and we discussed that how flexibility is the
important answer for handling these uncertainties.
In our last session we discussed in details about the adaptability of the supply chain that is
one important dimension of flexibility, that as the market structure for changing as you are
modifying the supplies network strategies and your supply chain should adopt to these
changes. The analytics components where you are making decisions in real time, we have
seen during the discussions of forecasting that how adaptive forecasting can help us in
improving our decisions.
You are using the most recent data for better forecasting, so these methods are known as
adaptive forecasting methods. So these things are required for building flexibility in your
supply chain. You cannot have the old regression equation Y=A + Bx giving all the time the
values of Y with respect to X. Because as the time will move you need to change the values
of A and B, the point which I am trying to say that is you have one regression equation y = A
+ Bx.
351
In the demand forecasting where we know that X is some independent factor and we take the
value of x from such other sources we have calculated the values of A and B, these are the
constant values and these values are taken based on some past data of Y and X. Now for
future today I am in 2017 March and from some independent sources I get the value of x for
December 2017. I will put that value of x and corresponding value of Y I will get.
This is my conventional method of forecasting, but what I am trying to make you clear that in
adaptable forecasting I need to continuously update the values of A and B also. So I will have
a much better forecast for the values of December 2017 for X. Earlier it was not so, so we
used to have higher forecasting errors. Our forecasting may lead to higher stocks or under
stocks.
But nowadays we want adaptive forecasting where we can change the values of A and B in
real time and you can have a more realistic adaptive forecast, so that is the first important
element of flexibility in the supply chain. The second is alignment, what is the meaning of
alignment, we create incentive along the partners within the supply chain for better overall
performance.
Now in supply chain we all know that there are good number of partners, you have
wholesaler, retailer and smaller retailers. Now I am not taking the supply chain to the left side
of this wholesaler, manufacturer and suppliers are also there. I am considering right side of
the supply chain. Now if I am not creating some kind of incentive system that is uniformly
giving advantage to all the parties.
352
If this incentive system is biased or for this wholesaler, or for this retailer or for this retailer I
will not be able to achieve a good alignment, so the second important element of supply chain
flexibility is the alignment and this alignment is possible that I need to create incentive
system which should take care interest of my all the partners of the supply chain then I can
have a better overall performance.
Otherwise you can have plentiful of examples where most of the supply chains before this
wholesaler you have the manufacturer also and the incentive system is designed in such a
way that most of the profit, most of the share of the incentive goes to the manufacturer. And
therefore these wholesalers retailers are not able to take much of the benefit and this is a poor
alignment of the supply chain.
In your earlier session we were talking of middleman and we have large number of examples
in our country particularly in India where you have handicraft items, you have cottage
industries, you have small farmers. Now these small farmers are those who were involved in
the cottage industry, in the Handicraft activities they have limitation for accessing the
markets.
And the role of middlemen is very important to take their product from those far places where
handicraft items are made, where these vegetables food items are grown to bring it to the
urban market, but for that obviously they also want a share in the profit, the incentive system
design by these middleman is such a faulty system that it does not give appropriate amount of
incentive to the producer, to the villages, to the farmer.
And therefore we have so much debate about the role of middle man in the supply chain. If
you can design a proper incentive system where middleman also take a right amount of profit,
the right amount of profit goes to the retailer and right amount of profit goes to the actual
producer the farmer, the handicraft manufacturer, the cottage industry owner, then probably it
will help us in achieving the better overall performance, but this type of alignment between
the incentive system is not there.
And therefore you have a problem in almost all types of unorganised supply chains in our
country and these supply chains are not finding them self changing aligning with the
353
uncertainty of the external environment. So this alignment component is also very important
and alignment is possible when you can design appropriate incentive system. Unfortunately
the coordination, the trust between the partners that is not there.
And because that trust is not there we all want to maximize only our incentive, so each one of
us is looking for incentive of individual ,no one is looking for the incentive of this overall
supply chain system and that is another very important source of problem and in case of those
supply chains where you can clearly defined the ownership of an individual, when I take the
example like this is the supply chain of Walmart.
This is a supply chain of Maruti, this is the supply chain of Dell, this is a supply chain of
Hero Motocorp. In these types of examples it is quite possible to develop an incentive system
which can give good amount of equality in the incentive sharing, but when I talk of supply
chain of vegetables, when I talk supply chain of food items and fruits etc. here I am not able
to decide that who is the owner of the supply chain.
And in these cases there is a problem that we are not able to develop a good incentive system
along the partners of the supply chain and in that case the problem of alignment will be there,
but certainly where ownership is properly known supply chain of Apple, supply chain of
Samsung, in these cases you can create a good incentive system along the partners of the
supply chain. After adaptability and alignment the third important dimension of flexibility is
agility.
Agility is another very important aspect of supply chain flexibility. Now agility we mean that
how fast you can fulfill the demand of the customer, your response rate, that is the agility.
Now what agility means, the ability of a supply chain to respond to short term changes in
demand of supply quickly and handle external disruptions smoothly, that is the meaning of
agility. Now as we have discussed in the last session that supply chains required high speed
and low cost.
When I talk of agility this means high speed and high speed is easy to achieve when you have
steady conditions, you can follow things like make to stock, you have products readily
available in your warehouse, you have products which are readily available in your retail
counter and as soon as a customer comes you can offer product immediately to the customer,
354
so that is very high level of speed, customer comes and you offering immediately without
even a waiting time of 5 minutes.
But problem starts coming when you have short term changes in demand then problem start
coming, how to achieve the same level of speed, when demand is changed and when demand
is changing you will not be able to maintain good stock, make to stock type of policy in these
cases, because some time demand is more, sometime demand is less.
And some time you were over stock, sometime over under stock and both these things are not
desirable so therefore you require some kind of agility in your supply chain that you should
be able to fulfill the demand of the customer within the short time even if demand is
changing.
Now for that purpose there is a term we all know that is lean supply chains, we will discuss in
our coming classes in detail about lean supply chains, but lean supply chains are those types
of supply chain where we try to reduce the waste in the supply chain. We consider that those
who are lean and thin they are more agile, they can change very fast, their reflexes are very
fast and those who are bulky they are slow.
And same thing applies to the supply chain. If I am lean my supply chain is lean, the meaning
of that lean supply chain when is that I have less amount of inventory at each stages. I have of
minimum waste in my supply chain at each stage. So my supply chain can change very
355
quickly as per the demand of the customer. Otherwise what will happen for example I have a
product A, today I have a product A and at add retailer 2 I have 10 units of this product A.
Tomorrow because of some reason demand of this product A goes down and now customer
wants a new variant of this product A that is a A ’. Now I already have 10 units of A in my
stock, so I will be hesitant to procure A ’, because I will wait that I should sell my 10 units
first and then I should procure A dash, but if I procure A’ because customers are wanting A’
now, so I need to force to purchase A’.
So in that case these 10 units of A may turn to the overstock and maybe after sometime these
overstock metre into the dead stock, demand will totally be out and only A dash demand will
be there. Now there is another retailer who is having only 2 units of A, for that retailer it is
much easier to procure A’, so the meaning of lean is that you need to have a minimum
amount of inventory.
So that as soon as demand changes in the short term, you can immediately procure new
products and if you have a stocks of your previous products in your warehouse, in your retail
counter or wherever you keep your finished goods inventory, you will be hesitant to change
or to procure new product and by that time your competitors must have procured the new
products.
And then you will be behind in the competition, it is also very important to understand one
more phenomena because that will help us to understand the meaning of agility. Now what
356
happened when you are in this process of lean and agility, we also have a particular
phenomenon of double advantage, now what is the double advantage, here normally it is seen
that this is a product life cycle very simple representation of product life cycle where I am
introducing a product going to come growth period and then maturity period and finally it
will be declined.
So this is product life cycle of those who are responding to the changes at the early time and
another life cycle I am making for those who are responding to the product life cycle at a later
stage, so this curve A represents product life cycle of those marketers, those supply chain
decision makers who are able to introduce products to the market at early stage and the curve
B represents those who are introducing a later stage.
Now you yourself can see that for the early introducer those who are agile those who are able
to adapt to the changes early they are able to achieve higher market share, so this height, this
height indicates the market share, so they are able to go to the higher levels of market share
and B is at the lower level. So those who are coming late to the markets they will not be able
to achieve that higher level of market.
Second thing you see the span of A, the life of the product A is for longer duration and the
span for this company B is for the shorter duration. So this is the double advantage, you will
have higher market share and you will be there in the market for the longer duration, for B
you will not be able to achieve that level of market share and you will have a smaller span of
market period in this case.
So the concept of agility is very very important, obviously company A is more agile and
company B is less agile and the reason maybe because leanness of your organization. A
maybe more lean and therefore could respond quickly to the requirement of the market, B is
bulky, so B took time to change itself to the requirement of the market and therefore B came
late into the market with the same product and you can see finally the advantage associated
with the agility.
So this agile supply chains are very very important and we need to understand that what are
the different types of waste in my supply chain and with all possible effects I need to reduce,
I need to eliminate the waste and one important waste like I discussed about the inventory,
357
inventory is 1 type of waste in my supply chain. The other type of waste can be the waiting
time, if I do not have a good route planning.
My trucks are going for last mile delivery and at the last mile delivery trucks are waiting for
unloading, that is also a waste because I am wasting my time, so I need to eliminate that
waiting time also. So when we will go for lean supply chain class we will see that what are
the different types of waste.
And we need to eliminate those waste, so that we can achieve in higher degree of agility in
my supply chain and that high degree of agility in the supply chain will help me to achieve
the objective of flexibility in the supply chain. So and because I am not keeping much
inventory, I am not keeping more drugs, I am not keeping much of my infrastructure and
therefore I am very fast in responding to the short-term changes in the demand.
As well as I can also handle external disruptions very smoothly without changing much in my
system. If I have developed huge warehouses across the country and now you see the time of
online retailing is coming, so in that case my big stores will be of less use and it will be a
challenging task to sell those stores because no one else will be requiring those types of store.
So it is important that you need to identify that what type of resources you need to build and
what type of network you need to build, so that whenever question of agility comes you can
respond smoothly without changing much in your organisation. Otherwise you need to adopt
or you need to change as per the situation but many a time these changes create lot of
resistance, it is not very simple to adopt to new things.
So there are courses on management of change, but the concept of agility says that if you
have minimum resources, if you have optimum resources you can change smoothly without
creating much disruption in your organisation, so the ability is third important dimension of
your flexibility, so with adaptability, alignment and agility if we have these three things in
our organisation we are a flexible organisation.
And flexibility is one of the key for making our supply chain network suitable for handling
the uncertainty, it can minimise the impact of uncertainty in my network decision of supply
358
chain. So we stop here in this session and in our next session we will discuss some of the
mathematical aspects of network decision under uncertainty. Thank you very much.
359
Lecture-31
Optimal Level of Product Availability In Supply Chain
Welcome friends, we are discussing about various types of decisions in the supply chain and
in last few sessions we discussed about inventory management in the supply chain, along
with inventory management there is one more issue which we will like to discuss in today's
session that is about the level of product availability in a supply chain and it is very crucial to
discuss about the level of product availability.
Because this on one side can take your supply chain to the responsive supply chain, on the
other side it can take you to the efficient supply chain.
Now let us first understand what is the meaning of product availability?. The meaning of
product availability is that customer is going to a supply chain or customer is going to a
retailer and how many times the customer demand is satisfied from the available inventory. I
am a customer I am going to a shop let say in a week 100 times, so out of 100 times my
demand is fulfilled from the available inventory in the shop let us say 90 times.
Ten times I am not able to get the product which I am desiring, so I will sell the product
availability is 90%, sometime in literature we also called as it service level or the filtrate,
360
these are other terminologies which we used to define the product availability. Service level
is a very common term which we used to define the product availability. Now it is very easy
to understand that customer satisfaction is directly related with the product availability.
If we want higher level of customer satisfaction we need to increase the product availability.
Now increasing product availability is directly related with the stock keeping unit, you should
keep more number of units in your stock, so that you can offer higher product availability. So
when you are offering higher availability it means your supply chain is more responsive
whenever customer comes you are able to fulfill the requirement of the customer.
So that is a responsive supply chain, without any time delay you are fulfilling the customer
requirements. On the other hand if a customer is coming and you are not providing product
immediately and you ask customer to wait and you are looking for a pull type of supply chain
where you will ask your next number in the supply chain, next stage in the supply chain to
supply the product as per the customer requirement.
So obviously because time delay is there the satisfaction level of customer maybe slightly
less, so if you go for very high product availability, you go for a very high service level, you
incur more inventory cost, and if you go for very efficient supply chain where you see that I
should not stock and inventory and I should only ask for inventory whenever demand is there.
So in that case customer satisfaction level is low.
361
So we need to see that what should be the optimal product availability, so that we can
maintain, a balance between the customer satisfaction and the cost of keeping the product in
the stock, so that is the optimum level of product availability we are talking in the session.
Now the product availability, the optimum level of product availability depends on these two
important factors. The first is cost of overstocking the product and the second is cost of under
stocking the product.
Now what is the cost of overstocking and what is the cost of understocking, to explain these
things considered a very simple situation that I am stocking let us say I am at the top of the
some picnic spot and where tourists are coming and I am keeping some rain coats with me, at
that point I am keeping some rain coats with me because it is very likely that at any point rain
may occur there.
Now when I am keeping rain coats with me and these are I am paying some price for that,
cost price and let us say today I stocked 20 raincoats and only 10 customers came to my shop
because there was little rain, so I have stocked 10 additional raincoats which were not sold
today and because of very poor quality of rain coats the rain coats which I am keeping today
at the end of the day are not able to sell tomorrow.
These are use and throw type of rain coats. So these cost which I have incurred for those 10
additional raincoats which I could not sale today that cost will go for the cost of overstocking,
because the demand was of 10 raincoat today only and I stocked 20 raincoats, 10 additional
raincoats, so that will go for the cost of overstocking. Cost of understocking again I stocked
considering the yesterday's experience I stocked 10 raincoats today, but today because of
cloudy weather because of rainy weather around 25 customers came to my shop.
But I stocked only 10 rain coats, so I could sell only 10 raincoats and for remaining 15
customers I could not give the required products. So now the profit which I could have
earned from selling those additional 15 raincoats I was deprived of that profit and this
likelihood of the profit which I could have earned and I could not earn because I did not stock
enough number of rain coats this is going to be the cost of understocking.
So I want to have a balance between these two things, cost of overstocking when I am
stocking simply now we put in a generic form the cost of overstocking is cost when I am
362
stocking more than the demand, the additional cost which is there because of stocking more
number of units than the demand. Cost of understocking I am slightly pessimistic and
because of my pessimism I stocked less number of units and demand was more.
So I was deprived of that additional profit, so that is cost of understocking. So we need to see
that a balance has to be created between these two types of cost and for that purpose we will
see with the help of a numerical example that how do we model such a situation where we
need to decide about the optimum level of product availability.
For that purpose we have this type of demand data available with us where the demand of the
product in hundreds of the units are like 400, 500, 600, 700, 800, 900 and 1000 and on the
basis of our past experience we have given these levels of probability of these levels of
demand. Then the probability of demand is 400 units is 10%, probability of demand being
500 is 20% and so on that demand is 1000 the probability is 10%. Now I need to calculate out
of this demand table that what should be the optimum product level I should keep in my
stock.
And some of the additional data available to me that is the cost price is for this item for which
we are discussing is 45 dollars per unit. I am selling this product at the rate of 100 dollars, the
unsold items if there are certain unsold item because of overstocking the unsold items I am
able to sell at some exhibition at the rate of 50 per item, but in exhibition the cost of holding
the inventory and transportation to the exhibition side is 10 dollars per item.
363
Now there are different things which we will like to calculate. The first thing which we will
like to calculate that is the expected demand with the help of this data first thing is to
determine the expected data, the different levels of demand and their respective probabilities
are given to us. But the expected demand considering this whole situation is
so your expected demand becomes 4 x 0.1 + 5 x 0.2 + 6 x 0.3 + 7 x 0.1 + 8 x 0.1 + 9 x 0.1 +
10 x 0.1. Now I will like to stock any number of units as long as by stocking that additional
unit, my expected profit increases if I my expected profit does not increase I will not like to
stock additional unit.
So the underlying condition because this is my expected demand, so this expected demand
becomes 0.4 + 0.1 + 0.18 + 0.21 + 0.8 + 0.9 + 1. Now cost price is 45 dollar, the selling price
is 100 dollar, so profit per unit is from the given data 55 dollar per unit. Now I will multiply
this expected demand with this profit per unit this will give me my expected profit that with
this type of data this is my expected profit will be expected demand multiplied by 55.
Now the meaning of expect optimum level of the product availability is that I want to
increase this expected profit, as long as by stocking additional units my expected profit is
increasing, then certainly I will like to stock additional units, but if expected profit is not
increasing I will not like to stock additional units. We can show the meaning of that what is
364
the meaning of that additional unit with the help of another way of doing this calculation of
expected profit and probably then this calculation will be more clear to us.
Now in this table if we see the data the maximum probability is associated with the demand
level of 600 units, that is the 0.3, so just by seeing this data it looks that I should stock 600
units because that is the maximum probability. So I also take this assumption initially that I
should stock 600 units because this is going to maximize my profit, because this is I feel just
by seeing this data that maximum probability is associated with 0.3.
So obviously I should stock 600 units, now when I am stocking 600 units, so let us say what
will be the calculation of my this is the demand that is 4, 5, 6, 7, 8, 9 and 10. Now this is the
sell when I am stocking 6, so I will be able to sell all 4 units which are the demand, when 5 is
the demand I will be able to sell 5 units, when 6 is the demand I will be able to sell 6, but
when demand is 7, 8, 9 or 10 because I have stocked only 6 units.
So I cannot sale 7, 8, 9, 10 out of 6, so my sell will be 6, 6 ,6, 6 in all these cases whether
demand is more but since I have stock the less I will not be able to sell more number of untis,
so these are the sells data. Now the cost data, cost price is 45 dollars, let us slightly simple
cost price, cost prices is in this case is 1 dollar and selling price is 2 dollar, so it will make our
calculation slightly simpler.
So since cost price is 1 dollar and I have purchased 6 number of products, so my cost is 6, 6,
6, 6, 6, for all these levels my cost is 6. Now comes to my revenue I sold 4 newspapers, my
365
selling price is 2, so 4 x 2 revenue is 8, 5 x2 revenue is 10, 6 x 2 revenue is 12 and then for all
other periods I am selling only 6 newspapers, so it is 12, 12, 12 and 12 and therefore I
calculate my profit which is revenue – cost, 8 – 6 profit is 2, 10 – 6, profit is 4, 12 - 6 profit is
6 and then subsequently for rest of the rows profit is 6 rupees or 6 dollars or 6 units whatever
you say.
Now please see the probabilities these are probabilities 0.1, 0.2, 0.3 and then for remaining
four levels it is 0.1, 0.1, 0.1, 0.1. So now just to summarise the expected profit is the
multiplication of 2 x 0.1 the probability of happening of profit of 2 dollar is 10%, probability
of happening the profit of 4 rupees 4 dollar is 20% and so on these are the probabilities of
different levels of profit, so my total expected profit is summation of all these products that is
2 x 0.1 + 4 x 0.2 + 6 x 0.3 + 6 x 0.1 + 6 x 0.1 + 6 x 0.1 + 6 x 0.1.
Where I am stocking 600 number of products. Now somebody will argue that I should stock
7 number of units or 8 number of units, so I will take this decision whether two stock 7 or 8
or 9 only when I do this whole calculation again by considering the seven or considering the
8 and calculate this total expected profit.
If this total expected profit is higher by is stocking more number of units I will like to go with
that decision but because of cost of overstocking it is quite possible that my stocking more
number of units I go on negative direction, my total expected profit may decrease, so that is
the meaning of optimum level of product availability, that which level of queue I should
keep, so that my total expected profit is highest, is maximum.
So for that purpose we have already discussed about the cost of understocking and cost of
overstocking. This cost of overstocking and understocking help us in developing a model for
which we can use that optimum quantity which we should stock and for that purpose let me
take you to one more step in this table.
In the table we will add one more column and that column is of cumulative probability, and
that cumulative probability will be 0.1, 0.3, 0.6, 0.7, 0.8, 0.9 1. Now what is the meaning of
cumulative probability that let us say I take this data 0.7 or 70%, now this data. 0.7 says that
of the time demand is less than 7, now if you see the earlier quantity Q=6.
366
For that the cumulative probability is coming 0.6, now the meaning of this 0.6 is that 60% of
the time demand is less than 6, but 40% of the time demand can be more than 6 also, so
whenever demand is more than 6 ,I will not be able to fulfill that demand so that will go for
my cost of understocking but 60% of the time is less than 6 or up to 6, the quantity ordered
equal to 6 will help me to achieve my service level.
But now days we all know that 60% service level is a very low service level and in most of
the products which are becoming commodity this service level is absolutely very low, so we
need to see that how to increase that service level and for that purpose this concept of
cumulative probability is going to help us. Now in this case we now calculate a probability
that demand is more than Q, should be more than or equal to Cu / Co + Cu.
Where Cu stands for cost of under stocking and C o stand for cost of overstocking, so now
taking this data which is available to us let us take this date first of which is a simple data,
here cost price is 1 rupee and selling price is 2 rupee. Now if I have some unsold items, so on
each unsold item I will incur a loss of cost price, so my cost of overstocking in this case
becomes 1 dollar per unit.
If demand is more than what I am stocking and in that case I will not be able to incur the
profit which I can get, so in that case the loss is 2 – 1, the profit which I could have earned
but I was deprived of that profit, because I stocked less, so that is cost of understocking, so
that is selling price - cost price, so this is again one, so now you will like to use this data of
Cu and Co in this case so it becomes Cu 1 and as a matter of fact Co is also one
That is 1/2 = 0.5, now this calculation of 0.5 tells us that we will take a level of cumulative
probability which is just higher than 0.5 and I am coming from the top 0.1, 0.3 and
immediately higher than 0.5 is 0.6 and corresponding to 0.6 the demand is 6, so this is my
optimal level of product availability. I should maintain this level of product availability 6 in
my organisation.
If I increase the product availability beyond this point, so my expected profit may decrease.
This is the point where my expected profit and loss are imbalance, so this is the meaning of
this calculation. Now take this data in this slide, in this data the cost price is 45, the selling
price is 100.
367
So my cost of overstocking is simply 45 and cost of understocking the profit which I am not
able to get that is 100 – 45= 55 dollars, but there is some other information also given. That
information says that unsold items, those items which I am not able to sell can be sold in
exhibition at the rate of 50 dollars, but in the case of going to exhibition, so the holding and
transporting cost is 10 dollar per item.
So cost price is already 45 and in that cost price of 45 you add this cost of 10 dollars per item.
So now if you are selling some item in the exhibition the cost of that item becomes 55
dollars, cost of that item becomes 55 dollar and you are selling that item at the rate of 50, you
are selling at the rate of 50, so the only lost the cost of overstocking ,so the cost of
overstocking in that case because you have some opportunity to dispense of your unsold
items.
So your cost of overstocking reduces to just 5 dollar per unit because you are taking a margin
of 50 dollar, you are taking a revenue of 50 dollar on unsold items but 10 dollars is additional
because of transportation and inventory at the exhibition, so your loss is just 5 dollar and cost
of under stocking is 55, so if I go with this data so your new calculation will be cost of under
stocking is 55 and cost of overstocking is 5, so 55 + 5 that become 60 and if I do this
calculation so it is 11/12.
And this is very close to around 0.9 and with this data now I will like to is stock 9 units, so
now the point which you can understand after this calculation and with the data, now in the
first case when we took this data here our cost of overstocking is actually very high, cost of
overstocking is almost equal to cost of understocking, so therefore we become conservative.
Conservative means we will like to stock less because otherwise our losses will be very high,
in the second case when we use this data here the cost of overstocking is less as compared to
cost of understocking it is much less actually today, so when cost of understocking is much
less like this so we can take some chances and when we can take some chances the meaning
is that we can think of stalking higher quantities.
Because with higher quantities the loss is not much even if I go with this 10 number of items
10 units, so loss is 5 units per dollar per item and therefore I can take this type of risk in my
368
decision making, so the relative proportion of cost of stocking and cost of understocking will
decide my actual decision that what should be the optimal level, but this formula cost of
understocking /(cost of overstocking + cost of understocking).
This formula will help us to actually model the situations that what is the optimal level of
inventory we should maintain or what should be the optimal level of product availability we
should maintain in the supply chain at different stages so that we can have the maximum
expected profit. So I hope we all can use at different stages in the supply chain this decision
making pattern, so that we can give a decent level of customer service level and at the same
time we do not incur excessive cost of offering that service level. Thank you very much.
369
Lecture-32
Time Value of Money in Supply Chain
Welcome back friends, in last two sessions we were discussing about uncertainty in network
design decisions, uncertainties are of different types, we discussed that network decisions are
long term decisions, so you need to be very careful about developing those networks where
you are developing the factories, where you are developing the warehouses, where you are
developing the depos, and what are the sizes of these things.
And we discuss different types of uncertainty which may come because of demand, which
may come because of price, which may come because of fluctuations in the exchange rate
and because of business models also. We also discussed that flexibility can be one answer for
handling this types of uncertainty. If you have 3 A’s in your supply chain, if you are supply
chain is adaptable, it can align and it is agile.
So these 3 A’s we discussed in detail in our earlier two sessions and in this particular session
we will see that because the decisions related to network, these are long term decisions, so
you need to take extreme care with respect to financial aspects of these decisions, so in this
particular session we will see with the help of some numerical examples that how we take
care of those aspects.
370
All these things we have discussed in our earlier session about flexibility in the supply chain,
we discussed about adaptable alignment and agility in the supply chain, now coming to long
term financial decisions in the supply chain we will like to discuss this discounted cash flow
analysis, you are making a project a supply chain decision of making a warehouse at a
location X.
Now you are expecting some cash inflows with that decision, so you have a choice whether
you want to develop at X location or at Y location and how much cash flows you will get in
future 10 years from warehouse at location X, how much cash flow you will get from the
warehouse which you are making at location Y and then you can compare the inflows of cash
for X and Y, because the cost of developing the warehouse at two locations may also vary.
So you will have a net present cash flow from these two alternatives and which ever
alternative gives you better net cash flow that is obviously your choice. Similarly you have a
very common situation in this type of discussion, and whether to develop your own facility or
you can develop some kind of lease agreement or you can have some on the spot market
decisions, so for the same location X we have these multiple options.
I can develop my own warehouse, I can purchase the land, and develop a warehouse, I can
also take a decision that somebody is offering me a lease agreement for 5 years, so I can sign
that lease agreement and it is also possible that in some cases I may not have consistent use of
that facility for 5 years, so whenever there is a requirement I can go the spot market and can
see whichever option is available I can use that for my purpose.
371
So all these possibilities are there and when all these possibilities are there I need to evaluate
these possibilities and later we will also see that how can we build element of uncertainty into
these decisions, so that is what we are going to discuss in this particular session, so now
because supply chain decisions are in place for a long time so they should be evaluated as a
sequence of cash flows over that period.
So as I just discuss that I am planning for next 5 years, I am planning for next 10 years, so
what is the sequence of cash flows over that 5 year period, 10 year period that is going to
decide my supply chain decision with respective network design. Now discounted cash flow
DCF evaluates the present value of any stream of future cash flow and allows managers to
compare different cash flow in terms of their financial value.
The point is that when we are getting future cash flows today it is 2017 and in 2017 I am
getting some cash flow let us say 1000 rupees, in 2018 I may get 1200 rupees, in 2019 I may
get 1500 rupees.
So in these discounted cash flow I will calculate the present value of these future cash flows
and what is the total present value what is the total present value of all these future streams of
cash flow that I will evaluate and on the basis of that I will take a decision about my network
choices, it is very very important to understand that a dollar today is worth more than a dollar
tomorrow.
372
This is a very very important concept in our management and we all know about this concept
of time value of money. If today with 100 rupees I can purchase 1 kg of rice, so tomorrow
with same 100 rupees I may not be able to produce 1 kg of rice it may be 900 grams of rice
that I will be able to purchase by that 100 rupees, so the meaning is the money today is which
is there in your pocket that is more valuable than what you are going to get tomorrow.
So for that purpose it is quite possible that in absolute terms 1200 is more than 1000 rupees
and 1500 looks more than 1200 and 1000 rupees but it is quite possible I have not considered
the rate of discounting, but it is possible that the present value of 1200 rupees is just 900
rupees and present value of 1500 rupees maybe 925, so though in the absolute term 1200 and
1500 these values are, these figures are more than 1000, but when I am talking their present
value it is equal to 900 or 925 maybe.
If I consider suitable discounting factor, so I need to see what is the total present value of all
these future cash flows that is my analysis of discounted cash flow. Now for that purpose it is
very simple and those who have some knowledge of financial management and even in the
classes of project management also we discuss this time value of money in greater details,
here we are just touching these aspects.
Now discount factor is 1 /(1 + K) where K is the rate of return and on the basis of this 1 /(1 +
K) becomes discounting factor, so for example if I consider the value of k = 10%, so my
discounting factor becomes this discounting factor is 1 / (1 + 0.1) =0.1 and if I want to
calculate the present value of 1200 rupees = (1 /1.1) x 1200
373
so the present value of 1500 for 2018 = (1 /1.1) x 1500 then for 2017= (1 /1.1) x 1500 x(1
/1.1) and similarly if I have a future cash flow of 2020 of let say rupees 1700 and I want to
calculate the present value of this 1700 for 2017.
So this will be a step-by-step process (1 /1.1) x 1700 will give me the present value for 2019,
then this factor multiplied by again 1 /1.1 will give me the present value for 2018 and then
this value this factor multiplied by the 1 /1.1 will give me the present value of this 1700 for
2017, so if I want to generalize the formula for calculation of this present value is [1 /(1 +
K)]tCt
where Ct is the future value of a particular time period t, now I want to calculate the future
cash flow for all the periods, then formula will be
So what I am doing in 2017 if I want to calculate the cash flow of all future streams, so I will
add this 1000+900+925+1700(1/1.1)3 all these, so the generalized way of putting this formula
is this cash flow for the original period. The period of starting that is Co that is here and then
Ct’s are C1, C2, up to Ct is the stream for all the period 2018, 2019, 2020 etc. and sum total
of that from t=1 to T.
So this will become my net present value, the formula for calculation of net present value, so
this formula we will compare net present value of different supply chain design options and
option with the highest NPV will provide the greatest financial data, so that will become our
options, that will become the choice for us. So let us have some data and with the help of the
data we see that how can we use this formula for determining a supply chain decision.
374
Now in this example one numerical example we consider that how much space to lease in the
next 3 years, that is the question, the demand is 100,000 units and we require 1000 square
feet of space for every 1000 units of demand, that is the rate of demand which is there, the
revenues 1.22 dollar per unit of demand, now the alternatives are decision is to taken whether
to sign a 3 year lease or obtained warehouse on the spot market.
You can sign one MOU for 3 year lease or you can have as per the requirement go to the spot
market and whatever alternatives available you can select that alternative from the spot
market, so these are the 2 choices available to you, now the cost is that if you are signing a 3
year lease in that case the cost is 1 dollar per square feet and these spot market the cost is 1.2
dollar per square feet.
Always you will find that spot market rates are higher than the lease and the reason is also
