Exercises

in Supply Chain Optimization and Simulation

using anyLogistix

Level 3 Advanced+

Prof. Dr. Dmitry Ivanov

Berlin School of Economics and Law

Professor of Supply Chain and Operations Management

To be cited as: Ivanov D. (2023). Exercises in Supply Chain Optimization and Simulation
using anyLogistix, Berlin School of Economics and Law, 5th Edition

© Prof. Dr. Dmitry Ivanov, 2023. All rights reserved.

0
Table of Contents
1. Introduction ......................................................................................................................... 2

2. Case study 1 ......................................................................................................................... 5

2.1 Description of Case Study ..................................................................................................... 5

2.2 Greenfield Analysis (GFA) for Facility Location Planning: Selecting the Best Warehouse
Location for Polarbear Bicycle .................................................................................................... 6

2.3 Network Optimization (NO) for Facility Location Planning: Comparing Polarbear’s Supply
Chain Design Alternatives ........................................................................................................... 7

2.4 Simulation: Dynamic analysis of the Polarbear’s supply chain design ............................... 11

2.5 Risk analysis ........................................................................................................................ 15

2.6 Comparison experiment ..................................................................................................... 15

2.7 Validation using Variation .................................................................................................. 16

2.8. Simulation with BOM......................................................................................................... 17

2.9 Recommendations .............................................................................................................. 17

3. Case study 2........................................................................................................................... 18

3.1. Description of Case Study .............................................................................................. 18

3.2 Problem Statement ............................................................................................................ 20

3.3 Greenfield Analysis (GFA) ................................................................................................... 24

3.4 Network Optimization (NO)................................................................................................ 25

3.5 Simulation ........................................................................................................................... 27

3.6. Risk analysis: Two-month disruption at one of the DCs .................................................... 29

3.7 Comparison experiment ..................................................................................................... 29

3.8 Validation using Variation .................................................................................................. 30

3.9 Recommendations .............................................................................................................. 31

3.10 References ........................................................................................................................ 31

1
1. Introduction
Supply chain network design and operational planning decisions can have a drastic impact on
the profitability and success of a company. Whether to have one warehouse or two, close a
factory or rent a new one, or to choose one network path over another are all consequential
decisions a supply chain (SC) manager must make. However, these decisions must be the result
of more than experience or intuition, and, as a result, research in SC management (SCM) is
geared towards providing the data, tools, and models necessary for supporting SC managers’
decisions. One of these decision-supporting tools is anyLogistix, a software which facilitates
Greenfield Analysis, Network Optimization, and Simulation.

anyLogistix has become more and more popular with the provision of the free PLE version, and
because it is an easy-to-use software, includes simulation and optimization, and covers all
standard teaching topics (center-of-gravity, efficient vs responsive SC design, SC design
through network optimization, inventory control simulation with safety stock computations,
sourcing (single vs. multiple) and shipment (LTL vs FTL) policy simulation, and milk-run op-
timization). This guide has been developed to support a course in Supply Chain Optimization
and Simulation using anyLogistix software using two sample project reports. The following
themes will be considered:

 Facility Location Planning (COG, Trade-off Efficiency vs Responsiveness)

 Supply Chain Design (Network optimization, CPLM)
 Inventory Control Policy (simulation, safety stocks, ordering policies)
 Sourcing Policy (simulation, single vs multiple sourcing)
 Shipment Policy (LTL/FTL, aggregation rules)

This collection of exercises is designed as an application add-on to the main ALX Handbook
which provides technical descriptions of how to build ALX models. The main ALX handbook
is available at www.anylogistix.com.

The ALX exercise book addresses the application of quantitative analysis methods and software
to decision-making in global supply chains and operations. Understanding of optimization and
simulation methods in SCM is the core of the course. Technical skills for using simulation and
optimization software in praxis can be acquired with the help of anyLogistix software. This
course is designed to stimulate and enhance conceptual and analytical decision-making skills
in actual operating situations, and will lean heavily on case studies developed throughout the
course. Class sessions will be spent mainly discussing the cases and software implementation,

2
where students will be required to evaluate actual business situations and apply their relevant
skills, experience, and judgment to develop viable resolutions to business problems using pro-
fessional software for decision-making support. Cases are drawn from both industry and service
environments, calling for decisions in different facets of supply chain and operations manage-
ment (SCOM). The case method requires you to prepare a decision based on careful evaluation
of case facts and numbers to the extent possible. As with all business situations, there may be
insufficient facts, ambiguous goals, and dynamic environments.