very simple because in the lease for the suppler as well as for the customer it is a guaranteed
agreement for a particular period of time. In case of spot market it is not so today demand is
there tomorrow not be there and you are fulfilling the immediate requirement and therefore
the normal circumstances a spot market rates are more than the lease rates. Now let us see
how to handle such type of questions and the rate of return is 10%.
375
Now in this case for leasing warehouse space on the spot out market, expected annual profit,
when you talk of expected annual profit so 100000 is the annual demand, this value is the
annual demand, this is the revenue that is 1.22 dollar per unit, that is the revenue you have
and in the spot market you are paying this much price 1.20 for per square feet, so your
expected annual profit is 2000 rupees.
Now 2000 rupees you are getting in all 3 years, for all 3 years the profit is same 2000 2000
2000, so you have same cash flows C 0 in the first year, C1 in the second year and C2 in the
third year, so C0 2000. So the value of net present value the total cash flow in terms of present
value is 2000+(2000/1.1)+(2000/1.1 2 )=5,471 dollars.
This when you are not going for the lease agreement, you are doing on the spot arrangement
for your requirements. The other alternative is you can go for the lease arrangement and in
376
that is when you are going for the lease arrangement the same revenue 1.22 that figure we are
carrying from the value that is the revenue 1.22 per unit of demand, so same figure come
here that 1.22 x 100,000.
Now the cost is less, cost of lease agreement is 1 per square feet, so this expected it becomes
22000 and because of this expected annual profit 22000 you are going to have 22000 in each
of next 3 years so 22000 here, 22000 here and 22000 here. Same rate of return you will apply
and 1.1, 1.12 and this is the net present value when you are having the lease for 3 years
So the net present value is much higher of signing the lease is 54711 dollar higher, therefore
you can decide to sign the lease agreement and this is a very very important thing that you are
signing the lease because you are getting the higher NPV, but there are in this question we
have taken some very simple assumptions and these assumptions will automatically give you
idea that we should do for this kind of lease arrangements.
Because the cost of lease is less 1 dollar and in that case it is 1.2 dollar, so you are expected
annual profit is less. Now take a situation now we have understood how to go about the
calculation but actually the problem will happen in that case when you are going to have
different annual demand and in that case you will be stuck by different alternative.
So let us have that situation that in a particular situation because now we know how to
calculate the NPV, again for the same three period in this case we had the constant demand
for all 3 years, now we have a varying demand for 3 years, so we are trying to build some
element of uncertainty in this particular case there is no uncertainty in demand and cost, so
there is no need to think much, you can directly calculate and get the answer that leasing is a
better option. But in case when we have some kind of uncertainty so I am trying to build one
case of uncertainty where we have uncertainty in the demand of the product.
377
First year you have demand of 50000 unit, the second year the demand drop suddenly and
you have a demand of just 10000 units and next year you have again a very high demand of
40000 units, so these are the demand labels for three different years, now when you are going
to lease in that case the space is fixed or you can say the cost of space is fixed for all the 3
years, if I am going for a lease arrangement so the cost of space is fixed that is 40000 per
year.
I have to pay 40000 dollars per year, when I go for the spot marker in that case I will take a
space as per the requirement, whatever is my requirement accordingly I will select this space,
so I will not pay constant 40000 in case of the spot market, my demand is 50000, so I will
take slightly bigger space and maybe at that time I will pay 50000 as rent in case of spot
market, when demand decreases to 10000 I will pay 12000 as the rent.
And when my demand is 40000 in that case also I pay 45000 as the rent, the profit per unit
which are able to generate not exactly the profit revenue, the revenue which I am generating
let us say that is 1.5 dollar per unit, so I multiply the value by 1.5 x 50000, 1.5 x 10000 and
1.5 x 40000, the difference of 4th column and second column will give me the net cash flows
in lease. This is column number 1, this is column number 2, this is 3, this is 4, this is 5.
So in case of lease I am getting the differences from column 4 and column 2 so I write these
values here and then I can also calculate the net cash flows in a spot market, this is column 6
and these will be the values of 4 – 3, so here I calculate 4 – 3, and now considering the same
378
rate of return that is 10% I will calculate the present value of these column number 5 and
column number 6 and you can do the calculation.
And then you can see that it is quite possible that when you complete the calculation in that
case it is advantages for me to have the spot market, to go for the spot market and not to go
for the lease arrangement, because here you are constantly paying 40000 in respective of
whatever level of demand is there, in a spot market I am able to adjust this though I am
paying slightly higher cost here the space cost me one dollar per square feet.
Here the space cost may be 1.2 dollar per square feet and that may also varying period to
period, initially it can be 1.2 and in the third year it can be 1.4 dollar per square feet, but
because I am able to optimise my space requirement in the spot market which I am not able to
do in the lease market, so on the basis of this net present value calculation of my all future
streams of cash flow I can take a decision that yes it is more profitable to have spot market.
So this is one small element of uncertainty with respect to demand that we built into this case
and we saw that it is more profitable to have a decision which is going to help us in creating
the more cash flow in the present system. Now going further into these things you will further
say that we need to build more uncertainties into our demand and cost related issue, so going
further we want to have the binomial representation of uncertainty and we can have some
other representation of uncertainties also.
You have a price P today and tomorrow either price can improve, improve means increase or
decrease, there are two probabilities with probability P the price can increase and with
probability 1 – P price can decrease, so in our next session we will see that how to build those
types of uncertainty in our network design decisions and that will give us more practical
insights about handling uncertainties in our real time analysis, So we stop here in the session
today. Thank you very much.
379
Lecture-33
Different Types of Analytics In Supply Chain
Welcome back friends, we are discussing the role of analytics in supply chain analytics in
supply chain title of this course. Now let us discuss that what are the different types of
analytical approaches is which are possible and what type of decisions is different analytical
approaches will help us in case of supply chain.
Now you can divide the analytics into these three types of analytics descriptive analytics,
predictive analytics and prescriptive analytics. In this session today we will discuss what are
the different characteristics of these three types of analytics and in our last many sessions
whatever we have discussed, so what can be classified under descriptive analytics, what can
be classified under predictive analytics and what is prescriptive analytics.
So now this is something which will help us in understanding the use of various software
tools and in which particular segment of decision making which type of data is required what
type of report is required and what type of problem solving approach is required.
380
Now the descriptive analytics that is the first type of analytics we know, so now the one line
answer for descriptive analytics is to know that what has happened that is descriptive
analysis, you are describing a particular situation, so what has happened the story about that
situation is your descriptive analytics, so narrating a particular environment is descriptive
analytics.
So this is the first type of analytics that collecting the information and telling that what has
happened, now in this particular case you gather data, you summarise the data and visualise
large data sets to reveal new business insights, because what has happened on the basis of
that you are trying to get new business insights, so this is one type of analytics, so the most
important thing is gathering and summarising and on the basis of that summary of large data
sets you try to get some kind of new business insights.
381
The other type of analytics is predictive analytics, what is likely to happen to predict about
the future, so one things is what has happened, the other thing what is going to happen. In this
case the descriptive data using historical data that is the descriptive data to predict future
outcomes that what can happen in the future. If you remember that in our forecasting
discussions we have done lot many cases of time series analysis.
Where we have historical data and using the historical data that demand was this much in a
particular period the demand is this much in a particular period and so on we are predicting
the future outcomes, we are extrapolating that what has happened in the past similar things
will happen in future also and that is very very important that you can predict the future on
the basis of your historical data. So this predictive analytics will tell you that what is going to
happen in the future.
And it can use either statistical learning which we have done when we have solve the
questions of forecasting, when we have done the inventory analysis, so all these things are the
statistical way of handling the predictive analytics or you can also use machine learning
where you go with the help of lot of algorithm, developed in the computer science and with
the help of those algorithm you try to predict the future. So that is the second type of
analytics.
382
The third type of analytics is the prescriptive analytics, one thing is there that is about past
that is your descriptive analytics, on the basis of this past what can happen in future that is
your predictive analytics, but what should happen, one thing is this is a natural course of
action if this past the same trend is going in the future, so you will go to this. But what should
happen we always look for a higher performance better performance.
So how goes those performance can take place that is what should happen that is the
prescriptive analytics that if I am let us say having in my supply chain the defects per million
opportunities that is one parameter DPMO defects per million opportunities. The current
level of DPMO is my supply chain is somewhere around 2000 so I will say that same level
will continue in coming periods also.
383
So my predictive analytics DPMO is 2000, descriptive analytics told me, so predictive
analysis say that I am not doing anything special, so in the future also DPMO will remain
around 2000, some natural variations will take place but it will remain around 2000, but the
prescriptive analytics because it is using operation research, it is using how you can improve
your system ,what should be the ideal sign, what is the optimise condition of the system.
So if I go with the prescriptive analytics it will say that you should achieve the DPMO
somewhere near 5, you are having so many defects in your million opportunity why don't you
go to the Six Sigma level that is the optimise level of defects in million opportunity, so the
use of prescriptive analytics will help us to get the best outcomes, so these are three levels of
analytics, the easiest the most common form of analytics which we all see that is descriptive.
Just giving a brief detail or as per the situation the complete narration about the something
which has happened, then the predictive analytics since in every election we feel a condition
of anti incumbency, so in this election also anti incumbency will be there and therefore there
are chances that ruling government will be changed by the opposition, but if I see from the
ruling government side how I can change this phenomena.
How can I improve my performance, so that the issue of anti incumbency will not be there so
that is the prescriptive analytics, so these are three forms of our analytics descriptive,
predictive and prescriptive and we have also understood that in which particular situation
means what are the different characteristics of these three types of analytics approaches.
384
In understanding this 3 analytics from this table is a very useful table for us because it tells
you that what are the needs of the user and what user needs to know and how analytics can
answer these requirements of the user and then what leaders do to make these things possible
so on the basis of these 4 things we will see the comparative discussion about descriptive
predictive and prescriptive type of analytical approaches.
We have taken this from a source which is our the group and they are one of the pioneers in
using analytics for the supply chain, so this particularly we have adopted from this source.
Now what the users needs to know , to do, users want to increase cross-channel fulfilment
and shipment, that is what users need to do, users need to reduce the transportation cost, the
inventory cost, the order fulfilment cost. These things user needs to do.
Then user wants to predict was to predict about infrastructure failure, user wants to know
about the forecasting related to space and facility demands , users want to predict it cost and
profit by customer product transportation fulfilment, inventory stream forecasting related to
profit and margins, users also wants to predict capacities by customer, product transport,
fulfilment, inventory stream and forecast capacities.
So users want to predict these things, and user want some kind of optimised solution to
increase asset utilisation, optimise scheduling of the resources, remain agile and competitive,
increase asset and inventory utilisation optimise efficiency and supply demand in face with
the ships, so all these things you need to do irrespective of whether this is to be done by
descriptive, or to be done by predictive or by prescriptive.
Now you can also to understand this all these things can be plotted in a sequence also and
then we can classify out of that long list that yes these 2 things can belong to descriptive
analytics. These two things belong to predictive analytics and this where you want some kind
of solution where you want to have some kind of solution related to optimisation these things
belong to prescriptive analytics.
So on the basis of that we have already categorised that what the user needs to do, these are
the things which form our descriptive analytics, these are the issues where we want to predict
something obviously these things are the part of your predictive analytics and for those things
385
where we want some kind of optimise solution in my supply chain, these are the part of
prescriptive analytics.
Now what the user needs to do with respect to these things, now you can follow these
columns top to bottom, so with respect to these things which needs to do you, user needs to
know the number and types of assets for here ,which has happened why transportation
logistics cost are high, what are the reasons that my transportation and logistics cost are high.
Now I need to collect data if I am talking on this particular aspect that why my transportation
and logistics cost are high.
May be I need to calculate the waiting time if my truck is moving from Delhi to Mumbai,
now in movement from Delhi to Mumbai how much waiting time my truck is incurring
because their maybe traffic jams, there may be some kind of toll system, there may be some
kind of other accident, some kind of prepared incidences also and all those things make count
for 1.5 days of waiting time.
The truck took from Delhi to Mumbai maybe around 7 days and in that 7 days 1.5 days are
the waiting time, so now I need to know these things if I want to see that why transportation
and logistic cost are high, so if I have this data with me that 1.5 days are wasted in waiting
only, so how can I minimise this waiting time and that answer will come in the prescriptive
analytics, but first I need to know that what has happened.
386
So I came to know that on an average my truck are wasting somewhere between 1.5 days to 2
days if they are going from Delhi to Mumbai in waiting only, so that is a non value adding
activity at all these non value adding things are increasing the cost of my operations and
therefore we will see that we need to eliminate these kind of non value added things from my
system.
So this is one example but there can be many items in my dashboard which can be attributed
for higher transportation and logistic cost, maybe my truck can fail and because of that my
maintenance issues may be highlighted because I am not doing right kind of preventive
measures and therefore more breakdown maintenance is are taking place and as a result of
breakdown maintenance I am in curing loss of due dates and I have to pay penalty for that
purpose.
And that is also higher transportation and logistic cost. So all the items what all item I can
think of whatever the point of generation of those data information that I need to list and then
only new insights can be available from the same data, it is very very interesting that same
data is there but some of us may be able to find new insights from the data and some of us
may not be able to see something interesting from the data.
So it is important that descriptive analytics where it looks very simple that it is simply the
presentation of past data but at the same time if we can visualise that last data set properly
certainly new business insights may come from the same data and that is what we are going
to see. We also want to know the value of multi party inventory, we have inventory not only
from one source in a supply chain we have inventory from variety of sources and we need to
know that what is the value of inventory from different different sources.
And therefore we can apply concepts like ABC analysis which will come to prescriptive
analytics. So you need to know the value of multi party inventory and this list is not
exhaustive by the way you can add many more parameters many more items which are very
important from your point of view like in this case I have not added one particular factor that
how many times you have given late supply to your customer.
That can also be a very interesting parameter now days because delivery speed is becoming
day by day a very important criteria, so in last one year that 30% of the times I was defaulter
387
I delayed by more than 2 days, 10% of the time i delayed my supply by more than 4 days, so
this type of data is also required in some cases, in some cases not be required. So how many
times you defaulted on the schedule delivery dates that is also important.
Then how analytics gets the answer for all these things you need to see first standard
reporting system, what has happened, simply the presentation of data that these are the
waiting times, these are the defects per million opportunities, these are the lead times and so
on whatever is there in your dashboard, so you need to give a fair description of all the data
points that is standard reporting.
So earlier in organisations we used to have a very detailed logbook type of system and in that
logbook we used to enter so many information and this information was circulated to top
management of the organisation so that was the system of reporting. Then MIS came
automation started and as a result of automation we started using computers for giving quick
information to all the persons to be reported in that particular report.
So it created it enabled faster reporting of the descriptive analytics, then you can also prepare
query and drill down reports. So this query and drill down report will help you to identify the
exact location where is the problem. In my supply chain from initial vendor to the customer
there are many stages from initial vendor to final customer there are many stages, you have a
manufacturer, you have a warehouse, you have a wholesaler you have retailer and then a
customer.
So there are many stages and to exactly the prescription you need to know where is the
problem exact location of problem, so it is very similar to Medical Science when you have
the exact location of the problem then only you can provide proper prescription. Otherwise
you will be doing a lot of hit and trial and it will take lot of time to get the location of the
problem and to prescribe it properly.
So query drilldown type of reports will help us in locating the problem where it is and then
sometime not always some Ad Hoc reporting is also possible intermediate reporting you can
say these are also required in some emergency kind of situation that how many how often
where some very specific information in a limited format, these type of things are also given
in emergency situation.
388
For that purpose if you consider the case of humanitarian supply chain, disaster supply chain,
in these types of supply chains Ad Hoc reporting may be required and in this ad hoc reporting
you give very specific information about this kind of data that how many units are required,
how often, where what is the frequency where you need that type of medicine where you
need that type blanket or other food items or other kind of help etc.
So that type of Ad Hoc reporting is also part of your descriptive analytics. Then what leaders
do to make this possible as I was discussing we need to see what are the alert items, you need
to create some kind of signals in your supply chain which can give you some kind of
alertness. So whenever something you may be knowing because we have already discuss the
inventory systems, when we discuss that basically EOQ model. In that basic EOQ model you
remember that we have a point of reorder ROP. So nowadays we have a system of canman as
soon as you reach to that reorder point all of a sudden that green canman comes into picture.
And it signals that now material is required, so that is a kind of alerts signal whenever you are
driving a scooter and petrol is below a particular mark you get an alert that your metre start
coming into the red side, so that you know that it is time to replenish the fuel and these type
of things are required some regular reports are required you need to have some kind of
dashboard items, you need to have a screen in your office and that screen will have various
dashboard items.
And using those dashboard items you can continuously monitors the progress of year supply
chain and business intelligence can be used for that purpose for creating the dashboard items
for generating the reports on the basis of your past happening. So that is all about descriptive
analytics, but all these are very mechanical activities mechanical means these are very routine
type of activity, but now my exposure my experience will help me to find something new,
something interesting from these reports.
We know the same test is being conducted by different doctors, but the specialist doctor the
doctor who has better exposure she can find the exact source of problem, exact location of
problem with the same amount of information, with same ultrasound, with same x-ray, with
same scanning and a doctor with less exposure may not be able to find the problem, may not
be able to understand the location of the problem etc.
389
So therefore these reports can help a manager with wide exposure to develop better insights
or who understands the supply chain, who understand that market, who understand that
product in a better way that a manager who is not so expert, that is about the descriptive
analytics, coming to the predictive analytics, in this case we want information for future,
based on this descriptive activities we want to predict for the future and it is more like a
forecasting activity.
So you want to know anticipate sales shipments for specific channels assets facilities that
how many products I will send through rail, how many products I will send through road,
how many product I will send through air etc, so using different specific channels when to
consolidate underutilized facilities to expand the uses of them, I want to know these things, I
have so many facilities in my supply chain.
Some facilities are warehouses, some facilities are factories, some facilities are my
distributing points. Maybe all the facilities are not properly utilise, so I also want to know
when to consolidate some of these facilities which are underutilized, so that I can have a
better use of those facilities, how to determine and segment cost to improve service level. I
want to improve the service level of my customers, my down supply chain.
And how to segment cost for that purpose that is another thing which I want to determine and
that then how to determine and rank baseline and future scenarios to improve service level.
These things are also part of my predictions for the future and for that purpose for knowing
these things for predicting these things for the future the analytics will help with the
predictive model that is the most commonly used word and time and again we have use the
predictive modelling.
That what will happen next and then we do the trend forecasting in the same trend continues
what is going to happen, we will do stimulation, again based on the past data you have some
observations of the past and basis of those observations we will do the simulations exercises
and with the help of that leaders do to make it possible that we create different types of
predictive models, better algorithms.
390
So that we are more accurate about the future, we try to mix different types of trends, and
different types of you can say characteristics of our historical data to be more closer to the
future, and all these things will help us continuously lot of mathematical researchers are
involved to develop the methods models of predictive modelling for determining the issues
which we want to know which we want to do under the predictive modelling.
And then coming to prescriptive analysis where we want to optimise the use of resources and
for that purpose we want to increase the asset utilisation, we want to increase the our
customer service level, we want to reduce the wastage and for all that purpose we are using
optimisation, the operation research tools are the best solution for the prescriptive analytics,
you do the optimisation what is the best possible outcome.
Earlier in the classroom discussion we used to do optimisation in a paper, but nowadays very
good softwares are available and because of availability of good enablers the optimisation or
you can say use of prescriptive analytics in supply chain is continuously increasing. The
optimisation and then for the purpose of what leaders need to do we need to see that what are
the rules of the business.
When we are doing optimisation we take lot of assumptions, we try to simplify the real world
so that a particular mathematical model can we develop, but the knowledge of real world
game business and process rules will help us to organize our models closely to the
requirement and then the mathematics the prescriptive analytics will give us better results
which are more responsive which can help us achieving both these things simultaneously,
agility which is responsiveness how fast you are responding to the customers requirement.
And efficiency in minimum possible cost because these are the 2 important objective of the
supply chain, the whole discussion is revolving around this 2 words. Our ability to respond to
the customer requirements to clean and then how much lowest cost you can fulfill the
requirement of the customer, so both these things can be simultaneously achieved if we use
proper optimisation model with more real-world conditions.
So in this way now we are clear about descriptive analytics, predictive analytics and
prescriptive analytics, with this we want to close this session and in our next sessions we will
391
go that how predictive analytics in a supply chain can be used for the forecasting purpose.
Thank you very much.
392
Prof Dr Rajat Agrawal
Lecture-34
Predictive Modelling in Forecasting In Supply Chain
Welcome back, we were discussing in the last session about the role of analytics in supply
chain and we discussed that there are three types of different analytics, one is descriptive
another is predictive and the next is prescriptive. So we discuss that in descriptive whatever
has happened describing connecting the data, presenting the data in a summary for, so that
you can get some useful insights about the previous events, the past events.
Then following the same trend you assume that something may happen in the future also,
same thing will carry to the future also and what will happen to predict about that, that is the
predictive modelling, since forecasting as we have discussed in our earlier sessions also is the
basis of entire supply chain and therefore in todays session we will see that how predictive
modelling can be used for the forecasting in the supply chain. The forecasting plays very
important role because there are different types of entities in the supply chain.
We have vendors then the manufacturers are there, then we have the wholesaler or
distributors, then we have retailers, so in a general type of supply chain we have all these
different partners, the vendors, the manufacturers, the distributors and retailers. Now earlier
393
what was happening that all of them vendors, manufacturers, distributors and retailers, they
were doing forecasting individually.
And as a result of that individual forecasting they all were having because it is a prediction
and prediction maybe very close to the actual things and maybe because of your inability to
predict properly there may be large amount of forecasting errors, so to some extend the
forecasting errors are okay, but if these forecasting errors is start increasing, then problems
start coming.
Because the forecasting errors can lead to 2 types of situations, these forecasting errors may
lead to over stocking or under stocking, when you are having under stocking where you are
very careful, and it is in a nature of a person that I want to play without much risk, I want to
play a very safe and in that situation these are the chances that I where under stock.
And if I under stock my service level may go down, the customers coming to me I may not be
able to fulfill their requirements many a times because I will be out of stock, that is the under
stocking situation and when customers come to me and repeatedly I am not able to fulfill
their requirements what will happen, they may not come to me in future and they may go to
some competitor, so that is a loss of business.
In case of overstocking when I am stocking more then the anticipated demand, so what will
happen that I may end up with some extra inventory at my level may be at retailer, maybe at
distributor, maybe at manufacturer, maybe at vendor. At all label there will be some extra
inventory and time and again we have discussed that extra inventory may kill, may take away
the entire profit of your supply chain.
So therefore that is also not desirable that you create unnecessary inventory in your supply
chain, so both these errors, both these sides of the error whether you have access inventory or
you have a situation of out of stock, both are undesirable, and both will ultimately result into
loss of business, here customer may go to other supply chains, customers may go to other
retailers, if repeatedly a customer to me at my shop.
And I am not able to fulfill the requirement of the customer after two or three attempts
customer will not come to me and that customer will go to some other competitor. So I need
394
to see whether I am not under stocking too much, so we see that when we are under stocking
the customer satisfaction level may go down and when we lose a good customer to our
competitor, when we are overstocking then we are keeping more than the required inventory.
And in this situation another typical problem in a supply chain may emerge and that we all
know is the bullwhip effect, this is another important problem and you will find large amount
of literature is available to handle this Bullwhip effect, now because all these people they do
not want to go for the situation of under stocking, so what they want to do they want to go for
higher stocks.
Now everybody is going for over stocking and therefore as we move from the customer side
this retailer side to this vendor side the amount of excess inventory increases and you can find
this kind of increasing trend in stocking the inventory and therefore these are the excess
inventory that we are keeping and this excess inventory is another problem in my supply
chain, because inventory is a non value added thing.
And this inventory will take away my profits, so I do not want this extra inventory also. So
therefore we will look for predictive modelling where we will see that how we can improve
our forecasting in real time and predictive modelling as we have already discussed that what
will happen in the future.
So predictive analysis used historical data and various algorithms also to intelligently predict
the outcomes of various what if type of scenario, if this happens what will be the outcome, so
395
if you remember your B-Tech classes of first year, second year level we discussed this type
of what if scenario in Computer Science subjects, so that type of what if scenarios we
develop in different types of algorithms.
And the level of predictive analysis looks at all possible situations within the demand
planning, cost, profit, inventory, optimisation, logistics and transportation and wherever you
have the source of data. This predictive analysis allows organization to better anticipate
anomalies through their supply chain network and respond appropriately to the given past
conditions, so you understand the past condition on one side.
And at the same time you try to anticipate since it is all about future, so anticipation is there
but anticipation avoiding all possible anomalies in the supply chain and to name a few days to
give the example that businesses are able to determine shipment, consolidate, underutilized
capacities and rank various future looking scenarios, so you can rank your different scenarios,
whatever possibilities are there, you will rank them that this is possibility 1, this is possibility
2, this is possible 3.
And with the help of that we will try to see that what best solution is possible, to make
predictive analysis reliable additional external data may be required, you see more data about
your historical events are available better the predictive modelling will be, so with this idea
we will see that how predictive modelling can be used for making the supply chain planning
more effective.
396
Now let us see how predictive modelling works?, now predictive modelling are known
reasons to develop a model that can use to predict values for different or new data, modelling
provides results in the form of predictions that is the forecast, that represent a probability of
the target variable based on estimated significance from a set of input variables. Our ability to
get reliable input is one important guarantee for developing the reliable forecast for the
purpose of predictive modelling.
Then why we use predictive modelling in the supply chain, so this slide gives us that idea that
why are we using predictive modelling, through predictive analytics we are able to get some
results which we were doing earlier also, but because of development in the field of IT, in the
field of computer science and our ability to get data from different point sources with the help
of Internet.
All these things are major enablers for using predictive analytics more efficiently and
effectively nowadays in the supply chain environment, so therefore more and more
organisations are turning to predictive analytics to increase their bottom line and the
competitive advantage. And some of the reasons which I just told that faster cheaper
computers are available, easy to use software are available.
Then growing volumes and types of data, more interest in using data to produce valuable
insights, all these things because it is the time of data, the whole business is above data and
therefore we see nowadays lot of new business model are coming with the help of business
397
analytics only and the role of business analytics therefore in supply chain also becomes very
very important.
And obviously tougher economic conditions and a need for competitive differentiation, we
are finding that competition is increasing day by day and therefore we need something which
can differentiate us from my competitors and all these things are forcing me to use predictive
modelling in my supply chain.
Now these are some of the areas where predictive modelling is used, now to detect fraud, to
reduce risk, to improve the operations and to optimise my marketing campaigns. These are
some of the areas where you can find the use of predictive modelling, detecting fraud and
reducing risk. These are two important things in a global supply chain. In a global supply
chain you see that we have international supply chain and there are possibilities of risk
related to exchange it.
There are possibilities of risk related to International Political situation also, then there are
possible frauds also where you have a situation that you are getting products from a less
reliable vendors and he may not supply you the right kind of quality which you have
approved during the sample stage or you supply a product to downstream supply chain to a
less known customer and you have supplied and you are not receiving the payments.
So these are possible fraud and risk and predictive modelling can help us in detecting the
fraud because where is the location of problem, that is one important things which predictive
398
modelling can help us, it can help us in reducing the risk of exchange rate, it can help in
reducing the risk of uncertainties, lot of uncertainties are there in the supply chain, so we
have discuss them time and again.
So those type of risk can be reduced by using the predictive analytics, then obviously you
have more accurate information about the future and these accurate information will help us
in improving the operation and we can optimise according to demand, according to
consolidation of our various unused facilities you can optimise your marketing campaigns
also.
So detecting fraud as we just discuss these are combining multiple analytic method which can
improve pattern of detection and prevent criminal behaviour, so in international supply chains
where you are dealing with those unknown customers and vendors you are possibly have a
situation of fraud and with this predictive modelling you can detect with the help of multiple
analytical methods that where is the possibility of fraud.
then risk then we are using things like credit scores for assessing the buyers likelihood of
default for purchase and these are some of the very good possible you can say tool of
predictive analytics where even for the individuals you have this credit scores and our ability
to repay the loan is checked on the basis of the credit risk, credit score, so these are examples
of the predictive analytics based on my historical information.
399
These credit score are given, and on the basis of that you know that this customer, this
applicant is a genuine applicant or what is the level of risk involved with the particular
customer, so that is also a very popular many of us have experience the use of this type of
tool in our life.
Then optimising marketing campaigns that is also a very interesting use of predictive
analytics where predictive analytics are used to determine customer response or purchase as
well as promote cross-sell opportunities, but a customer who is looking for camera, so you
can have the cross-sell an opportunity to sell the memory card to that person also, you can
have the cross-sell opportunity to sell the bag of the camera also, so these type of things
consolidate or optimise my marketing campaign. I need not to start a new marketing
campaign for selling the bags of the camera.
I can optimise my lot of marketing efforts with proper integration of my sale opportunity, so
that is also very very important, that nowadays we know that if I am booking a hotel, I am in
Roorkee and I am booking a hotel in Bangalore, so it means when I am booking a hotel in
Bangalore I am going to Bangalore and then those companies which are offering the travel
booking system.
They can optimise their marketing campaign if they are into this predictive analytics that if
someone who is going to Bangalore from Delhi or from Roorkee, so he will also need to have
some kind of travel requirement, so they may start me start sending me the quotation, they
400
will send me the offers available to travel from Delhi to Bangalore and maybe MakeMyTrip,
yatra.com and all these types of travel agencies will be behind me.
So that is the optimisation of marketing campaigns. Next is improving operations, many

companies use predictive model to forecast inventory and manager resources. For examples
Airlines use predictive analytics to set ticket prices, hotels try to predict the number of guest
for any given night to maximize occupancy and increase revenue, you know that there is a
seasonality in India about the travel during the festival period.
During the period of summer and winter break for your children and with the help of
predictive modelling Airlines predict analytics to set ticket price during those festival offers,
so that they can maximize their revenue, similarly hotels they know that one Christmas or
New Year eve or in a routine way maybe on the weekends they may have more number of
guest in the hotel and how to attract and if it is a lean period when you know that not many
guests are going to come.
So accordingly you can reduce the prices to increase the occupancy and increase revenue, so
you will have improved operations on the basis of this predictive analytics where you can
find better utilisation of the resources of the organisation for getting higher level of revenue.
Now what do you need to get started the predictive analytics in your organisation, the first
thing you need to get started using predictive analytics is a problem to solve. It is very
simple, but it is simple to say when we are using predictive analytics it is important that I
401
should be able to write the problem, the meaning of writing the problem is that when I am
writing a problem on the board.
When I am writing a problem on my diary, on computer I get more perspectives, I get more
insights about the problem. So clarity about the problem is the first important thing to start
the predictive analytics, without clarity it is difficult to start the predictive analytics for your
supply chain purpose, so what do you want to know about the future based on the past, what
is the things for which you want to link past for the future, that is one thing.
Like you talk of independence in 1947 India got independence, but at that point of time
predictive analytics will talk of what will be the future of India, but you cannot link that
future of India with past events because past was ruled by the Britishers and the future was
going to be entirely new set of circumstances which was I think at all was linked with the past
and therefore you should know that what do you want to know about the future based on the
past.
So at that time if I you want to know about the governance of the future based on the past that
was a wrong problem to solve, but now since is most of the things at stabilized with respect
to our governance, so many things are possible that which we can determine on the basis of
the past that what will be the future, what do you want to understand and predict you will also
want to consider what will be done with the predictions, if I know these things what
managerial implications will be there, so all these things are some of the examples to clarify
that what is the meaning of problem to solve.
402
The next step is you will need data once you are clear about your problem the next step is
data you need data for solving the problem, now data is to be found out from variety of
sources, data means for a lot of places that maybe your point of sales also, that maybe the
policies is of government also, that maybe the moves of your competitors also, so there are
large number of places from where you can get the data.
So like transactional systems data collected by censor, third party information, call centre
notes, weblogs etc. All these are the different places, different sources from where data has to
be captured and it is very very important in our predictive analytics that we should have a
good idea that from where the data can come and we need to list out all possible sources of
data. Earlier in our manual discussions we were only limited to this transactional systems.
The customers coming to my shop and I am giving a product that customer that is a kind of
transaction we used to have, so that was the only source of data earlier, but nowadays because
it is possible to collect data from variety of sources, so we need data from all these different
sources, you will need a data wrangler or someone with data management experience to help
you clean and prepare data for analysis purpose.
Now the data which comes maybe it will have variety of information and all that information
may not be required for the type of problem I am looking to solve, so you need to do cleaning
of the data, the unwanted data you need to remove from your total set of data and that is the
second important thing in this process of data and data preparation. After that once you have
403
data also depending upon the type of data you have to build a predictive model that is the
third important step in this process.
That model building is the important part and here comes the role of you can say the cross
functional people also because this work collecting data they are the field persons mostly
marketing people, those who understand the problem they may be marketing or operational
people, but those who are developing the model they may be from the IT and computer
science and the mathematical background also.
So they are the different people those who developed the model and nowadays we are using
easy to use software with people can use easily and you want some kind of readymade
software for that purpose and though I am talking of readymade type of software but initially
you still need someone who is expert in the field of data analyst and so that he or she can
refine the model as per your this is what we call as customised the model as per your
requirement and employ that model in the system of your organisation.
404
Then another important thing with respect to predictive modeling is that its a team approach
as I just mention that someone is there to understand the problem, then there are another
group of people who are expert of understanding that for this problem what are the different
set of data from where data has to be captured and then they also need to have some kind of
exposure.
If I say that you go to villages and collect data about their preferences for a particular
product, now someone who has never gone to village, someone who has no connectivity with
the villages will not be able to collect data from the villages, so we require people from that
background also in our team who can help us to facilitate the data connection from the
villagers, so same thing is here that it is a team approach and we require people from different
background.
So that they can help us in doing the right kind of data analysis. There is a very interesting
story which we discussed during operation research classes that there was a problem in 1
multi storey building, only one elevator was there and it was taking lot of time for all the
users of the elevator to wait and they were giving complaints to the owner of the building that
please install few more elevator, so that our waiting time can reduce.
So he thought that it is a queuing problem and as a result he invited one ORT operation
research team to study that how many elevator should be there in that building, in ORT also
we follow this team approach and under this team approach one psychologist was also there
in that team and he was also meeting with the people he was also collecting the data and
405
when it came about solving the problem that how many elevators can optimise the problem of
reducing the waiting time for the users.
So that psychologist said that there is no need of any extra elevator, I have a very simple
solution of the problem and solution was that please install large full size mirrors in the
waiting area and when passengers users are waiting for elevators they will see their full body,
they will see their full dressing and they will be busy in just making them self more smart and
with that they will not feel that they are actually waiting.
And this simple solution solve the entire problem and rather it increase the satisfaction level
of the users also, so the team approach is very very important that many a times you alone
cannot solve the problem and a different person in your team will give a different perspective
to the solution of the problem, so we need team approach, we need people from different
background.
So with this you can see that how predictive analytics are increasingly important to supply
chain management because it makes the process more accurate, reliable and it reduces the
cost of your supply chain activities from raw material to suppliers to manufacturing to
distribution to customers and finally it is coming to consumers and these are the various steps
in the supply chain from the procurement inbound logistics, then you have manufacturing
activity, then the finished goods inventory is going to the customers ,all others steps we have
discussed many times.
406
So at all these steps your predictive analytics will help us to predict what is going to happen
in the different stages of the supply chain.
So just to give you the example that how things can change so using historical data of we
could determine that a part takes on average X days to arrive and even calculate standard
deviation to make some fairly sophisticated adjustments in our procurement plan and
similarly on the demand side we could look at historical demand data and try to extrapolate
demand into the future converting that to forecast production requirements and backward into
procurement and logistics requirement.
So these are some of the examples where predictive modelling can help us our resource, it
can help us in better planning the activities and therefore we will be in a better situation to
deliver as per the requirement and deliver at the low cost. So this is the use of predictive
modelling in the supply chain activities and in the next class we will see that how some of the
models can help us in reducing the uncertainties where we are talking of predictive
modelling, so how some of the models can help us in reducing the uncertainties related to our
future activities. Thank you very much.
407
Lecture-35
Representation on Uncertainty in Supply Chain
Welcome back in our last two sessions we discuss the role of analytics in the supply chain
and be focused about the predictive analytics that how we want to determine that on the basis
of past what can happen in the future and what can happen in the future, this is also link with
lot of uncertainties. If there is a well-defined period what has happened in the past
accordingly things will happen in the future.
But we all know that future is full of uncertainties and therefore we will talk of uncertainties
in today's session. I can take you to the situation before 1990’s, in most of the pharmaceutical
companies before 1990s the environment was very much certain and company is used to have
dedicated capacities in their plants, the meaning of dedicated capacity was that in a particular
plant you will have the entire facility dedicated for a particular kind of medicine.
But nowadays if you go to a pharmaceutical company you will find that most of the
companies are having flexible capacities, because it is highly uncertain that what kind of
problems will be there and problem means what type of new decisions will be there and
accordingly you will require new types of medicine and therefore on a very fast basis your,
search are changing and accordingly the new types of medicines you need to manufacture in
the same plants.
So therefore flexible capacities are required and these things are the subject matter of
uncertain environment which are going to be there and we will discuss some of the specific
cases that how some of the tools can help us to handle the issues related to uncertain
environment.
408
The type of uncertainty which we are going to discuss that is binomial representation of
uncertainty, there can be normal representation of uncertainty also, but in this session we are
going to discuss the binomial representation of uncertainty there can be 2 cases.
One case where we have the multiplicative uncertainty and in another case we can have
additive uncertainty, this slide gives us the example of multiplicative uncertainty. Now there
can be uncertainty with respect to so many factors, your are demand may be uncertainty in
the future, your price of the commodity maybe uncertain in the future, the exchange rate
maybe uncertain in the future, the supply can also be uncertain in the future.
So you can take any underlying factor at once to understand the method of solving such type
of situation, so here I am taking one underlying that is price and I am representing the price
with capital P in my starting time at 0 my price of the commodity is P, now price can increase
409
or price can decrease in the subsequent period. So I am taking two factors, one is u and
another is d, u is greater than one that price is increasing.
If I multiplying this u with my P, so that is the increase price in the next period and d is less
than 1 if I multiply P with small d the price is decreasing in the next period. So for example if
price can increase by a factor of 5 % the value of u is 1.05 and if I say that price can decrease
by a factor of 5% so the value of d is 0.95, so if price P is in time 0, so in time1 period one
that price can have two values Pu and Pd.
So if P is 100 so Pu is 105, if Pd is there it is 95, now since it is uncertain environment the