Upon completing the course, students should be able to do the following:

 Develop critical thinking skills, be able to identify, generalize, prioritize, isolate, and
reduce complexity in complex and ambiguous operational situations,
 Understand how strategic considerations influence operational decisions,
 Apply analysis and improvement tools learned in previous courses to actual business
situations,
 Strengthen qualitative and quantitative reasoning skills for operational decision-making.

This course seeks to convey the following skills:

Analytical Skills: Students will possess the analytical and critical thinking skills to evaluate
issues faced in business and professional careers.

Technical Skills: Students will possess the necessary technological skills to analyze problems,
develop solutions, and convey information using optimization and simulation software.

Along these lines, throughout the course we will examine two case studies:

 Building a new SC from scratch - a case study of the Polarbear Bicycle company, which
must create and optimize its SC in order to maintain profitability and keep its competitive
edge in an increasingly global market where sales prices are driven down while costs re-
main stable and
 Improving the existing SC - a case study of the beer producer BERLIN BREWERY, which
seeks to analyze the performance of their existing SC and optimize its distribution network,
while considering the risks and ripple effect.

Using the models available in anyLogistix, we will conduct analyses to (1) determine an optimal
location using Greenfield Analysis (GFA) for a new warehouse, given the location of their cur-
rent customers and those customers relative demands, (2) compare alternative network designs
using Network Optimization (NO), (3) perform a Simulation of different scenarios, (4) validate

3
the models using Validation, Comparison experiments, and (5) analyze SC behavior under un-
certainty using the Risk Analysis experiment.

The author thanks The AnyLogic Company for their invaluable feedback, comments and sce-
nario updates to this exercise book. The author also wishes to thank Ms. Meghan Stewart for a
thorough proof-reading. The author thanks students Meghan Stewart, Jannes Zuch, Chantal
Reimann, Moritz Albrecht, Stephanie Paeschke, Julia Dyck, Lily Creed, Christin Kemper,
Ragna Maria Berg in MA Program Global Supply Chain and Operations Management |
GSCOM at Berlin School of Economics and Law for case-study samples used in this exercise
book. Finally, the author thanks Mr. Nurlan Mammadzada, a former student in the GSCOM
master program as well as Mr. Hiran Prathapage, a research associate at HWR Berlin for the
updates of all exercises and educational materials.

4
2. Case study 1

2.1 Description of Case Study

We consider a company called Polarbear Bicycle. Polarbear Bicycle has been newly founded
as an e-commerce start-up selling bicycles. Polarbear’s portfolio includes four different types
of bicycles: x-cross, urban, all terrain, and tour bicycles. Polarbear needs to find a good location
for a new distribution center (DC). First, they estimate customer demand. Polarbear distributes
their bicycles to four locations throughout Germany: Cologne, Bremen, Frankfurt am Main,
and Stuttgart. Table 1 shows customer demand, which is equal to 245 bicycles per day.

Customer Bicycle Type Demand per day

Cologne x-cross 2
Cologne urban 50
Cologne all terrain 15
Cologne tour 10
Bremen x-cross 7
Bremen urban 30
Bremen all terrain 20
Bremen tour 20
Frankfurt am Main x-cross 6
Frankfurt am Main urban 5
Frankfurt am Main all terrain 4
Frankfurt am Main tour 5
Stuttgart x-cross 15
Stuttgart urban 15
Stuttgart all terrain 1
Stuttgart tour 40
Table 1. Customer demand

Polarbear Bicycle has hired a consulting firm to analyze supply and distribution network alter-
natives and to develop a best-case scenario for Polarbear Bicycle. They are charged with con-
ducting a GFA to determine the possible location of a new DC in Germany, as well as a network
optimization to compare several options for network paths. The consulting firm was also asked
to run a simulation to validate several KPIs and plan inventory, and to conduct a sensitivity
analysis to verify all results as well.

5
2.2 Greenfield Analysis (GFA) for Facility Location Planning: Selecting the Best
Warehouse Location for Polarbear Bicycle

Now we conduct a GFA for the outbound network of Polarbear Bicycle considering the four
customers located in Cologne, Bremen, Frankfurt am Main, and Stuttgart. The aim of this GFA
is to determine the optimal location of one new DC in Germany subject to total minimum trans-
portation costs.

Creating an ALX model

Step 1. Create scenario PB GFA Advanced+.

Step 2. Enter data from Table 1 into tables Customers, Demand, and Products. The data
in these tables should correspond to Table 1.