probability are also required, so the probability is p for this u and 1-p for this d, so the
chances of getting this price Pu is having a probability of p and chances of having this price
Pd is having a probability of 1-p. Now going further in the next period after period 1 , Pu can
further increase, Pu2 it can become and Pu can decrease by a factor of d also.
So Pu can become Pud. Similarly Pd can also increase with a factor of u, so Pd can become
Pud and Pd can further decrease with the factor of d, so Pd can become Pd 2, then Pu2 which is
there in the period 2, can further increase with the factor of you, so it can become Pu 3 and Pu2
can decrease with the factor of d, so it can become Pu 2d. Pud can also increase or decrease it
will increase with a factor of u, it will decrease with the factor of d.
So when it increases it becomes Pu2d and when it decreases its becomes Pud2 and so on fpr
Pd2 it can also increase or decrease when it increases it becomes Pud2 and when it decreases it
becomes Pu3 and similarly from period 3 when we move further to period 4 Pu 3 can also take
two values it can increase with the factor of u, it can decrease with the factor of d, so it will
take Pu3 to Pu 4, and Pu3 to Pu3d.
Pu2d will take Pu3d and Pu 2d2, Pud2 will take 2 values Pu 2d2and Pud3 will also take 2 values,
Pu4 and Pud3 and you can go for period n also by doing these type of calculation each time
you have two possibilities that your present underlying factors may increase or decrease and
probability of increasing are p and probability of decreasing is 1-p.
410
Now these are the way in which you can write on a generic basis that how things are going to
change in a multiplicative binomial representation and this gives you a very general type of
understanding, if you recall the previous slide that the possibility can be represented as
Pu tdT-t , t=0,1...,T.The For period t for period 2, period 3, period 4 you can find all possible
outcomes in this manner.
And therefore it is also very similar simple to understand that at period zero we are having
this price P at period 1 price may take two values that is Pu and Pd, so for that let us go back
to the previous slide, so the price is taking two values Pu and Pd, further in period 2 Pu may
take two values one is Pu2 and another is Pud when this factor is decreasing you can see on
the screen also.
411
Similarly Pd will also take 2 values one when it is increasing it will become Pud and when it
is further decreasing it becomes Pd2, going to the third period Pu2may take two values, it can
increases, so it becomes Pu3 and when it decreases by a factor of d it becomes Pu 2d. Similarly
Pud can also take two values it can it increase and when it increases it become Pu2d.
And when it decreases it becomes Pud2, similarly Pd2 can also take 2 values, it can increase
when it increases it becomes Pud2 and when it decreases it becomes Pd3 and you can
simultaneously match the development of this kind of branch diagram with uncertain values
of our underline factor which we have given on the slide. Then further if I go to period 4 Pu 3
will also have two possibilities that is Pu 4.
If it increases and if it decreases it becomes Pu3d, Pu2d also has two possibilities if it
increases it become Pu3d and if it decreases it becomes Pu2d2. Similarly Pud2 also has two
possibilities, if it increases it becomes Pu2d2, and if it decreases it becomes Pud3 and similarly
Pd3 also has two possibilities, if it increases it becomes Pud 3 and if it decreases it becomes
Pd4.
And you can match this diagram with the values which we got. So this is a very simple way
to develop the uncertain environment this is a very simple way and just by going into the
same methods you can develop this diagram.
And therefore the next slide which gives us the idea about how you can have a generalized
representation of any state in the development of this uncertainty diagram, that is Pu tdT-t for a
412
particular period T , T can be period 1, period 2, period 3, period 4 and in these period you
have a variety of states and these formula will help us to go for all those different types of
states.
There are in each of these things in a particular period from this state the price may move in
increasing or decreasing order you see let say in state 2 Pud at time 2 Pud is a state, now from
Pud price can increase it becomes Pu2d, price can decrease it becomes Pud2, so both these
movements are possible, so this is again the generic representation of 2 possible states that
the probability of increasing the price is p and probability of decreasing the price is 1-p from
a particular state to the next time interval.
So this is the generic representation of uncertainly and this is the specific movement in this
particular case. This is about the multiplicative uncertainty, in most of the cases in our supply
chain decision we will be using the multiplicative uncertainty that uncertainty is increasing or
decreasing as a multiplying factor to our underlying factor, it is also possible to have additive
representation of the uncertainty.
In that case these P will have addition of this factor u and this will have negative of the d, so
it is P + u the underlying factor is increasing by u and underlying factor is decreasing by d, so
you are adding and subtracting a fixed quantity from your underlying factor, so if this type of
issues there it is the additive representation of the uncertainty, so similarly you can go for all
subsequent period if you want to use this P + u or P – d, so all these representations will
change.
And we will have a new table for representing the additive binomial uncertainty, but there are
certain cases where it is not possible to have the additive binomial uncertainty, particularly in
case of price, demand etc, which are very important in case of supply chain, because you all
can understand when you have this additive or subtractive values + u ,–d you can get to
negative values also and it is not possible to have negative demand, negative price etc.
You can understand for example if price is initially Rs. 10 and you are talking of uncertainty
that price can increase by + 2 rupees and price can decrease by - 1 rupee. So in that case it is
possible that if you talk of this type of length P – d, P – 2d, P – 3d, P – 4d, P – 5d and in 11th
413
period it will be P – 11d and at that time price will be negative, at that time you can imagine a
situation where the price will be negative.
And that is not possible you cannot have a negative price, similarly demand is also there you
have initial demand of the product, let say 100 units in a particular time period that in month
1 you have the demand of 100 units and there are values of + 10, - 15 that demand can
change by 10 units or can decrease by 15 units. So if I talk of this last leg there may come a
time when demand make go for the negative values, so that is also not possible.
So many times you will see because our underlying factor cannot take the negative values it
is not possible to use additive binomial uncertainty, so we will use in most of the cases the
multiplicative binomial uncertainty, then there is another reason for that reason one reason is
this that you cannot have the negative values of underline factor, the other factor is that it is
not possible to have some kind of change in a very small time like for that purpose if my
price of a product is 15 dollar.
In that case it is not very appropriate to say that price may change by a factor of let say 5
dollar, it is very you can say things which is not possible, it is not appropriate to say that price
can change by a factor of plus minus 5 when the original price is 15 dollar. However same
thing makes sense if the prices 500 dollar, if the price is 500 dollar in that case if I say that
price can change by plus minus 5 dollars in a period it makes sense.
So you need to see that what is the size of my original underlying factor and in that case
probably you can use the additive uncertainty, but if the size of the original underlying factor
is relatively small again it is not advisable it is not possible, it is not good to have the additive
uncertainties because you will not have this type of fixed values which can be added or which
can be subtracted from the original underlying factor.
So whenever the size of the initial underlying factor is relatively small we go for the
multiplicative binomial uncertainties, so that is another important reasons, so two things are
there that sometime your underline factor cannot take the negative value which probably are
possible in case of additive uncertainties, so you will not take those things and when this size
is also a small it is not advisable because in that case fixed uncertainties are not possible.
414
So uncertainties are percentage of the original values, so in that case also will go with the
multiplicative type of binomial uncertainties. Now once we are clear about this binomial
representation of uncertainty it also important to say that here we are only taking two
possibilities that price increase by either by factor of u or price can decrease by a factor of d.
So when my original price is P it can take only two states Pu or Pd.
But in real life it is not so, in real life price can increase by different values, price can
increase by u1, u2, u3, there are different probabilities for increasing the price from p to pu, Pu ,
1
Pu2, Pu3. Similarly price can decrease to d1, d2, d3 there can be different underlying factor,
different values of d that d1 is 0.95, d2 is 0.96, and d3 can be 0.92. Similarly u1 can be 1.05, u2
can 1.07, u3 can be 1.10.
And you will have different probabilities for all these things that the probability of u 1, u2, u3,
d1, d2, d3 are different from moving P to that state, but here we are considering only two
states, so there are two reasons for that one is that we are starting, we are just learning that
how to incorporate uncertainty in our discussion, so for that purpose it is easy to take only 2
states, that is one answer.
The second answer is that though if your time period is slightly long then all these u 1, u2, u3,
d1, d2, d3 will come, but if you take a very small time period then probably you can divide the
uncertainties only into two states u and d. So this type of situation can also be taken
practically, it is not only just for the academic discussion, it can be taken practically also, but
415
then we need to take the very smaller time period where you can only think of two
possibilities whether u or d.
As you increase the time period you will have chances of different states with different
probabilities, all those things can also be handled but that will be slightly more complicated
model and then thanks to IT, thanks to computer science, these things will help us in handling
those complex situations also. So if I reduce my time period if let say I am doing the
determination of pricing on the monthly basis, probably u and d is fine.
If I take demand factor on a weekly basis u and d is fine, but if I take price on yearly basis
then probably u1 to u3 and d1 to d3 can come, so I need to see that what is the optimum, you
can say time period for taking this into my account, so that u and d only two possible states
are required, so these are the initial things and when you do this binomial representation of
uncertainty for fairly small time period.
And you do it for a long time you have T=n and n is fairly large, then this binomial
representation of uncertainty becomes almost equal to normal distribution of uncertainty, so
that is also important to understand that then it is nothing but simply the normal distribution,
so we should take in better decision making better predictive analytics we should take a
smaller time period and smaller time period for long time.
So that you can have the normal distribution of the uncertainty in the long run, so with this
background now this is the additive representation of the uncertainty just as we have
discussed if initial underlying factor is P, so it can take 2 states P + u, and P-d and similarly
as the explanation was there for the multiplicative uncertainty P + u will again have 2
possibilities, so if it is increasing it becomes P + 2u and if it is decreasing it becomes P + u –
d. Similarly P – d also has a 2 states where it can increase or decrease.
So when it increases it becomes P + u – d and when it decreases it becomes P – 2d. I am not

going to discuss the state 3 and period 4, I request the students that you write that what will
be the different states when we are moving in the same way for the additive uncertainty and
try to match your answer for what we have written on the slide that these are the different
states in period 3 and these are the different states in period 4.
416
So please do not see these 2 lines and try to write on your own and then you can match that
whether you are also getting the same kind of states for period 3 and period 4 and again the
simpler way of achieving these states is to go by this type of tree arrangement that how you
are getting different types of states during the different time periods, so that this will help you
to achieve that tree for the additive uncertainty.
And here also you have generic representation of additive binomial uncertainty that T has all
possible outcomes at a time period t, that is P + tu (T-t)x d and for this t you can have all
possible States from t = 0 to T, so that is the way you can develop your different states for a
particular time period T. So like for an example if I take T = 1, so your different possible
outcome using these binominal representation are P + Tu – T – t x d.
And I will take 2 values of small t and these 2 values are t are 0 and 1 T is one, so
possibilities are P + 0u (1-0)x d so one state becomes P – d, the other state the next time I will
take the value T = 1, so this becomes P + 1u (1-1)x d, so this becomes P + u and you can see
that at time period 1 we have these 2 states P + u and P – d.
So similarly you can determine the states for any time period using this generic representation
of additive uncertainty, but the additive uncertainty as we discussed can only be use where
we have large size of our initial underlying factor, so we stop our discussion at this point and
in our next session we will take a numerical example to discuss the role of uncertainty in
predictive analytics. Thank you very much.
417
Lecture-36
Using Decision Tree for Handling Uncertainty
Welcome back and we were discussing in our last session about use of uncertainty in the
predictive modelling and there are different types of uncertainties, we can have uncertainty
related to demand, related to price, related to inflation, related to supply. So all those
uncertainty is are there. There are uncertainties which we need to handle immediately and
then there are uncertainties which are we handle in the long-term.
Now in last session we discuss the binomial representation of the uncertainty and we discuss
two types of binomial representation, one is multiplicative and another was additive. And we
saw that how if we take one underlying factor P and it can increase in the subsequent period
or it can decrease in the subsequent period. So here in that case we took price as our
underlying factor P and small u is the factor by which it can increase in the next period. And
d is the factor by which it can decrease in the subsequent period and we saw these type of
table and this we also discuss with the help of a binomial tree.
418
Similarly we also discussed in the last session about additive uncertainty where u and d are
the amount of uncertainty is in the subsequent period, so if it is additive uncertainty our price
will increase by P + u and price will decrease by the factor of d, so similarly we can have
binomial tree representation for these additive uncertainties and that we discuss that why
binomial uncertainty and particularly multiplicative uncertainty is more used in the supply
chain decisions.
Because in case of additive uncertainties it is quite possible that your underlying factor may
get some kind of negative value, if this is price P and it is decreasing by fixed value d over a
period of time, so you see that in fourth period it is becoming P - 4d and nth period it may
become P – nd and in that case this Nd factor may be much more than P and there can be a
chances of negative price also which is not possible.
So in reality most of the time we will see that we will take multiplicative uncertainties in our
consideration and the other reason we also discussed that many a times the size of your initial
underlying value is also very important, if the initial price is 20 rupees, so 20 rupees may
change with the factor of 2% 3% 4% but it is difficult to say that 20 rupees may change with
a factor of 2 rupees in the subsequent period.
So you cannot assign a fix value if your initial values are small, if your initial value is let say
1000 rupee then the price may change by a factor of 5 rupees in the subsequent period. So
that is also a limitation in the use of additive uncertainties in this case. So the example which
419
we discuss in the last case that in today session we are going to have a numerical example to
discuss these representation of uncertainties in our decision making.
Now for that purpose we will take the help of binomial decision tree we discuss in our last
session, now in today session we will see that how we can represent the decision tree for the
purpose of uncertainty and with the help of probabilities associated with each node of
decision tree, before I go for a specific example let me give you the idea of decision tree, that
how it all evolves for that involves.
For that purpose this is a condition at state 0, now there are two possibilities this period zero
and in period 1 you have two possibilities, the probabilities for the sake of simplicity we have
same probability for both these states, then we can go further from this state of time period
420
one there are further two possibilities or let us say 4 possibilities and each with the
probability of 0.25.
And these are the further states, and so on with the state number 2 of time period 1 you can
have another 4 possibilities each with a probability of 0. 25, this is time period 2, and in the
same manner we can develop this diagram, so at each these are known as nodes, so all these
nodes will give rise to branches, these are the branches coming out of these nodes and further
you get node and then another branches are there.
and each branch will terminate at a node and then another branches will be there, so it is like
a tree type of arrangement where you have nodes giving rise to branches nodes giving rise to
branches and so on .Now this graphic representation of decision tree is a very useful tool we
already have gone through this type of decision tree discussions in the classes of statistics in
mathematics also.
But here we will see a practical application of decision application of decision tree in
evaluating our decisions with respect to supply chain where uncertainty is there, because all
these possibilities if uncertainty is not there so there will not be multiple branches, there will
only be one branch, you have one thing which is 100% sure, but since there are uncertainties,
so for those uncertainities is to inbuilt in your system you have different branches with
different probabilities.
421
Now let us see what this slide says that a manager must make many decisions when designing
a supply chain network, so you need to make decisions like a simple example I can say you
that I want to develop facilities in my supply chain and I am deciding about a warehouse, this
is a facility to be developed in my supply chain. Now for this warehouse I have three
different alternatives, what are the three different alternatives?
One alternative that I go for a lease arrangement for certain number of periods, so there is a
lease arrangement for 3 years, 4 years. So I am going for lease arrangement where I have a
fixed capacity dedicated capacity given to me for a certain number of period. The other
option is the spot market. In spot market I go and take the space for the warehousing purpose
as demand is there.
In case of lease and arrangement even if my demand is not there I will incur that much cost
because I have taken it full hall for that purpose and whether I use that full hall or part of that
hall, it is a immaterial for the person who has given me that hall. So I will incur a fixed cost
for all through the period. In case of spot market the chances are there that I will incur
slightly higher cost.
Because whenever in demand is there I will go and take the space so slightly higher cost I
will incur, but here I will not incur the cost for unused capacity. I will take only that much
space for warehousing which is required, which is in my demand. So though the cost maybe
higher but cost means the rate at which I will be paying the rent that maybe hire in case of
spot market then the lease maybe in the lease arrangement.
I pay just to give you an example I may be at a rate of 1 dollar per square feet that may be the
rate and if I go for the spot market I may pay at the rate of 1.25 dollar for per square feet, so
that is kind of difference you will normally see and but the advantages that I will only pay for
the used capacity here because I will take only that much space. Here in lease case I may take
more space because of cheaper availability of the space.
But later on I may find that some of the space is not used and then the third option is flexible
system flexible arrangement. In flexible arrangement it is possible that I take some space
fixed that this much minimum space I am going to take on the lease, it is a combination of in
422
fact both these things lease and spot. Flexible is a combination of lease and spot, where I take
some space on lease and the remaining requirement I fulfill on the spot market.
In that also there is a fear that whatever minimum I decide for lease even that may not also be
used, but it still the loss because of higher and unused capacity can be minimise to some
extent if I go with the flexible. So normally flexible is done to take the advantage of lease as
well as of a spot. Both these things are combined in this flexible case. So these are certain
type of decisions which a manager has to take in a supply chain network.
Similarly there are many other decisions maybe related to inventory how much you stock,
want to stock, so you can follow any such practice where you have a fixed quantity and then
you may order as per the requirement for some quantity you follow JIT and for some quantity
you follow up EOQ type of model. So both these types of combinations are possible, so
managers need to take decision in your supply chain and these decisions are with respect to
facilities.
These decisions are respect to transportation facility, I am talking of warehousing same you
can take to the transportation also. The same situation we can take to transportation also
where in transportation also all these three things are possible, how much trucks, how much
vehicles you want to own that maybe the lease kind of arrangement, you want to own some
number of vehicles.
Then some number of vehicles you take from the market and then there is a possibility that
you have some minimum number of vehicles with you and remaining requirement you fulfill
from the market. So all three alternatives available even in the case of transportation. So as a
supply chain manager you need to make all these decisions once you go up in the hierarchy
and please see that these decisions are long term decisions, I took the example of warehouse
you cannot change the capacity of warehouse in a day or two.
It is a long term decision and it can very well affect the surplus of your supply chain, so you
have to be very careful about selecting these kind of options and therefore these mathematical
models will help us to take appropriate decision. So as the second point mention , many of
these decissions involve a choice between a long term or less flexible option and short term
or more flexible option.
423
Like in this particular case if I see when I go for lease type of arrangement, so it is a less
flexible I got a fix capacity with me and it is a long term kind of arrangement nobody is going
to give me lease for one month or two month I need to have lease for 4-5 years. So then only
lease arrangement is possible, but it is less flexible if the capacity unused I have to bear the
losses. So that is one option.
And another is the short term or flexible option where spot market type of options are there
depending upon the requirement how many units I want to store and depending upon the
space requirement for those units I can go to the market and I can take the space on rate
which is highly flexible and in each period I can change my decision. So you have a trade off
between these two things.
Obviously when long term decisions are there and less flexible these decisions are there,
these are slightly cheaper and when you are taking short term and more flexible decisions
these are slightly costlier. So you need to see the overall benefit in these two types of
decisions depending upon the amount of uncertainty in your environment, if it is highly
uncertain environment obviously we will like to take short term decisions, if it is less
uncertain there you can take long term decisions.
If uncertainty is ignored the long-term option will almost always be selected because it is
typically cheaper so when uncertainty is less you will like to have long term decisions
because in that case you will incur minimum possible cost of taking that decision, such a
decision can eventually hurt the firm, however because actual future prices or demand may
be different from what was forecasted at the time of the decision.
Because uncertainty is going to be there, you cannot eliminate uncertainty and if uncertainty
is going to be there certainly these types of decisions which initially we thought are cheaper
may end up in a very costly affair for the company because there will be changes in the
demand there will be changes in the price is there will be fluctuations of inflation rate, there
will be fluctuations of exchange rate.
So all these things will contribute towards making our initial decision slightly costlier. So
therefore it is possible that we should include elements of uncertainty in our decision making
424
and probably in that case our decisions will be more realistic. Then finally as I was
mentioning that we will use decision tree for evaluating these decisions with the uncertainty
cases.
Now what is this decision tree methodology, so let us see that how are we going to use this
decision tree methodology for evaluating decisions with uncertainty. The first thing is we
need to identify the duration of each period and the number of period T over which the
decision is to be evaluated. Like in this particular case here first thing is we need to decide
that what is the duration of each period.
Like this is period one, this is period 2, this is period 3 etc, etc, I am making this decision
tree, but what is the duration of this period one after how much time this period is going to
come, so this may be monthly that this is January 2017, this is February 2017, this is March
2017 like that, it may be a quarterly if it is January 2017 then this period one is April 2017
and so on it can be a half yearly issue also it is January 2017 then period one is July 2017 and
so on.
In some cases it can be annual affair also it is January 2017 then January 2018, then January
2019, so first is what is the duration of each period, whether it is month quarter half yearly
early etc, so that is one thing you need to decide. If you remember in our last session we
discuss like here from node one we are coming up only with two possibilities that either your
price can increase or your price can decrease and price can increase by a fix value that is u
and price can decrease by its is fix multiplying factor that is d.
425
But in reality there can be various factors by which price can increase, there can be various
factors by which price can decrease. So it is advisable then it is a slightly more complicated
situation. So it is advisable to keep your duration of period slightly limited, slightly smaller.
so that in that small period you can come up with only two possibilities then our decision tree
is more realistic and simpler also to handle.
So it is normally said that if it is a price kind of thing, demand kind of thing we can go by
monthly periods in that case and then the number of periods t over with the decision is to be
evaluated then 0 period, first period, second period, third period, fourth period etc. So now
these are the three periods over which I am evaluating the decision and for a bigger case
when I want to take decision for slightly more longer period you need to take higher values of
T
So we will see a case of 3 period but in reality you may take cases of 5 period also and
because nowadays we can do all these things with the help of computers, so we can take a
period upto 10 years also or 10 period, 10 months, so all those things are possible so first we
need to take decision about the duration of period and the number of periods for which we
want to develop the decision tree. The second factor is identified factors such as demand
price and exchange rate whose fluctuations are to be considered for the T period.
So like price we have discussed time and again that we are going to discuss price P, Pu, Pd
and then there are different possibilities of Pu and different possibilities of Pd also. Similarly
you can have demand also. Demand is increasing or decreasing, so you need to decide that
what the factors you are going to consider for your uncertainty calculation. Then identify
representations of uncertainty for each factor that is determine what distribution to use to
model the uncertainty.
So you need to see that what type of distribution is best suited for your case of uncertainty,
which particular probability distribution suits your case of uncertainty, we are taking in our
class the binomial distribution of uncertainty, though you can take normal distribution also
and if you remember in our last session we discuss that if we take binomial distribution for
long period of time with the same values of up and down.
426
So it can become like a normal distribution also. So you need to see that which type of
distribution you are going to take to represent the uncertainty. Then we also need to see what
is the discount rate that is K for each period, it means that because we need to take decision at
time = 0, so whatever is the outcome of a particular series of nodes and branches we need to
calculate the present value of that outcome.
Whatever is the outcome coming after this period 2, so we need to calculate the present value
of this and for determining the present value we need this discount rate factor K. It is possible
that from period 1 to 2 you have a particular discount rate and from period 1 to 0 you have a
different discount rate. In reality it is possible and it may happen also. But in our case for
understanding the uses of this method we are taking a constant discount rate K for all the
periods.
But please be sure that in reality it can be given k1, k2, k3 etc. for different time periods. So
we will see in our discussion that how to handle the variable discount rates that it is also
possible. Then represent the decision tree with define states in each period. These are the
states so you need to develop and these are the time period.
So you need to develop the decision tree with define states in each period as well as the
transition probabilities between states in successive period. These 0.5 these 0.25 0.25 0.25,
these are the transition probabilities between states in successive period, from period 1 to
period 2, from this state you are getting these four new states, and the transition probabilities
for these states are 0.25 each.
From this state in period zero you are getting 2 states in period 1 and the transition
probabilities for these two states are 0.5 each, so that transition probability also need to be
defined for all the different nodes which you are going to get in your nodes means a different
state you are going in the successive period. Then the calculation starts from the most recent
period means the last period in your decision tree.
So if I consider only this much as my decision tree so my calculation will start from period 2,
form period 2 I will go to this direction, first I will calculate the values here, then here and
there here, so this is the way in which I will move for calculation. So starting at period t work
427
back to period zero identify the optimal decision and the expected cash flows at each step. So
because all these things are probabilistic in nature.
So therefore we need to see what is the expected cash flow and therefore whatever cash flow
is there at this stage this will be multiplied by the respective probabilities and therefore we
have the expected cash flow value and then expected cash flow from period 2 to period one
will come with the help of time value of money concept using the discount rate K for this
period.
So expected cash flow at each state in a given period should be discounted back means we
need to calculate the present value of the cash flow when included in the previous period. So
you need to calculate the cash flow of this period and you also need to calculate the cash flow
the present value of the cash flows of the second period, so that will be the total cash flow at
state 1.
Similarly when I am talking at state 0, so I will calculate the present value of cash flows of
time period one and the cash flow at time period 0 that will be the total value of cash flow at
time period 0. So this is how I will go for the backward calculation from time period T to
time period 0. Now for understanding this we will use some kind of numerical example so
that we can understand properly the use of decision tree methodology for handling the
uncertainty. And this is what our predictive analytics will tell us.
428
Now it is a situation where we want to hire some kind of warehouse and in that warehouse we
have these are the three options available to me, decide whether to lease warehouse space for
the coming 3 years, so this coming 3 years this is the T, period of my decision tree and
because I am talking of yearly basis so each period is 1 year. So now the first thing if I go
back that what is the duration of each period that is one year. And the T is 3 years.
Then long-term lease is currently cheaper than the spot market rate the situation is that which
is very obvious that if I go for a long term lease agreement, this is cheaper than the spot
market rates. The manager anticipates uncertainty in 2 things, 1 is he anticipate that there
may be a fluctuation in the demand. So he is not very sure that the space which I take today
the same requirement will be there tomorrow or not.
So that is one level of uncertainty and the second level of uncertainty is the price. The price
of the commodity, the price of the spot market will also change. So today spot market price is
different, tomorrow spot market price will be something else. As we discussed long term
lease is cheaper but could go on unused if demand is lower than the forecast. And spot market
rates could also decrease.
Today the rate of a spot market is 1.4 dollar per square feet but it is possible that tomorrow
the rate may decrease to 1.25 dollar per square feet, so today it will look costlier to go for the
spot market but since there is a possibility that my demand can also decrease. So if I take
more space that space may go unused and another possibility is today the price of the spot
market is high, but tomorrow is this market price goes down.
So that may also give me some kind of attractiveness to go for the spot market. So these are
the dilemma of manager. Spot market rates are currently high and the spot market would cost
a lot a future demand is higher than expected, and future demand is high so today it is already
a higher price of the spot market and if more demand is there in the future and if the same
price trend goes.
So it will be really very costly for me to bear the cost of this warehouse. So we will see
different scenarios in this case, we will develop different scenarios and with the help of those
scenario we will try to calculate that how this type of calculation can we handle. So that
entire calculation part we will see in our next class. Thank you very much.
429
Lecture-37
Example of Using Decision Tree Incorporating Uncertainty in Single Factor
Welcome back, in our last sessions we were discussing about use of binomial uncertainty in our
network design decisions. And, in the last session we discussed in detail about the use of
decision tree for the purpose of evaluation of uncertainties in the future decisions. Now we have
already gone through the representation of uncertainty, we have already gone through the
concepts of time value of money and now in the last session we were discussing that what are the
important things we should keep into the mind when we are going for a decision tree model.
And, we discuss that we need to have the period, we need to have the duration of each period,
then we need to find that what are the underlying factors for which we are evaluating the
uncertainty then we need to represent the factor of uncertainty for all these types of underlying
factor. Then we also need to know the time value of money and for that purpose we require some
kind of discounting rate.
Then, we also need to have the what is the transition probability from one state to another state.
And then we start calculating from period T, the last period to period 0. From where, we have
430
started the building of this decision tree. So, all these things in a very detailed manner we
discussed in our last session.
Then we came to a particular specific example about deciding the warehousing situation that we
have alternatives where, we can take the long term lease decision, and we can take a spot market
also. There are certain issues related to long term lease and spot market. A spot market is
currently costlier, but there may be fluctuations with respect to prices of the a spot market at in
future the prices of a spot market get go down.
Then there is a situation of demand also, today there is a demand and according to that demand, I
take lease space, but it is quite possible that tomorrow demand may go down and in that case
when demand can go down I have some kind of unused space in my warehouse, so that is also
extra cost, so all these things we discussed.
431
Now coming very specific so, there are 3 possibilities considering all those situations which we
discuss in the last session and in the previous slide as well. We have three options available to
this company, these options are get all warehousing space from the spot market as required. So,
today you require 60 square feet, so that take that much. Tomorrow you require 100 square feet,
you take that much, then you require 90 square feet lease some 10 square feet, and take only 90
square feet.
So, that type of flexibility is there in this first option, the second option that you sign a contract
for 3 years, and it is a fix amount of warehouse space, and get additional requirements from the
spot market. So, you have taken let us say some 100 square feet area, from the lease agreement,
and if demand exceeds beyond that so the additional requirement can be fulfilled from the spot
market at whatever price it may be available at that period.
And the third is sign a flexible lease that you have taken a minimum amount of fix area. and then
you have some kind of variable uses of warehouse space, up to a limit and additional
requirement from the spot market. So, this is taking the advantage of the spot market condition,
as well as for the leases condition you can have this kind of flexible ware for some space.
May be for, 60 square feet you have this rate of uses, and beyond that, you have some other rate
of uses, and then there is the upper limit of the availability of the space in this agreement. And,
432
then even you go beyond that, if demand increases beyond that the additional requirement can be
fulfil from the spot market. So, it takes the case of both spot as well as your lease agreement.
Now, coming to some specific data. So, the data says you require 1000 square feet of ware house
space, for 1000 units. One unit take 1 square feet, so for 1000 units you take 1000 square feet.
See current demand is 100,000 units per year. That is the current level of the demand. The
binomial uncertainty is demand can go up by 20% with a probability of 0.5, and demand can also
decrease by 20% with a probability of 1-p, that is again 50%.
So, chances of increasing and decreasing are same, demand can increase same here, we are
taking the same values 20%, for the increase in 20% for the decrease. But, in reality it can be
different also, it is possible that the chances that demand can increase by 30% is 50%, and
demand can decrease by 10% is 50 %. So, that type of data also there. It can also be modified
that you have same value of u and d but different probabilities.
The demand can go up by 20% with the probability of 60%, and demand can decrease by a factor
of 20%, with a probability of 40 %. So, all these combinations are possible. Here just for the sake
of simplicity you can say that simple data is taken. Where demand u if you remember introduce
this term u, and d in our last sessions. So, that u and, d these are same.
433
The value of u becomes 1.20, and value of d becomes 0.80. And, the probabilities for both these
things are 50% and 50%. Though in reality it can be different also. The u can have a probability
of 60%, it can have a probability of 40%, and all other alternative combinations are possible, so
as long as you have p and 1-p. So, you can have variety of cases. Then lease price is given to us
that is that 1 dollar per square feet per year.
That is the lease price, if I go for lease agreement, so I will be charged at this rate. The spot
market price currently is 1.2 dollar per square feet, per year. That is the spot market price. But,
there can be an uncertainty in a spot market price also, and a spot market price can also go up or