Performing experiments

Step 1. Go to GFA Experiment and run it for “Number of sites = 1” and the period of one
year. (select “all periods”). In order to activate the experiment settings, the settings button
should be clicked.

Step 2. Analyze the results using statistics “Flows” and “New Sites”:

a) What are the optimal coordinates of the DC?

b) What is the maximum distance from the optimal DC location to a customer?
c) What is the minimum distance from the optimal DC location to a customer?
d) What are the total costs of the SC? (Note: to compute the sum of costs or flows in GFA
Results, just slightly drag the heading of the column “Period” or “From” in table “Prod-
uct flows” in the space over the table).
e) Compare the data in statistics “Flows” and Table “Demand”. Do we satisfy all customer
demand from the optimal DC location?
f) What other costs were not considered in selecting the optimal facility location in the
GFA?

Step 3. Go to GFA Experiment and run it for “Number of sites = 2”.

Step 4. Analyze the results using statistics “Flows” and “New Sites”:

a) What are the total costs of the SC?

b) Compare the results with one and two DCs in terms of costs and responsiveness.

6
2.3 Network Optimization (NO) for Facility Location Planning: Comparing Po-
larbear’s Supply Chain Design Alternatives

After selling the bicycles from the newly established DC according to the GFA results, Polar-
bear decided to produce their own bicycles. Their production facility has now been established
in Nuremberg and 250 bikes are produced each day. Recently, they have received an offer from
a Polish production factory to rent a DC in the Czech Republic at a reasonable price. The same
company also wants to offer them rental of a factory in Warsaw, Poland, even though they
already have one factory in Germany. Polarbear must now decide which SC design is more
profitable:

 Option 1: DC in Germany and Factory in Germany

 Option 2: DC in Germany and Factory in Poland
 Option 3: DC in Czech Republic and Factory in Poland
 Option 4: DC in Czech Republic and Factory in Germany

In Fig. 1, the different possibilities for the path networks are shown. The dotted lines show
possible alternatives and the solid lines the existing structure of Polarbear’s SC.

Figure 1. Network optimization alternatives

The aim of the NO is to determine which network design is optimal based on Polarbear’s se-
lected KPIs, e.g., profit.

Therefore, the factory in Warsaw, Poland, the DC in the Czech Republic, and the DC in Steim-
elhagen were added as inputs to the model along with the Nuremburg factory. To enable the
model’s calculation, the reality of the case must be simplified: all demand is assumed to be
deterministic without any uncertain fluctuations. To define the two-stage NO problem
(transport between factories and DCs and between DCs and customers) from a mathematical
perspective, several parameters must be input as data. These are shown in Table 2.

7
 Costs Value in USD
Factory Nuremberg: fixed (other) costs, per day 15,000
Factory Poland: fixed (other) costs, per day 5,000
DC Germany: fixed (other) costs, per day 15,000
DC Germany: carrying costs (per bicycle) 3.00
DC Czech Republic: fixed (other) costs, per day 5,000
DC Czech Republic: carrying costs 2.00
DC Germany: processing costs (inbound and outbound ship- 2.00
ping per pcs)
DC Czech Republic: processing costs (inbound and outbound 1.00
shipping per pcs)
Factory Nuremberg: production costs (per bicycle) 250
Factory Poland: production (per bicycle) 150
All bicycles: product purchasing costs 30
Transportation costs; Paths: from factory - to DCs 0.01 * product(pcs) * distance
Transportation costs; Paths: from DCs - to customers 0.01 * product(pcs) * distance
Unit revenue 499
Table 2. Cost inputs to optimization model

The costs of the rent for the factory in Poland and the DC in Czech Republic are included in
“other costs”. For transport, it is always assumed that each truckload fits 80 bicycles, and trucks
travel at a speed of 80 km/h.

Creating an ALX model

Step 1. Create scenario PB NO Advanced+.

Step 2. Enter data from Table 2 and Fig. 1.

Performing experiments

Step 1. Go to NO Experiment and run it with the Demand variation type “95-100%”.

Step 2. Analyze the results using statistics “Optimization Results”, “Flow Details”, “Produc-
tion Flows”, “Demand”, and “Overall Stats”:

a) What is the most profitable SC design?

b) Is demand for all customers satisfied?
c) What is the total revenue of the most profitable SC?
d) What is total profit of the most profitable SC?