down, by a factor of 10%. So, this u and d is for the demand. And, there will be another u and d.
That is 1.10 and d is 0.90, this is for a spot market prices. So, you have 2 u and d, and it is a
matter of chance that for both these the probabilities are 50% again. So, spot market prices can
go up by 1.10. The probability is 50% can decrease by a factor of 0.90, probability is 50%. And,
the revenue is from these units, which you are storing in your warehouse 1.22 dollar per unit.
And, for determining the present value of the future cash flows. The discount rate is 10%. So,
that is the discount rate K. So, that is 10%. Now, let us see the development of decision tree for
this set of data.
434
So, it is good if to understand this development of decision tree, we keep this data readily
available on one side of our notebooks somewhere you note down this data. And with the help of
this data we will see the what we develop this decision tree, on the board, and then finally we
will see that whether our looks like that the decision tree which there in the next slide.
So, let us see how to develop the decision tree with this data, and I expect with that you will also
be able to develop decision trees on your own.
First is the time t=0, this is the period 0, where you have some demand, and the price of the spot
market, so initially is current demand is let us say 100,000 units. So, for the sake of simplicity I
435
am simply writing the 100. And, it is into D is n units, so you are clear that three zeros are more,
after this 100. And, the initial price of the spot market is 1.2 dollar per square feet per year. So,
that is the initial price that is 1.20.
Now, we move to period 1, in period 1 what are the different possibilities 1 possibility is that
demand increases, and price also increases. So, demand is increasing, let me give you this with
the help of + +. Demand is increasing and, price is also increasing. Both these things are having
the positive things. So, the factor of u for demand is 1.20 and so, this demand will become 120.
And, price for that factor is 1.10, so price will become 1.20x1.1=1.32. So, you can see that this is
the first case that in period 1. From here you are moving to this place. That your demand is
becoming 120, and price is becoming 132. The next is second case where your demand can
increase, demand is increasing, but price is decreasing, so demand is increasing, so demand
becomes 120.
But, initial price is 1.20, and u is 0.90 for price, so it becomes 1.20 x 9, so it remains 1.08 that is
the another state, that d becomes 120, and price becomes 1.08. Then another possibility is, now
demand can decrease and price can increase, so demand is decreasing by a factor of d, and d is
0.80, we wrote initially so d becomes 0.80. And, p is increasing so p becomes 1.32.
And, similarly another possibility is that demand is decreasing, and price is also decreasing, both
negative. So, in that case d becomes 80, and price becomes 1.08. So, these are the 4 possible
states, 2 states where demand is increasing, 2 states where demand is decreasing. 2 states where
price is increasing, and 2 states where price is decreasing. So, in all these 4 states are possible.
And since the price and demand, these 2 things are independent of each other not affecting each
other.
So therefore the probabilities, now we need to assign if you go back to our earlier slide. That we
need to assign the transition probabilities also. This is step number 5, so we will assign the
transition probabilities, that the probability of increasing the price is 0.5 probability of increasing
the demand is also 0.5, so the transition probability here is 0.25.
436
And, so on for all other decisions also all other states which we are getting for all these different
states, the transition probabilities are 0.25, 0.25, 0.25 each. So, you get the complete set of 4
states at time period 1. Now, we need to go further, because we are considering for 3 years, so
now this is the next year decisions, from here you can have another 4 possibilities. And, these
possibilities are same, your demand can increase.
Demand can increase there are 2 possibilities, here price is decreasing here price is increasing,
then demand can decrease 2 possibilities, and in these 2 possibilities again price can decrease,
and price can increase. So, further you have 4 possibilities. But, these 4 possibilities will be
evaluated from this state now. So, if demand is increasing let us say if demand is increasing, so
here it is 120, demand increase factor is 1.2. So, it may become 144.
when demand is decreasing by a factor of 80, 0.80, do demand can become 0.80 x120, that is 96.
So, you should remember that I should write here u for d is 1.20, a small for d is 0.80, u for price
is 1.10, and d for price is 0.90. So, similarly when my price is changing, when I am having these
2 values where price is increasing. So, 1.32 multiplied by u that is 1.1. So, 1.32x1.1
And, when price is decreasing this is here, price equals to 1.32x.90. And, we need to have the
transition probabilities, chances of increasing the demand, 50% chances are decreasing the
demand. 50% chances are increasing the price, 50% chances of decreasing the price 50%. So,
again the transition probabilities are 0.25 for all this 4 states in period 2.
So, I request my participants that for remaining 3 states of period 1, you can develop similar type
of state for period 2. And, when you develop states for these 3 states. The successive states many
of those states will be the common which you get here in the, this discussion. Like when the
demand is 120, if it increases it will become it will become increase by a factor of 20%, so 1.2,
and then in that case it may become 144.
Demand may become 144.So, many of those are possible that where you have the common
values in these final state. So, I request all my participants that you please develop, and once you
437
develop this entire tree. Decision tree, so this is the representation of our various states, the
transition probabilities. And, now the evaluation of various alternatives.
That these 3 alternatives which are there, will start. So, but before that please complete this
decision tree. Because, this decision tree will help us and, I have this please be careful in reading
this values. These are 0.25, 0.25, 0.25, and 0.25. So, slightly misadjusted values are there. So, all
these I have written very clearly on the board. So, these are 0.25, 0.25, 0.25 as my transition
probabilities. Similarly, in this case of period 2.
This has 4 possibilities, which are represented here, and from this also you have 4 different
possibilities, from this state also you have 4 different possible states which are emerging. And,
from here also we have 4 different states which are emerging. And, now you can see, now when
you make your own decision tree diagram, you will see that many of these states are common.
For example, if we talk of this state D=96 and, price is 1.19, this is common in almost, because
you have 1 arrow coming from here, 1 arrow coming from here, 1 arrow coming from here, and
1 arrow coming from here, so this state is being represented by all the 4 states of the period 1.
You can have possibility. So, it is not exclusive for, you are going to get from here.
438
Your states will have various you can say common is gets, in these 3 other calculations. So,
otherwise 4 into 4 you will have 16 unique states, but you will not have 16 unique states you can
just count that we have in period to 1, 2, 3, 4, 5, 6, 7, 8, 9. 9 unique states, so that is something
we should understand that and this again rooted to that type of discussion which we have in last
to last session,
Where, when I am talking of Pu, when it can increase, and it can decrease. So, when Pu
decreases it becomes Pud, when Pd increases it becomes Pud, so this state becomes common for
both these previous states. So, same logic apply in this particular case also. That there are various
common states in developing this decision tree.
But, developing this decision tree, is very good visual representation of our entire state of affairs
which may be there in the future. Now, let us start discussion on the evaluation of the
alternatives. That how are we going to evaluate these alternatives?
439
And, for that purpose first we are going to analyse the option of not signing a lease, and
obtaining all warehouses space from the spot market. So, we are considering this option first, that
I am going to take the entire space on the basis of a spot market. So, as we mention that we need
to start from the last period. So, we will start with the period 2. And, calculate the profit at each
node, so at this case for D equals to 144, and price 1.45.
So, this is probably this situation where, 1.32 x 1.1. This is 1.45. So, I am starting with this
particular state, and how we will go backward, that you will come to know. For that purpose to
determine the cost, here we will write it in such fashion, D 144, p is 1.45, 2, this 2 signifies that
the time period it is 2 if I write comma 1, so this is the state at period 1. So, D 144, P 1.45 this is
the state and , this is the time period.
So, this is the situation, and what will be the cost, because the demand is 144 here. You see how
do we calculate this demand is 144, and price is 1.45, a square feet of space. So, and my
company is committed to fulfil the entire demand. Whatever demand is there I am going to fulfil
the demand completely, so I will multiply this price with this demand. So, and I have already
mention that D is in 1000 units.
So, this 144,000 x 1.45, this is the cost of the space, and if I am having this much demand 144,
so my revenue is for this option, my revenue is I will be selling, because my revenue is 1.22
440
dollar per unit of the demand , so I will multiply because I am fulfilling this demand 144,000. So,
144,000 x 1.22. That is my revenue. And, when I subtract cost from the revenue. So, 144,000
into 1.22 – this cost. That will become my profit.
So, I am subtracting 144 –this cost, and this is 175,680- 2800, this is -$33,120. It means that if
this is the situation 144 is the demand, and I am fulfilling the complete demand using the spot
market I will be charged at the rate of 1.45. My revenue is not going to change, whether I am
using spot market or lease. So, my customer is not worried about that.
So, I will get the fix amount of revenue on at the rate of 1.22. So, the simple logic is that my
revenue is 1.44 x1.22. that is the revenue which I am going to get. Revenue –cost, cost is 144 x
1.45 that is my cost, and so, this is the profit. So, 144 is common, so you can say that this -0.23.
So, this multiplication yields me 33,120 dollars. So it is a loss, so I will evaluate all possibilities
of such.
And, will develop the tabular kind of arrangement, and then will see that which option is better.
So, we stop here, we understood the method of calculation, but we stop here today, and in our
session, we will develop the tables and then we will evaluate, we will compare the 3 alternatives
which are available to our warehousing company. Thank you very much.
441
Lecture-38
Example of Using Decision Tree incorporating Uncertainty in Two Key Factors
Welcome back, in our last session we were discussing about developing the decision tree for a
particular type of situations where we had the problem of developing the decision for our
warehousing requirements. Three alternatives are there, and we developed the complete decision
tree, which is there now for the ready reference of us.
And this type of decision tree we finally developed in our last session. That the initial demand
was 100 units, the initial price in the spot market was 1.2 dollar per square feet, and chances of
increasing the demand are 20% , chances of decreasing the demand are 20% ,chances of
increasing the price is 10% and chances of decreasing the price is also 10%, and increasing the
demand, and decreasing the price means the uncertainties of demand and uncertainties of the
price are independent.
So I cannot say that by change of price demand is going to affect or by change of demand price
is going to be affected. So these two things are independent, and therefore the probabilities of
0.25, 0.25, 0.25, and 0.25 are there for combine effect of change in demand and change in price.
442
And, therefore this type of decision tree is available in front of us, now when this decision trees
are developed.
After that we need see that what are the three alternatives which are there in front us. So the first
alternative is that we can take all the requirement, we can take all the space from the spot market
and for that purpose. In the spot market we did this type of calculation that for a particular
situation where demand is 144 and the spot market price is 1.45, so the total cost of this type of
decision where my price is 1.45.
443
And demand is 144, 000 units, so we calculated in the last session you remember that total cost
of a space, because I am taking 100% from spot market, since I am taking 100% from the spot
market, so the cost will become multiplication of P, and D is the price, so the multiplication of
these two things, and then the second thing I need to calculate revenue from selling the units. So
I will be selling because, I am fulfilling the entire demand.
So revenue will be the cost the selling price that is given to us as 1.22 and into the demand and
then will calculate the profit that is revenue-cost and in this calculation with this data we got the
value that is coming -33,120 dollars. in fact this – sign signifies that it is a loss. If this type of
situation comes where demand increase from original 100 to 144 units price increases from 1.20
to 1.45 and we are fulfilling our total requirement from the spot market. So we are going to get
this is negative it means this is the loss we are going to have.
Now we need to see from this particular case, this is the value we got from period 2 and with this
type of state. Now we as discussed in the previous sessions from this we will go to time t=0, and
for that purpose from this is state I will come to this is state and let us how do we come to that
period 1. Now expected profit at each node in period1 is the profit during period 1 plus the
present value of the expected profit in the period 2.
444
Now when I am talking of period1 and this is 1 node, so the profit at this when I am trying to
determine, so what is the actual profit because of these things 120 and 1.32 is the spot market
price, and the present value of various expected profits which are coming from this is state there
are four states which are coming and these four states having the possibility of 25% each, so I
will see the possibility of getting this negative profit possibility of getting this loss of 33,120 is
25% in period 2.
Similarly, with respect to 144 and price 1.19 I will calculate the profit with D equals to 96 price
is 1.45 I will calculate the profit and the other state is D 96 and price is 1.19 I will calculate the
profit. And, for all the profits probabilities is 0.25 and that will be expected profit at period 2
emerging from period 1, and then the total I will calculate the present value of that total for
period 1, so that all calculation will discuss.
So, this is what I mention in the first point that expected profit at each node in period 1 is the
profit1 is the profit during period 1 + the present value of the expected profit in period 2. Now
expected profit EP that is D and P at a particular state1 ,1 means at time period 1, and for a
particular label of demand and a particular label of price. So when D is 120 price is also 1.32 that
is the first state if you go back to your note books.
And see the decision diagram at a node is the expected profit over all four nodes in period 2, that
may result from this node that we already explained then present value of expected profit. The
next thing is present value of expected profit in our some earlier session, we have discussed
about the time value of money concept and with the help of that time value of money concept,
we understood that whatever is the future streams of cash flow you need to calculate the present
value of those future cash flows.
And for that purpose we use the discounting rates, and in this example we have taken K equals to
10%, because we are going to calculate finally at time t=0, so we need to determine that this is
possible after time t2. And what will be the value of this in the present circumstances, and for that
purpose the present value concept is required.
445
So, present value of expected profit and the total expected profit is the sum of the profit in
period1 and the present value of the expected profit in period 2, so this is about time period 1.
Now, just to see the calculation that how this expected profit calculation has taken place for that
purpose you see from node D =120, and to have better understanding we have this diagram here,
let us go back to that diagram and quickly. Let us have that diagram on the board, so that you can
easily see that how things are moving this is D equals to 120 and P equals to initially 1.20 then
you have four possibilities and let we discuss this only`
Where the D is increasing to, this what initially the 100 and then it become 120 price was
initially 1.20 then it becomes 1.32, and then there are further four possibilities and these are 144
D is 144, and P is 1.45. The next is D is 144, and price has decrease it becomes 1.19 then
demand decreases it becomes 96, but price increases it becomes 1.45, and then finally you have
where demand also decreases, and price also decreases that is 1.19.
So, these are different alternatives which are emerging out of our tree. Now for the purpose of
calculation of profit at our state 1, period 1. This is state 1 period 1. Here our demand is 120
price is 1.32, and there are four alternatives as shown on the board. Now we will evaluate the
expected profit in the period 2 over all four states possible from this, so like this we have
calculated the profit for the first case where it is -33, 120.
446
Similarly, you need to calculate the profit for this state also, for this state also, for this state also,
so, that is what we have written here that 0.25 and for the situation of D =144, 1.19 D =96, 1.45,
and D =96, 1.19. So these all are the four different states and then the expected profit is
multiplied by the respective transition probability, that is 0.25. So that is the expected profit for
all these things. Now when you have the calculation, so these values are coming like that here it
is coming 4,320`
This value is coming -22,080 and this value is coming 2,880 and the probabilities are same 0.25
for all these different states, so the total expected profit is the sum of multiplying of these
probability with these profits. E.P.=∑(probability x profit). So, that comes to be -12000 that is
the expected profit at period 2 from D equals to 120 and P equals to1.32 that is state number 1in
period 1.
Now, the second thing is once we have calculated this expected profit of -12000 we need to
calculate the present value this the meaning is that this expected profit is in period2 this expected
profit you will get in period2. Now presently I am talking of period 1, so the present value of this
expected profit. So, I will add present value of expected profit this fraction and PVEP=EP/(1+K).
PVEP(D=120, P=1.32,1)= -$10,909
447
now the total expected profit for this node where demand is 120 price is 1.32 at node. At a state1
is the sum of profit in period 1 at this node + the present value of the future expected profits
possible from this node.
P(D=120, P=1.32,1)=[(120000X1.22)-(120000X1.32)]+ PVEP(D=120, P=1.32,1)=-$22,909
This is the total expected profit here is coming to be -22,909 which includes the profit generated
at this time + the present value of profits from all these future states, so that is taken get, and now
you already can understand the similar calculation will flow to period t0 also.
The same calculation will come to this state also, because this state is coming from this state
when the initial demand was 0 and P was 1.20, so now when I want to calculate the total
expected profit at t=0 I need to calculate the total expected profit for all the four states, because
this node is giving rise o four different nodes. There are four different nodes and in our last
session we already have discuss, what those four different nodes?.
So I will do the same level of calculation of determining the total expected profit for all these
four nodes different values will come, and then the expected profit for this particular node will
be the multiplication of this expected profit into the transition probabilities that is 0.25 for each
of them, and then I will calculate the present value for period 0. So that becomes the present
value of expected profit.
And then I will also calculate the profit coming from this particular node in the original period,
and that is the total expected profit for this particular option when I am going for 100% a spot
market. So let us see that what type of table emerges, because it is a quiet lengthy calculation and
therefore it is not very much advisable to show the complete calculation of developing the table
on the board.
448
So, let us see on this slides that how the table emerges, so for period 0 the starting period the
total profit for the original state that is 100 units is the demand, and the initial price was 1.20 is
EP[D=100,p=1.20,0]=
0.25[P(D=120,p=1.32,1)+ P(D=120,p=1.08,1)+ P(D=96,p=1.32,1)+ P(D=96,p=1.08,1)]
=0.25[-22,909+32,073-15,273+21,382]
=$3818
that is the expected profit during time period 1 but for time period 0 I need to have the present
value of this expected profit.
PVEP (D=100,p=1.20,0) =$3818/1.1=$3471
So, what does it mean that this is one part, this is only one part that this is the present value of
my all future cash inflows at time t=0. But, I also need to add into this 3471. The value of profit
or loss, which I am incurring in this particular period.
449
So that is done in the next slide where this 3,471 this component is there from the future cash
flows, and the present value of those future cash flow is 3,471 and the present value of my profit
is in the current period. I am selling 1.22 is the revenue. 1.20 is the original cost. So, it becomes
2000 rupees is the profit 2000 dollars, so 2000+3,471, so the total amount of profit, which I am
going to get is 5471. So, therefore the expected net present value of using 100% a spot market is
5471. This is the final calculation available to me.
Now, using the same approach for lease arrangement. So, then when lease arrangement is there
in this case it is 100% spot market. Now, I want to go for the lease arrangement, thus I want to
try for the second alternative.
450
And when I am going for the second alternative my net present value is coming 38,364. So that
is the total profit which I am going to get in case of lease arrangement, If I do a lease agreement
for 100,000 units initially and if I ignore the uncertainty so for this type of issue if you remember
in our 3, 4 sessions before the net present value was coming to the 60,182. Let me take you to
those slides, where we ignore the uncertainties, and we only consider the certainties and only on
the basis of time value of money. We determine this case, so when we took that data into
account.
451
Now in this particular case, there was low uncertainty initially rate is 1.22, and the rate of lease
was 1 dollar per square feet, and the profit was, annual profit was 22,000 and in this particular
case the uncertainty was not there. So we were having a very fix type of cash flows and, this case
gave us the NPV equals to 60,182 dollars. So, because uncertainty is there, so certainly, because
of these type of fluctuations in my pattern.
In my future obviously I have to sacrifice on my profit because, I am not very sure, that what
may happen, so this 60,000 reduce to 38,364, when some kind of uncertainty I have built into my
situation. But considering the situation of earlier case when I am going for the spot market and in
that spot market my profit was coming just 5471. So, that is much less as compare to 38,364
which is much, much higher.
So, even though uncertainty is there, but now we can see on this calculation basis, no certainty
obviously, there was a clear choice that I will go by lease agreement. But even with the
uncertainty the kind of data which is there, and kind of this calculation is showing to me that I
should go with lease arrangement, and the size of the lease is 100,000 square feet so this
arrangement is very well suitable because of higher profits as compare to the spot market. The
third option now, which we need to evaluate, that is flexibility.
452
Now we want to actually in this particular case the data was supporting the lease arrangement,
but it is quite possible that in some cases lease arrangement may not be that much useful. So, we
want to have a combination of both these things. So, we want to have some kind of flexibility in
this particular case. And when we want to have flexibility, so you can see that the condition
which we are developing is so how we can use this decision tree method for evaluating the
flexibility in the supply chain.
Because the whole predictive analytics, the whole real time data uses is to have flexibility in
your supply chain. If you do not have flexibility in your supply chain the issue of real time data
uses will not going to give you any kind of benefit. So, flexibility is a kind of you can say pre
requisite for adopting predictive analytics, for adopting the real time decision making.
If your supply chain, if your ware house is a lease arrangement which is a specified space, so,
there is no point in getting this real time information, because in daily basis, or monthly basis or
an annual basis you are not able to change the available space. So, therefore the point is that in
this particular case. We need to see that we can use the available space on the flexible basis.
And then only the usage is possible, because why I am saying that .if you go to this particular
example, and when you develop this type of calculation for the lease arrangement, because I am
taking 100,000 area for my lease arrangement. So, in that case when my demand is 144, when
my demand is 120.These types of additional requirements I may not be able to fulfil, because,
my area is only 100,000.
So, for 100,000 I will be fulfilling from my warehouse. The additional demand, I will not able to
fulfil, because I don’t have space. The second issue is when I have the less demand 96, when I
have less demand in that case also I will not be able to utilise my capacity to the optimum level. I
have unused capacity I am giving rent for no reason
I have wasted my money because, my company is having a system of fulfilling all the demand.
So, what will happen that 100,000. I will fulfil from my warehouse. And, the remaining 20, for
that I will go to the spot market. Here, 100 units I will fulfil from my warehouse. And, from the
453
remaining 44 I will go the warehouse, so this type of arrangement when I am going for the lease
kind of system.
So, this arrangement I have to follow, that a fix amount I will use from my warehouse, and the
remaining I will use from the spot market. So, therefore flexibility is advocated, where in some
of the cases, when you go to the continuously downward trend. Like 100 is here, and if it is just
80% of the original demand. So, it will comes 80, and then further in the next period it reduces to
it is 80% to so it is only 64.
So, in that case you have 36,000 square feet of area which is unused. So, that is also a loss, so
therefore we want to have flexibility in my supply chain, which can give me the benefit of both
these option. So, we stop here in this session and in our next session we will see that how
flexibility can be incorporated in uncertain environment. Because, that is the most important part
where, predictive analytics can help us and we can minimize the cost of offering and it will help
us in increasing our revenue. Thank you very much.
454
Lecture-39
Modelling Flexibility in Supply Chain
Welcome back till last few sessions we discussed about the uncertainty in the supply chain
and we try to understand the modelling related to uncertainty in the supply chain with the
help of a question where we want to discuss that where should we develop our facilities and
we took an example where we have a particular demand and there is a particular price of the
commodity and over the period of time demand can also fluctuate and price can also
fluctuate.
And we had two options either we can go for a lease contract or we can go for the spot
market, with the help of decision tree analysis we developed a decision tree and we saw that
how can we evaluate alternatives to get the best option, now this is all about predictive
analytics that we want to predict about the future as accurate as possible and obviously given
a variety of alternatives I would like to select one alternative which maximizes my expected
profit.
Now flexibility is one such area which we need to incorporate in our supply chain for
handling these uncertainties because earlier there was a time we used to have almost lot of
certainty with respect to various underlying factors of the supply chain. But as we all see we
are seeing so many things which are uncertain and all of a sudden these things makeup and to
handle those things flexibility is required in the supply chain.
455
In our earlier class also we had a little bit of discussion about the flexibility in the supply
chain that we need to have adaptable supply chains, we need to have aligned supply chain
and we also need to have agility in our supply chain, where our supply chain can respond
quickly in short time to the changes of demand and supply. Our supply chain should be
creating such type of incentives along the partners within the supply chain for better overall
performance.
If in a supply chain environment when we have a vendor, manufacturer, wholesaler, retailer,