8
 e) Compare the data in statistics “Production Flows” and Table “Demand”. Does the
production quantity correspond to the total demand?
f) Compare the optimal SC design as computed in the NO and the initial SC design
(factory and DC in Germany) in terms of profit.
g) What other costs should be considered when redesigning the SC according to NO
results?
h) What other factors, apart from costs, should be considered when re-designing the
SC according to the results of the NO?
i) Explain the role of penalties in network optimization problems (e.g., in table “Pro-
duction”).

NOTE! In order to run the NO experiment, units in experiment settings should be aligned
with product data (e.g., m3 and pcs should be used consistently, and unit conversions
should be defined if needed in case you use, e.g., m3 in Experiment and pcs in Data).

The optimization results show that the highest profit can be achieved in the SC design with a
DC in Czech Republic and a factory in Poland. However, the negotiations with the factory in
Poland revealed another constraint: the Polish factory would only consider annual quantities of
each bicycle type within the range of 10,000 units (minimum capacity utilization) and 25,000
units (maximum capacity utilization). Polarbear must now conduction another NO to include
this additional capacity constraint.

Creating an ALX model

Step 1. Create scenario PB NO Advanced+ Constrained.

Step 2. Enter capacity constraints data into table “Production”.

Performing experiments

Step 1. Go to NO Experiment and run it.

Step 2. Analyze the results using statistics “Optimization Results”, “Flow Details”, “Produc-
tion Flows”, “Demand”, and “Overall Stats”:

a) What is the most profitable SC design considering the capacity constraint of the
factory in Poland?
b) What is the total profit of the most profitable SC?

9
 c) Compare the optimal SC design with the capacity constraint, as computed in the
second NO, and the optimal SC design without the capacity constraint, as computed
in the first NO experiment, in terms of profit.
d) Which differences between solutions in incapacitated and capacitated scenarios can
be observed? Explain.

Polarbear’s management is interested in the optimization results and starts thinking of redesign-
ing the SC. At the same time, the marketing manager of Polarbear expresses concerns that the
NO model considers only one period and demand is deterministic. In reality, demand for many
products is stochastic and subject to seasonal fluctuations in four periods, i.e., winter, spring,
summer, and fall (Tables 3 and 4).

Customer Bicycle Type Mean Demand per Standard Deviation

day of Demand per day
Cologne x-cross 2 -
Cologne urban 50 8
Cologne all terrain 15 2
Cologne tour 10 3
Bremen x-cross 7 -
Bremen urban 30 6
Bremen all terrain 20 5
Bremen tour 20 3
Frankfurt am Main x-cross 6 -
Frankfurt am Main urban 5 -
Frankfurt am Main all terrain 4 -
Frankfurt am Main tour 5 -
Stuttgart x-cross 15 -
Stuttgart urban 15 -
Stuttgart all terrain 1 -
Stuttgart tour 40 6
Table 3. Stochastic customer demand

Period Coefficient Time

Winter 0.7 January 1 – March 31
Spring 1.2 April 1 – June 30
Summer 1.5 July 1 – September 30
Fall 1 October 1 – December 31

10
Table 4. Demand coefficients

Polarbear conducts an NO again to include these additional features.

Creating an ALX model

Step 1. Modify your scenario PB NO Advanced+ or PB NO Advanced+ Constrained in

tables “Demand”, “Periods”, and “Production”. Assume that the existing capacity constraints
for the factory in Poland are distributed equally among the periods.

Performing experiments

Step 1. Go to NO Experiment and run it.

Step 2. Analyze the results using statistics “Optimization Results”, “Flow Details”, “Storage
by Products”, “Demand”, and “Overall Stats”:

a) Explain the inventory allocation in table “Storage by Products”.

b) Explain the values of Demand Min and Demand Max in table “Demand”.
c) Compare the scenario with only a single period and deterministic demand and the
new scenario according to each scenario’s financial performance in table “Overall
Stats”.

2.4 Simulation: Dynamic analysis of the Polarbear’s supply chain design

In simulation, we extend our analysis by adding the following features:

- We transit from flows (as in NO) to orders, i.e., the customer demand is no longer con-
sidered an aggregated flow during a period, but it is now generated as orders at certain
intervals, e.g., 10 bicycles every day.
- We introduce inventory control to manage ordering processes.
- We introduce sourcing policies (e.g., single vs. multiple sourcing) to manage replenish-
ment processes.
- We introduce shipment control (LTL/FTL) to manage shipment processes.

First, we simulate the SC with two factories in Poland and Germany and a DC in the Czech
Republic subject to customer demand from Table 1 (see GFA exercise) and the following data
(Table 3).

11
 Object Inventory Po- Expected Trans- Production Sourcing Trans-
licy Lead portation Time per Policy portation
Time Time, Unit, days Policy
Min Max
(ELT), days
days
Factory Ger- 50 100 2 0.1 LTL
many
Factory Poland 50 100 2 0.07 LTL
DC Czech Re- 50 100 2 Closest LTL
public dynamic
Customers 5
Table 3. Parameters for simulation model

For the DC and the factory, three inventory policies have to be developed. We assign the DC
and factories a “min-max policy” for all products, where the minimum stock is 50, the maxi-
mum stock is 100, and the initial stock is 50 bicycles.

To evaluate the simulation results, we consider six KPIs according to the needs of Polarbear:

(1) Financial KPIs, such as profit, revenue and costs

(2) ELT service level by product, which is the ratio of products delivered within the ex-
pected lead time to the total ordered quantity
(3) Demand fulfillment (product backlog)
(4) Available inventory
(5) Production capacity utilization and
(6) Lead time.

With all of the parameters described, we now run the simulation for a period of one year.

Creating an ALX model

Step 1. Create scenario PB SIM Advanced+.

Step 2. Enter data from Table 3 in tables “DCs and Factories”, “Inventory”, “Production”,
“Sourcing”, and “Unit Conversions”. Explain the data in the aforementioned tables. The data
in these tables should correspond to Table 3.

Performing experiments

Step 1. Go to Simulation Experiment.

12
Step 2. Create KPI Dashboard that includes the KPIs “Revenue, Profit, Costs”, “Lead Time”,
“ELT Service Level”, “Production Utilization”, “Demand (Product Backlog)”, and “Average
Available Inventory”.

Step 3. Run Simulation Experiment.

a) What are the profit, revenue, and costs of the SC?

b) Is demand for all customers satisfied? Explain.
c) What is the production utilization of factories in Poland and Germany? Explain why
the German factory is utilized to 100% and the factory in Poland is utilized only to
75% even if we clearly see insufficient production quantities and the backlog?
d) What is your judgment on the inventory dynamics in the SC?
e) What is your judgment on the lead time?
f) What suggestions for improvement do you have which could increase profit and
customer satisfaction?

Polarbear’s SC managers suggest the following changes to the input parameters (Table 4).

Object Inventory Po- Expected Trans- Produc- Sourc- Trans-

licy Lead porta- tion Time ing Pol- portation
Time tion per Unit, icy Policy
Min Max
(ELT), Time, days
days days
Factory Germany 5 2 LTL
x-cross 30 60 0.05
urban 100 200 0.015
all terrain 40 80 0.04
tour 75 150 0.02
Factory Poland 5 2 LTL
x-cross 30 60 0.05
urban 100 200 0.015
all terrain 40 80 0.04
tour 75 150 0.02
DC Czech Republic 2 Closest LTL
dynamic
x-cross 60 120
urban 200 400
all terrain 80 160
tour 150 300

13
 Customers 5 Closest
dynamic
Table 4. Parameters for simulation model

Creating an ALX model

Step 1. Create a copy of your scenario PM SIM Advanced and modify it according to
Table 4 and name it PB SIM Advanced_Improved.

Step 2. Change data from Table 4 in tables “Inventory”, “Production”, and “Sourcing”. The
data in these tables should correspond to Table 4.

Performing experiments

Step 1. Go to Simulation Experiment and run it.

Step 2. Analyze the results using the KPI Dashboard “Revenue, Profit, Costs”, “Lead Time”,
“ELT Service Level”, “Production Utilization”, “Demand (Product Backlog)”, and “Average
Available Inventory”.

a) What are the profit, revenue, and costs of the SC? Did we improve?
b) Is demand for all customers satisfied? Explain.
c) What is the production utilization of factories in Poland and Germany? Explain.
d) What is your judgment on the inventory dynamics in the SC?
e) Explain how MIN (reorder point) and MAX (target inventory) values have been
computed.
f) What is your judgment on the lead time?
g) Explain why the changes made improved SC performance.
h) How can you validate the simulation modelling results using the previous network
optimization experiments?
i) Change the LTL policy: allow order aggregation for the period of two days. Observe
the changes in SC performance and suggest actions for improvement.
j) Investigate the impact of changing inventory policy on SC performance. Run exper-
iments for R,Q policy (continuous review with fixed order quantity) and regular
policy (periodic review with fixed order quantity). Compare the results.
k) Perform a Safety Stock Estimation experiment. Explain the results.