these different entities are there but normally we all know that if one OM is there in a supply
chain so normally OM designs this supply chain such a way that he takes the maximum
amount of profit, so now when we are talking of alignment the supply chain should be
developed in such a way that the equal share or the proportional share of all the partners of
the supply chain can be taken care of.
Then adaptable where my supply chain can adapt to the design to meet the structural shift in
market, it can modify supply network strategies product, technologies etc. We had the case of
Dell where earlier Dell was supplying only through online order and that was one of the USB
of Dell, but over a period of time when Dell realise that now it is no longer profitable to
supply through that online mode.
So Dell also started brick and mortar system the retail system of distribution of the products,
so you need to be adaptable as per the changing requirements otherwise you will be out of the
market. So these are the three important elements of the flexibility.
456
Now we will continue with our example of this ABC logistics which we discussed in detail in
our last session and from this point we will like to see that how the flexible options are better
than 2 dedicated options which we have discussed, one dedicated option if you remember we
discussed where we had all warehousing space from the spot market and then other dedicated
option was to sign a 3 year lease for a fixed amount of time and there were two worthy to
dedicated options.
Now we want to incorporate flexibility into this discussion and flexibility that I want to use I
have this 100000 square feet area which I can go for a lease of 3 years or I do not take this
area I take my spot market option, so these worthy first two cases. So the third option is to
457
sign a flexible lease deal where we pay some minimum charge of front which makes us used
to be for this 100000 square feet.
But there will be a kind of variable arrangement that if I don't want the complete space I pay
only for a limited space and then remaining space can also be paid on some kind of additional
rental. So here in earlier cases if you remember if my demand is not 10000 units the in that
case I was having some space which is unused and still I was paying the rent for that if I was
having the lease agreement. A spot market I always need to buy space on a higher price from
the market.
So there were limitations of lease agreement as well as of the spot market. So in this case in
this flexible case I want to take the benefit of spot market as well as flexibility or as well as
the low cost of the lease system. So now we will see that how are we going to model such
type of situation.
We already discussed in our earlier session that how did we develop this decision tree, so
same just for a quick re-capsulation this is the initial demand of the product that is 100 and
these are the rate of my spot market that is 1.20 square feet per for the spot market and we
have discussed that there are possibilities of increasing the demand or decreasing the demand
by 20%.
So in the next period demand can increase by 20% then it becomes 120 and demand can
decrease by 20% then it becomes 80, 80 similarly price can increase by 10% for the spot
458
market or price can decrease also by 10% when it increases it become 1.32, when it decrease
it becomes 1.08 so therefore these 4 states are possible in period 1. Now from these 4 states in
period 1each of these states will further give rise to 4 different states in period 2.
There will be many common states also therefore we do not have 4 x 4 i.e; distinct states in
period 2, like you see for this second state here where the demand is 144 and price is 1.19, so
you get this arrow here and this arrow also here. So this state is common for two possible
outcomes from state one and state two of period one.
So it is possible like in this case we demand is 96 and price is 1.19 dollars, you are seeing 4
arrows coming to this, this state is 1 states which is possible from all 4 states of period one.
So therefore we do not have 16 distinct states in period 2, we have only some 9 different
states in period 2. Now after and same logic applies to this period also that demand 120 can
increase or decrease by 20%.
So if it increases by 20% it becomes 144 if it decreases by 20% it becomes 96 and 96. The
price 1.32 can increase by 10% can decrease by 10%, so when it increases by 10% ,1.32
becomes 1.45 price in period 2 and when price decreases by 10%, 1.32 it becomes 1.19, so
these are the two states where price decreases and these are the 2 states where demand
increases.
Now the possibility the probability which we are taking that is 50% for increase or decrease
of the demand similarly 50% is the probability of increase or decrease of the price of the spot
market so and these things are independent the increase of demand or decrease of demand,
increase of price or decrease of demands, so both these are independent.
So therefore the transition probabilities from this state to this state or from this is state to any
of these 4 states will be the multiplication of two different things, 0.5 is the probability of
increase of the demand, 0.5 is the probability of increase of the price, so 0.5 x 0.5 becomes.
25, so therefore these are the probabilities 0.25 ,0.25, 0.25 and similarly these are
probabilities 0.25 0.25 0.25 and 0.25, so this discussion tree and this discuss we already had
in our previous session.
459
Now in the previous we evaluated the two possibilities one possibility when we have the
lease system and what is that and one possibility when we are going to have the spot market,
so these 2 things we have already evaluated and we saw that lease system and this is the
possibility of the profit expected profit when we are going to have the lease for 3 years that is
38364 and the possibility of profit from the spot market the expected profit from the spot
market is 5,471 dollars.
If I compare between spot market and lease I will obviously go for the lease because this is
giving me much higher profit. Now I will like to evaluate the third type of alternative which
is the subject matter of discussion today that is the flexibility in the supply chain decision.
Now I want to incorporate this flexibility in this discussion and in the flexibility you see we
will use the same decision tree which I just explained you in detail.
460
Now the issue is this that the manager at ABC logistics has been offered a contract where
foreign upfront payment of 10,000 dollar, I paid 10,000 dollar today that is that will make me
you can say a kind of tentatively I can hold this space with 10,000 dollars, so this is the initial
payment I am going to make that is 10,000 dollar, the initial payment and now the company
has the flexibility of using between 60000 square feet to 100000 square feet of warehouse
space at the rate of 1 dollar per square feet per year.
So I can use from 60000 to 100,000 I can use, but you can understand from this language that
I need to pay dollar 60000 for the first 60000, I need to pay first for this is 60,000 square feet
dollar 60000 that I need to pay whether I am using it or not, so that way you can say this
fixed component in this flexible case is this 60000 square feet. The remaining if I use more
than that if my requirement is 70000, 80000, 90000.
So I need not to pay for this 100000, so the remaining space which I use for that additional at
the rate of dollar 1 per square feet per year I will pay, so this is that 60000 for that purpose I
am paying 60,000 dollars and the remaining up to 40000 square feet on demand as per my
demand I will pay for dollar 1 per square feet per year. Now doing the calculation as we did
in the case of any of the earlier methods.
So we go back to this decision tree and will like to see that how this method is applicable.
When my demand is let say I take two scenarios to make you understand that how you
calculation will proceed. So we will start from period 2 and in period 2 let us take two, three
461
different scenarios one scenario where the demand is 144 at spot market prices 1.45 demand,
when demand is 144 units, price is 1.45 and this I am talking in period 2.
So when this is the data, so out of this 144000 units of demand I will like to fulfill my
100,000 demand from the lease agreement, so for that purpose my cost is
(100,000x$1)+$10,000+40,000x$1.45
100,000x$1 is the cost because of the lease agreement and this is the cost because of the spot
market, so 40,000x$1.45 is one scenario of the cost and obviously because you are selling
144000 units, so revenue will be 144000 and you can remember the data that we are selling at
the rate of 1.22, so that is the rate at which we are selling.
So, so revenue - cost will give you the profit for period number 2, now it is up to us that we
can eliminate this 10000 right now in this calculations and once we come with the final
calculation in that also we can add that whatever profit we are getting we will subtract 10000
in that, so that will also be possible.
So instead of adding 10000 here we will add 10000 in the last of our cost components and
from the total profit we will subtract 10000 for getting this initial payment of the lease
agreement, so this is your now profit will be revenue - cost that is the profit for period
number 2. When I come to period number 1 with this data I will calculate the present value of
the future cash by dividing be this value of profit by your factor of discount that we
remember is the 10%.
462
Not come to a different situation where D =64 and the spot market price is 1.45, since again
we have this initial arrangement and in the case the cost component is 60000 for that I am
already paying at the rate of 1 per square feet and I am also taking 4000 additional space.
So for 64000 the total cost will be at the rate of 1, cost =64000x$1 and I will not go to the
spot market because I have total availability of space 10000 square feet and requirement is
64000 square feet, so this price of 1.45 is not applicable right now in this calculation, this
price has no meaning I have zero spot market cost to this spot market cost in this case of 64,
1.45 is no reason it is not applicable to me.
So I will only incur this cost and then the revenue is 64000 x 1.22 and then I also calculate
the profit that is R – C. Similarly if I go to this calculation d = 96 and spot market price is
1.19. Now this 96 is also less than 100000 and therefore again there is no use of this spot
market price, because my warehouse for which I am making a flexible arrangement can
accommodate up to 100,000 units.
Because the size is 100000, so for 96000, 60000 at the rate of 1 and additional 36000 again at
the rate of 1, so for 96000 my cost will be 96000 square feet I will take at the rate of 1 dollar,
so that is my total cost is 96000x$1 and revenue will be simply 96000 x 1.22. So that is
against I can calculate the profit. So here by using this flexible system I will not incur any
additional cost because of unused capacity.
Now you see in these two cases 64 and 96 where I am not able to use entire 100000 square
feet of the warehouse, if I would have gone for completely lease arrangement, so what would
have happened that I will have some unused capacity of 4000, I will have a unused capacity
of 36000 and I am paying rent for that, so I am saving that amount of rent and therefore this
is much more profitable.
And then with this calculation you I request all my students that please make this calculation
complete and when you make this calculation complete you will see that the total amount of
profit is coming to be 46,545 dollars and in this calculation of profit I have already subtracted
the 10,000 dollars which we had made as upfront payment.
463
So that amount we have already subtracted, so now if you compare be 46000 we are getting
here, the lease agreement was giving us 38364 and spot market was giving us 5471, so
obviously out of this 3 alternatives I will like to select this flexible arrangement, because this
is giving me maximum expected profit. Now because here we have only two conditions and
price is either increasing or decreasing by 50%.
And demand is also increasing or decreasing by 50%, so you have only 2, 2 situation but as
the amount of uncertainty increases as flexible options give you more profitable or more
value in the answers, so when your uncertainties are higher it is always advisable to go for
more flexible options. Because in that case the dedicated options may go unused if you don't
have enough demand, enough supply.
So in that case the dedicated options will give you higher fixed cost of your plant and
machinery etc. So therefore be flexible option is very very useful and you can see that in this
particular example where we had only two options that how incorporating a very little bit of
flexibility, now you can do one more thing and you can take this as part of your assignment
also or as a practice question that for first 60000 with prices at 1 dollar per square feet.
So for 60000 you are paying 60,000 dollars, now for the remaining space up to 40000 the
price is increased by 10 cents, it is 1.1 dollar for any additional space after 60000 and I
request to do a revision of the entire calculation and in that again you will see that what type
of changes are there. Because this is a very simple case we took to understand this
phenomena.
But you can see that how increased cost of the because alternative for different, so you can
have the additional space at a slightly higher rate than this dollar 1 per square feet, so you can
try that option also. Now going further into the discussion we will see that you may like to
establish these two types of facilities.
464
So it is a different type of situation where you can have dedicated capacity of 100000 units in
Gujarat and similar dedicated capacity of 50000 in national capital region, then you have
another issue that you can have a flexible capacity of 100,000 units in Gujarat and 50000
units in NCR and again we are going to evaluated for 3 period from 0, 1 and 2.
So let us see that what is the difference between dedicated and flexible capacity. In these two
cases when we want to make any product and I am taking Gujarat and NCR it will be more
appropriate if you take two different countries and in that case you can incorporate the issues
related to exchange rates also.
So you can try that you replace Gujarat by India and taking place of NCR you take country
like Vietnam so it can be India and Vietnam also, so that will make this discussion more
interesting and not only the demand and price the other charges rate fluctuations will also
come into picture.
465
So the data available to us is like that for investment for the dedicated plant and for the
flexible plant, the size of the plant is 100,000 and 500000 respectively the fixed cost for the
dedicated plant is one million per year at Gujarat and NCR because the land prices are very
high it is 4 million per year. So we are making tyre, so the variable cost is 15 dollar here and
it is 11 here.
And similarly for the flexible plant it is 1.1 million years and 4.4 million per year and
variable cost is same it is as for the fix a dedicated plan that is 15 dollar and 11 dollars
respectively. Now the expected demand related to which Gujarat plant can fulfill is 100,000
initially and the demand from the NCR plan to meet is 50000 and there is a probability that
demand can increase or decrease by 20% over a period of time per period with a probability.
And let us say since we are considering only one probability of 20% of increasing or
decreasing the demand, so 50% up and 50% down let us go with this simple assumption.
466
Now when we consider this and these are the two plants of 100, 50 and to take this example
on the international level I have given a value of exchange rate also, so E stand for the
exchange rate and there are these possible states which way emerge from period zero to one
and one to two and I request my students that please try to write the various transitive
probabilities on these arrows.
So you do not see this graph I am not putting that slide for long and you can develop this
decision tree on your own and once you have this decision tree available with you then you
can have these possible scenarios that at both the places you have dedicated plant at both the
places you have a flexible plant, you have a flexible plant and Gujarat and dedicated plant at
NCR.
467
And this becomes a question for you that you can see that how with different plants you can
serve with different markets like this Gujarat plant will serve the Gujarat related area and
NCR related area also NCR plant conserve the Gujarat area also and NCR related area also so
you can have this types of scenarios possible and it will be very interesting if you develop the
entire decision tree.
And I expect that you develop this decision tree on a spreadsheet and then the calculation will
be much easier and finally you can have this type of net present value I am giving you the
final slide where if you have dedicated plant at both locations at Gujarat and NCR, the net
present value is this figure then flexible plant at Gujarat and dedicated plant at NCR, so out of
all these net present value calculation you find that the highest net present value is available
for the combination of dedicated plant at Gujarat and the flexible plant at NCR.
So this way you could find that which is the most profitable configuration of our plants and
this is what we mean by incorporating flexibility in supply chain. So whenever uncertainty is
there and when uncertainty is higher we need to include flexibility and as you go uncertainty
is highest as in earlier case we considered only two parameters, one parameter was the
demand and another parameter was the price of the spot market.
In this example we took three aspects we consider 2 markets their demands and second was
the exchange rate fluctuations also if it is a global kind of supply chain, so here we had three
different type of parameters and therefore more uncertainty is there. So as you increase the
number of underlying parameters in the decision tree you will have more states, more states
468
means more uncertainty and therefore the flexibility in the supply chain will give you the
better answers.
So with this simple tool of decision tree and developing the payoff table and on the basis of
that you calculate the net present value and finally we can see that out of various alternatives
which alternative is the most profitable one, means which alternative is giving you the
highest net present value of the future cash flows, so accordingly you can take the decision.
Thank you very much and with this we complete the discussion of uncertainty in the supply
chain design.
469
Lecture-40
Trends, Challenges and Future of Supply Chain
Welcome back, dear participants of this course so now we are coming to the end of this
course and in the final session of this course we will like to summarise what we have
discussed in earlier various sessions and what will be the future direction of supply chain,
what are the trends in the supply chain?, what are the challenges in this supply chain?, so all
these things we are going to discuss in the session. So that this course is not going to make
you an expert of the supply chain.
Rather this course is an attempt to create a kind of foundation about the various types of
supply chain decisions, it can sensitize you about the challenges of the supply chain, it
sensitize you about the requirement of supply chain for better decision making, it can
sensitize you about the environment in which a modern manager makes the decision for the
supply chain.
So this session will tell you about all those various aspects which are there which make
supply chain more interesting which makes supply chain as one of the function of the
organisation which is you can say key for the profits for the value of the organisation.
470
So we start with this discussion that supply chain as we discussed in the very first class is to
ensure the availability of the products to the customer where customer wants, what customer
wants, how much customer wants to give the answer of all these things, the discipline of
supply chain came into existence. Over a period of time when competition started growing
and we discussed the various phases of evolution of supply chain.
We discuss this phase where ford motor was there one side and ford motor was having a very
integrated supply chain, every types of supply chain where from mining of iron ore to the
distribution of made care was controlled by only one company that was ford, so that was one
extreme of supply chain, so initially the discipline started from that type of highly integrated
supply chain.
And over a period of time with the Japanese system of management where Toyota created a
very much disintegrated model, created so many vendors in the supply chain, that I am not
going to meet the all the components for car, I will procure components from different
vendors and that model became very popular and nowadays we all know that in a supply
chain of 1 OEM there are large number of vendors.
And you have so many distributors also in a supply chain and you guide them you ask them
to behave as per the company's policy but you do not own them, so you created it
decentralized model so the earlier model of ford was highly centralised, then this Toyota
model of supply chain is highly decentralized, so the whole world moved from that integrated
model to this disintegrated model.
Then over a period of time with the help of companies coming all across the world when we
started facing the heat of competition same came to supply chain also and this is the third
revolution which came from the ideas of Dell company, Dell power customer to design their
own products until late 2006 customers were giving orders online to Dell and Dell was trying
to give products as per the unique orders given by the customer.
So that was another important evolution in the concept of supply chain and the further
evolution of that we all see in present time is in the form of e-commerce where customers are
not visiting the brick and mortar stores they go to E-portals, they go to various types of E-
471
suppliers, e-commerce site we say in day today language and these E-Commerce sites are the
new retailers of the model era.
And all these changes which I just discuss are in last 100 years, so in these 100 years the
subject of supply chain has become such a vital issue that many companies, the companies
like Amazon, companies like Walmart, companies like Maruti in India. These are highly
successful because of their very very adaptable, very very efficient, very very effective type
of supply chains, but in all the supply chain discussions our focus is entirely as we discussed
in the first or second session to increase the supply chain supplies.
We want to develop supply chain which can increase the surplus of all the partners of the
supply chain, it is very interesting in the supply chain that we do not talk of profit of
individuals, we never talk of profit of individuals and I again give you a warning message
that supply chain is highly collaborative activity, so you should always think in a matter of
collaboration.
We never talk of individuals profit, individuals forecasting, individuals inventory, all these
terms are prohibited in our supply chain discussion, so we only talk in terms of how you can
increase the profit of entire supply chain, how can you reduce the inventory of entire supply
chain, how can you make forecasting more effective for the whole supply chain, so all the
discussions are happening for the entire supply chain.
So that is very very important in case of supply chain, but in all these discussions when we
are talking of continuous increase of surplus of supply chain and when surplus is increasing
we say that our supply chain is performing very well. But in this race of supply chain surplus
we have ignored this first. Which I mentioned here that is the human aspect. Human aspect is
very very important and without human aspect our business cannot survive.
So the first important challenge in front of supply chain is to address human related issues
also and therefore the extension of this human concept human aspect the word has come
where we talk of humanitarian surprises, a new type of supply chain discussion has started
that is humanitarian supply chain many of us will take the discussion of humanitarian supply
chain into the disaster supply chain that some accident has happened, some natural calamity
472
is there and to fulfill be requirement to provide relief rehabilitation kind of support to the
victims of those calamities.
We need to provide relief material, so that is human aspect of supply chain, but it is much
more than that when I am producing products like mobile phone and we all are consumers of
mobile phone it is almost impossible now days to think anybody without mobile phone and
when all of us are using mobile phones. So from where the raw material is coming, raw
material is coming from some of the African countries.
And what is the condition of those people who are providing raw material for our mobile
phones, so if you go to YouTube and see those videos of children getting into the mining
activities for getting the raw material for your mobile phone it is very difficult to use mobile
phone after that because those pictures will be haunting in your mind. So the human aspect is
also very very important that how much are we consuming?.
What are we consuming?, how do you think are given to us?, what is the process of giving
those thing to us, so all these are the subject matter of human aspects of supply chain, we
become many times in human in race of our supply chain surplus, so the first important point
which supply chain must address that we should not ignore humanitarian. We should not
ignore humidity for the sake of supply chain surplus.
The second important future trend which is there because of development of computer
science, various things in supply chain because if you are located very closely around your
customers and your vendors, so without even internet, without computer science you can
manage your business as long as your business is of low size. But as your business grows
your vendors become international.
So you need some kind of technology help and thanks to internet and computer science and
thanks to web technology that because of rapid development in the field of IT, computers and
Web Technology we are able to use those technology in the development of supply chain
management. So that is important area the whole discussion in last so many sessions is only
possible when we have good computers with us, without computers it is not possible to apply
those types of algorithms.
473
You are not going to make long pay of tables using your calculator on register, it is only
possible when you have some kind of spreadsheet program, without spreadsheet it is not
possible to develop such type of tables. So the computer is very very important, particularly
the level of automation in Indian supply chain is not very high, it is not very promising as the
supply chains of our developed nations are using.
So there is a challenge, there is a potential opportunity also. Both these things are there, that
we need to increase the uses of computers, internet in our supply chain, so that because we
require for the analytics purpose information in real time. And that information availability in
real time is only possible when you get information instantly from variety of sources and
probably the increased use of internet will help us, will facilitate to achieve this.
So that is a very important things that we need to incorporate more use of internet in our
supply chain. And nowadays many of you may be aware that it is not simply IT in the supply
chain. We are moving one step ahead it is IOT Internet of things, which is coming in a big
way in the supply chain. In India not much research has taken place in the field of IOT, so
those who are interested in a kind of research career they can think of that how IOT can be
used in supply chain.
There are certain journals in which some papers are also there about IOT and supply chain.
Now many of you know what is IOT when I use heterogeneous devices over internet that is a
IOT, so if I use I can sync my smart phone with my colour television, so it is a kind of IOT
that these two devices are not compatible with each other, but because of Internet of things I
can connect these two devices and I can command from one device to others.
Similarly in my supply chain if I start using IOT, so at each stage from procurement of raw
material, the inspection process of raw materials to the manufacturing, to the warehouse, to
the distribution to the consumer at different stages I can use IOT and this IOT uses will
certainly increase it is expected the efficiency and responsiveness of the supply chain.
Because these are the two most important things in a supply chain, efficiency and
responsiveness I want things quickly, and I want things at low cost. So when I want things at
low cost it means efficiency and when I want things quickly it means responsiveness, so as a
matter of fact we all are consumers and we want both these things whenever we purchase a
474
product, we want that the product should be available to us at low cost and simultaneously it
should be available within no time.
And therefore concepts like uses of IOT may help us in achieving these objective because
you can easily monitor the movement of your products in the entire supply chain, you can
control the movement of the supply chain, you can not only simply monitor, because
monitoring you are doing by using the RFID tags, barcode tags, and that monitoring was
already happening.
You know that when you are doing a speed post and that speed post envelope bears a
particular type of barcode and you have that consignment number available with you and if
you login to the Government of India department of post site and you put that consignment
number, so you can easily track the location the movement of the product, but with IOT what
is going to happen?, with IOT I can monitor, not only monitor I can also direct the movement
of the consignment that where it should go?, at what time it should be go.
At what time the customer will be there, so the product should reach to the customer when
the customer is there, we all know many time the courier wala, the Postman come to our
house but we are not there and then the return without delivering the packets, but with IOT it
is possible that I can connect all these things and I will direct that delivery boy sitting in my
office that you deliver this product at a particular time or the device will give signal that the
receiver is available and the product should be delivered at this time even I do not required to
be dictating it from my office.
The device itself will communicate that yes the receiver is there and the packet can be
delivered at a particular time. So these type of changes are possible with the use of IOT, so it
is a big thing for the future and the supply chain must stay in India particularly if I am talking
that be Indian supply chains must think of adopting tools and techniques which can make
IOT part of their supply chains.
Normally at the present time it is believed that only those players where the high unit value is
there of the product, they are interested in application of IOT. But I am very sure that with
more research and development in this field this IOT will be available to other players also
475
where the unit value is not very high. But another important issue in the supply chain is
environment and this is really a major threat to the Global world.
Now what is happening let us take a simple example because of globalisation your vendor is
in Africa, I am making product in India and supplying these products to America and this is
what we say we must take the advantage of globalisation, procurement from those countries
where you get the very cheap raw material, produced where you have the availability of the
skilled labour at the low cost and sell where you get the maximum value for your product.
So this logic is very simple and we understand this logic perfectly, but what is happening
because of this you are transporting the material first from Africa to India, so lot of carbon
emission takes place when the product is coming from Africa to India, then finished products
from India is going to America again lot of carbon emission will be there in transporting
products from India to America.
So now this is something which is against the environmental ethics we need to see that how
we take the advantage of all these low cost resources, how we make our supply chains more
efficient but for the sake of making them efficient we should not compromise with the
environmental issues. We have so much gifts from nature in each part of this world that if we
try we can sustain our life from the local resources.
But saying this is against the principle of globalisation. So we need to make some kind of
balance between these two things that you should be able to take the advantage of
globalisation and at the same time you should create some kind of balance with respect to
your environment issues. If this balance is lost then there are going to be enormous
challenges.
If our survival is in question then whatever amount of supply chain surplus I have there is no
use of that. So we need to see that what is the direction of growth and whether the direction
of growth is sustainable or not and that is certainly my next future issue that is the
sustainability. We are looking for growth, we are looking for hire supply chain surplus but
whether that whole process is sustainable whether it is long enough to survive.
476
Sustainability means a thing which can sustain for infinite amount of time, so whether this
whole process is actually sustainable or not. Now that sustainability will depend on various
issues like if my supply chain in our last session when we were discussing about the
flexibility in the supply chains one point came that is the alignment. Now my supply chain
should be aligned properly, it means the supply chain should take care be interest of all the
partners in right proportion.
If supply chain is bailed towards some of the partner and they take more by out of the overall
supply chain surplus or disproportionate by out of the total supply chain surplus that this is
the misalignment and if something is misaligned it is not going to be sustainable. So what I
am trying to say that if in a supply chain there are extra benefits for some of the stakeholders
at the cost of other.
These businesses, these supply chains are not sustainable. So we need to see that how supply
chains are to be sustainable, if it is part of a society, so the society must feel that this supply
chain is taking care of the society. There are male members, there are female members, my
supply chain should not differentiate between male and female. There are people of different
age group, there are people of different religious background, different language etc.
So I need to see that my supply chain is making a justice with people of all the background
all the ethnicity, all the races etc. My supply chain is taking care of environment, it is giving
the due recognition to the environment and my supply chain is also viable, it is also
producing sufficient amount of profit. So that it is attractive enough to be in that supply
chain, if my supply chain is not making a decent profit I will move away from that business.
And in that case I may think of some other business, in your area as you will see some
people, some shopkeepers they keep shifting the agencies, they keep shifting the business
interest and this happens because the supply chain in which they are dealing is not lucrative
enough to attract these people. So that economic sustainability is also required, but the most
important thing is that we need a proper balance between these three types of sustainability.
You have economic sustainability, you have social sustainability, you have environmental
sustainability no economic sustainability at the cost of social and environmental sustainability
not environmental sustainability at the cost of economic and social sustainability can never
477
happened social sustainability at the cost of economic and environmental sustainability again
not desirable. So you need to create again a proper balance between these three types of
sustainability.
So that you have a long term plan of sustainability and this is again a very important future
challenges for the supply chains that how to make a balance if you want to develop supply
chain which is only for one type of sustainability that is very easy task, if I only thing of
profit and profit all the time my economic sustainability I can develop the time of supply
chain.
But when I talk of developing a supply chain which takes care of all three types of
sustainability that is a challenge. So that is another future issue which we need to handle in
our supply chain that how to maintain balance between three different types of sustainability
and then one of the most important thing and lot of research is happening in this area also that
is risk management.
Now risk management is something like handling the uncertainty of the supply chain, our last
few session we heard about handling the uncertainly in the supply chain, so risk has 2
dimensions. There are risk which can be quantified and there are risk which cannot be
quantified, qualitative type of risk are also there. So we need to understand that what are the
quantitative risk and what are the qualitative risk in my supply chain.
For quantitative risk we can take care with the help of different types of risk management
softwares, we can use different type of risk management techniques particularly what we
discussed in the financial management, so those risk management techniques can be adopted.
Sometime you accept the risk, sometime you avoid the risk, sometime you mitigate the risk.
So there are different types of risk management strategies available.
And then there is a qualitative risk also which is difficult to quantify and it is our
foresightedness, it is foresightedness of the supply chain manager to understand to predict, to
visualise the possible risk in the supply chain and these possible risk in the supply chain will
help us to develop a robot supply chain. Otherwise we have done the sensitivity analysis also.
478
If I am not having a good risk management incorporated in my supply chain my supply chain
will debate with little bit changes in my input parameters. So it is important that I develop a
supply chain which is robust enough and we should not change, we should not deviate which
should not fluctuate much with the slight changes in the input parameters, though I need
flexibility in supply chain.
But if I take a supply chain where it is not robust enough then lot of fluctuations will happen
and in that case I may need to redesign my supply chains time and again. So therefore risk
management is also very important concept in the future supply chains where we take lot of
inputs from the financial management backgrounds to handle the risk management issues.
Since it is highly uncertain environment and daily new challenges may emerge.
So I cannot say nobody can say that we have a comprehensive list of issues for future supply
chain. There will be new challenges coming day by day, sometime you have daily new
regulatory environment, so when regulatory environment changes these environment will
also give you new type of challenges, when you see a new regulation comes from
Government of India, the new regulation make create demand for a new product admit totally
finish the market of some of the existing product.
So you don't know government of Bihar in India banned liquor so as a result of that the
demand of liquor in Bihar decreased it finished. So the regulatory environment is also very
much uncertain and it can also affect our supply chain to a great extent. So with these things
we discuss the various issues which are important in designing a future supply chain and the
analytics will help us in describing the events which has happened descriptive analytics.
The predictive analytics will help us in predicting the future what is going to happen in the
future with this type of description and the third part it will help us in getting the optimum
solution, the descriptive type of analytics, so with using the operation research techniques
with using techniques like decision tree etc, which we discuss these are the part of that
descrptive analytics.
Where we are using that what is the most optimum solution giving these conditions, giving
these scenario, so I hope that we had a good learning time during the course, during these
various session and now my best wishes to all of you for a very successful use of whatever
479
we have discussed during these 40 sessions of supply chain course and I also wishing you
very best for your examination of this certificate course. Thank you very much for being in
this course.
480
THIS BOOK
IS NOT FOR
SALE
NOR COMMERCIAL USE
(044) 2257 5905/08