14
2.5 Risk analysis

Creating an ALX model

Step 1. Create a copy of your scenario PB SIM Advanced_Improved and rename PB SIM
Advanced_Improved_Risk.

Step 2. Use table “Event” to setup a breakdown of the DC for two months starting July 1.

Performing experiments

Step 1. Go to Simulation Experiment and run it.

Step 2. Analyze the results using the KPI Dashboard “Revenue, Profit, Costs”, “Lead Time”,
“ELT Service Level / Service Level”, “Production Utilization”, “Demand (Product Backlog)”,
and “Average Available Inventory”.

a) How does a disruption of the DC impact the KPIs?

b) How would you suggest managing disruption risks in the SC?

2.6 Comparison experiment

In order to simplify a comparison of three considered simulation scenarios, we run a comparison

experiment.

Creating a Comparison Experiment

Step 1. Open scenario PB SIM Advanced and go to Comparison Experiment.

Step 2. Select scenarios PB SIM Advanced _Improved, PB SIM Advanced _Im-

proved_Risk, and PB SIM Advanced to compare.

Step 3. Create KPI dashboard. The KPIs for a comparative analysis of three scenarios should
be “Demand (backlog)”, “Profit”, “ELT Service Level by Orders”, and “Service Level by Or-
ders”.

Performing experiments

Step 1. Run Comparison experiment.

15
Step 2. Analyze the results.

a) What are the profit, ELT service level, Service level, and order backlog for the two
scenarios? Did we improve in the Improved scenario?
b) How can you explain the relationship between the Simulation and Comparison ex-
periments?

2.7 Validation using Variation

Rather than running the same simulation multiple times with different parameter values or com-
binations, the variation experiment allows multiple variations of the same simulation to be run
simultaneously. A variation experiment highlights how KPIs change depending on different
parameter values. This kind of sensitivity analysis can also be used to verify the validity of the
results of the simulation model.

A variation analysis must be performed to check Polarbear’s simulation model and the results.
The daily demand for “urban” bicycles in Cologne can be varied with a minimum demand of
50 and a maximum demand of 170 in steps of 40. The variation should be performed for a
period of one year.

Creating a Variation Experiment

Step 1. Open scenario PB SIM Advanced_Improved and go to Variation Experiment.

Step 2. Select the parameter we will vary (Demand).

Step 3. Create KPI dashboard. The KPIs for a variation analysis should be “Demand (back-
log)”, “Profit”, and “ELT Service Level”.

Performing experiments

Step 1. Run Variation experiment.

Step 2. Analyze the results using the KPIs “Demand (backlog)”, “Profit”, and “ELT Service
Level”.

a) What is the profit, ELT service level, and order backlog for different demands?
b) Why and how do the KPIs change as demand changes?
c) Is the simulation model sensitive?

16
2.8. Simulation with BOM

Finally, we extend our three-stage SC to a four-stage SC. Three different just-in-time suppliers
are introduced which supply modules for the bicycles based on single sourcing principles.

Figs. 2 and 3 show the general structure of the network.

Figure 2. Polarbear Bicycle’s network design

In Fig. 3, the inbound network is shown with three suppliers for seats, brakes, and frames. A
single sourcing strategy is used for each component.

Figure 3. Polarbear Bicycle’s inbound network

Create new scenario for four stage SC, define all necessary parameters and policies, and per-
form Simulation, Variation and Comparison experiments.

2.9 Recommendations

Develop recommendations for Polarbear Bicycle’s management. Consider GFA, NO, SIM,
Comparison and Variation results. What SC design would you recommend? Consider the im-
pact of inventory control and production capacities. Are there any other factors that should be
taken besides the results of the experiments performed?

17
3. Case study 2

3.1. Description of Case Study

BERLIN BREWERY is a Berlin beer brand well-known for its small, traditional, handmade
beer. Currently, they have one brewery (factory) and one DC in Berlin. Although they have
been in business a short time, they already have 50 customers in eight countries. Nevertheless,
they face several problems which need to be solved. On the one hand, the German beer market,
which is the primary market with the highest sales for BERLIN BREWERY, can be character-
ized as a mature market with strong prices and pressure from competition. A sales crisis has
been going on for years and competition is intensifying as a result of consolidation processes
implemented by international brewing companies. To expand their distribution network and
reduce costs by inventory management improvements, a simulation and optimization analysis
must be performed to evaluate their current SC performance and develop suggestions for its
improvement.

To provide context, an overview of the German and international beer market will be explained,
and the relevant technical details of the BERLIN BREWERY and its SC presented. Next the
current problem will be described, followed by developments of the GFA, NO, and Simulation
alongside the respective results. Finally, recommendations for management need is given based
on the computational results.

Beer is an alcoholic and carbonated beverage which is made by fermenting water, malt, and
hops. The production of beer can be traced back to the farmers who lived in the 4th millennium
BC during the Babylonian and pharaonic times. In Europe, the addition of hop to water only
became generally valid in the 14th century. In the early Middle Ages, beer was brewed primar-
ily in monasteries. Particularly in northern Germany, the top-fermented brewing industry flour-
ished. After the Thirty Years’ War, Bavaria gradually came to the fore as a beer country. To
save the purity and tradition of beer, the Bavarian dukes Albrecht IV and William IV regulated
the raw materials with strict purity laws. According to these laws, only barley, hops and water
could be used to make beer. The purity requirement of 1516 is the oldest, still valid food legis-
lation in the world (o.V.2017).

The German brewing tradition enjoys a prestigious reputation worldwide. According to the
German Brewers’ Federation (o.V.2017), there were around 5,000 beer brands in Germany at

18
the beginning of 2015 and over 6,000 in 2017. The trend is rising, although consumption is
decreasing.

The market for beer products and the German brewing industry are characterized by strong
prices and competition. As mentioned, a sales crisis has been going on for years and the com-
petition is intensifying because of consolidation processes implemented by international brew-
ing companies.

The SC of a brewery is highly complex. First, essential ingredients (hops, malt, yeast, and wa-
ter) and extra ingredients have to be shipped to the brewery. Shipping hops is particularly com-
plex as the product is very sensitive: moisture, heat, and oxygen reduce its brewing value over
time. As a result, hops must be stored between -2° and +3° Celsius. After the brewing process,
the young beer is stored in tanks at a temperature of 1° to 2° Celsius for a length of time between
three weeks and three months. In this time, the beer ferments and gets its characteristic colour.
In the final steps, the beer is filtered, bottled, and labelled (Plank, 2013). Figure 1 summarizes
the logistics process of the craft beer industry.

Figure 1: Logistics process of the craft beer industry

Warehousing, picking, and loading are carried usually out by company employees, but this
depends on size and internal factors for each company. Some companies set up their own ser-
vice companies for in-house logistics, but most breweries engage external companies for sorting
empties, cleaning empties, and the cleaning of the plant. Compared to other industries, the dis-
tribution of beer is diversified as many different channels have to be covered. These include the
classic food retail trade, beverage disposal markets, food discounters, petrol stations, and gas-
tronomy. However, the importance of different distribution channels has changed in recent
years. Food retailers (especially discounters) have gained more importance in the brewing in-
dustry, while the importance of gastronomy has declined. In general, all channels receive the
beer either directly from the breweries (it’s possible that a third-party-logistics provider takes
care of transportation) or through a beverage wholesaler.

19
3.2 Problem Statement

Currently, the BERLIN BREWERY’s Brewhouse consists of 5 tanks which have a total capac-
ity of 20hl. Assuming that demand for its beer is rising, the BERLIN BREWERY has the option
to expand its Brewhouse with more tanks. The whole process from the original brewing to the
finished product requires between four to six weeks depending on how long each type of beer
must be stored. As the location of production in Berlin is small, all beer produced is stored in
the only DC in Berlin. An external service provider takes care of all logistical requirements.
Currently, most of the beer BERLIN BREWERY sells is sold in Berlin, though the craft beer
is also sold to wholesalers all over Germany. Since 2018, customer locations include Switzer-
land, Austria, Sweden, Norway, France, Italy, and Spain. BERLIN BREWERY collaborates
with three suppliers in Germany who deliver empty beer bottles in crates, as well as the hops
and malt. Empty bottles in crates are delivered from a location close to Nuremberg, the hops
come from Koblenz, and malt is delivered by a supplier from a location close to Dresden. BER-
LIN BREWERY’s current sales’ figures and further financial figures, which are important for
the analysis, are summarized in the next step. One-off acquisition costs for the brewing equip-
ment is $300,000. Moreover, maintenance costs (including energy and electricity) for the loca-
tion of the factory are $80,000 per month or $2,630 per day.

Assuming that one beer crate consists of 20 bottles, the whole cost for one crate is $10. Note
that in the anyLogistix model, we consider a product “Beer” which is equivalent to one beer
crate. The costs of the product “Beer” can be broken down as follows in Table 1.