nptel.ac.in
swayam.gov.in

110107074

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

110107074

Uploaded by

Copyright:

Available Formats

INDEX

S. No Topic Page No.

(Refer Slide Time: 09:46)

It is integrating the development of supply chain management as a concept, as a theory, as a

Now whether I want to have a wholesaler in Maharashtra or I want to have a wholesaler in

(Refer Slide Time: 08:14)

So you have supplier, supplying products to manufacturer, manufacturer giving products to

If I am in the production department I use to have no communication with my HR

So my customer priorities, my ability to understand my customer priorities that is the first

So when I am talking of market intensive strategies or the market dominants strategy, I

There are centralized manufacturing, there are examples of decentralized manufacturing. So

I am continuously working on my retailers, so where to locate my new retailers, where to

Vice-versa is also possible today, I am operating in a very efficient manner in my outbound

We have from manufacturer to wholesaler, to retailer, to customer. Now whatever we are

Mostly in the case of transportation if I am taking help of somebody in transportation in

Now it is also very important to understand that to achieve a particular level of

Q1 is four times to Q2 because Q2 , Q1 is able to supplied four times to Q 2 out of it is originally

Nowadays because of influence of Japanese manufacturing systems in various domain of

So my holding cost because of increase in value, because of more value I am adding as I am

If mi is 0, 2 to the power 0, it is 1. Then 2 to the power 1, it is 2. 2 to the power 2 it is 4. 2 to the

So by doing some kind of mathematics in these quantities of Q 1, Q2 , Q3 can be minimise the

of inventory ci represented as . Now this formula can be changed in terms of

So whenever I am going to have a new facility so whether it is a manufacturing, whether it is

(Refer Slide Time: 21:51)

The formula for dn is written this G4 cell and that is

(Refer Slide Time: 02:42)

that you have

But now in this case it is

the values of yi can only take place 0 and 1.

So which combination of supply will go to the particular combination of destination, so we

So that is the optimisation of marketing campaigns. Next is improving operations, many

So if P is 100 so Pu is 105, if Pd is there it is 95, now since it is uncertain environment the

So when it increases it becomes P + u – d and when it decreases it becomes P – 2d. I am not

PVEP(D=120, P=1.32,1)= -$10,909

If in a supply chain environment when we have a vendor, manufacturer, wholesaler, retailer,

(044) 2257 5905/08

You might also like