Usage per beer Measure used in

BOM Costs per beer crate, $
crate anyLogistix
Hops 7.92 gr. 1 piece 0.32
Malt 1390 gr. 1 piece 2.22
Crate 20 bottles 1 piece 6.00
Production pro-
1.46
cessing costs
Total 20 bottles 10.00
Table 1: Bill of materials (BOM), based on information of BERLIN BREWERY

Table 1 depicts the BOM (Bill-of-Materials) for the product “Beer” which consists of 1 piece
of Hops, 1 piece of Malt and 1 piece of Crate.

20
Carrying costs, which include warehousing costs, handling inside inventory, and inventory
costs, are estimated to be $0.005 per beer crate ($0.2 per pallet) per day. The transportation
from the BERLIN BREWERY factory to the DCs are calculated as volume-distance based
transport and costs are $0.00175 per km/beer crate ($0.07 per pallet). The transport from the
three suppliers to the BERLIN BREWERY factory is paid by suppliers. Inbound and outbound
costs are shipment processing costs. Outbound costs are assumed to be $0.66 per beer crate
($26.40 per pallet) and inbound costs are $1 per crate ($40 per pallet). Hops and malt are deliv-
ered in a one-kilogram packaging unit: one pallet of malt or hops equals 40 packaging units.

BERLIN BREWERY wishes to expand its sales and work as efficiently as possible to increase
profit. To reach these goals, several problems must be overcome: as mentioned in the first chap-
ter, beer consumption in Germany, BERLIN BREWERY’s main market, is decreasing and the
market as a whole is highly competitive. The German beer market is a mature, nearly saturated
market. Two potential solutions exist: expansion into other countries or increased sales to ex-
isting customers. These options instigate further challenges as BERLIN BREWERY has only
one DC in Berlin. Long routes and long delivery times to individual customers are the main
problems. Because of the long routes, BERLIN BREWERY can respond only relatively inflex-
ible to spontaneous requests and a high number of unnecessary routes might be taken. There is
also the risk of manmade or natural disruptions which can influence service quality (e.g. a storm
destroys DC).

In sum, the goal of BERLIN BREWERY is to expand their distribution network, serve their
customers as efficiently and satisfyingly as possible, raise their sales numbers, and increase
profit. This is possible by optimizing their SC: an optimal number of DCs as well as good
locations for these DCs must be found to save as much logistics costs as possible. Loss of
quality and delivery problems should be avoided.

Fig. 2 depicts the current situation. Production is in the centre of Berlin and the raw ingredients
are shipped by truck directly from Dresden, Nuremberg, and Koblenz. To store as little as pos-
sible, raw materials are delivered on demand and used directly (JIT- just-in-time) for the pro-
duction. The beer is delivered to 50 customers all over Europe.

21
Figure 4: Current network of BERLIN BREWERY

To provide a better understanding of the circumstances, a few assumptions are made:

 All prices and costs are shown and calculated in $.
 All processes are considered in terms of (beer) crates or pallet specifications, rather than
bottles. This is because BERLIN BREWERY sells their beer only in whole crates, and
these terms help to simplify the model. In one beer crate there are always 20 bottles of
beer, which have 0.33 liters of content per bottle (6.6 liters per crate).
 1 pallet = 40 beer crates = 800 beer bottles
 The recycling deposit on bottles is not considered.
 Transportation costs from the factory to all DCs are the same.
 Transportation-/ handling costs from the DCs to the customers are adapted to the price
level of the actual country.
 One year consists of two periods:
o Summer period: 01.05.2021 - 31.10.2021
Demand coefficient: 1.2 (meaning 1.2 times higher demand as in winter due to
higher demand during warm months)
o Winter period: 01.11.2021 - 30.04.2022
Demand coefficient: 1.
 Orders are received every seven days (static demand).

22
  Transportation speed is 80km/h, the capacity of a truck is 1,320 crates, which equals 33
pallets which are single stocked.
 Price per crate for customer: $60.00.

We assume that BERLIN BREWERY will sell 260,405 beer crates within the coming summer
period (151,035 crates) and winter period (109,370 crates). Fig. 5 below shows the customers,
according to the summer and winter periods. Substantial sales are made in Germany, especially
in Berlin (32,807 beer crates per year) and the least sales are achieved in Basel, Switzerland
(871 beer crates per year).

Figure 5: Distribution of sales by country and period, own illustration

The main customers are beverage retailers, which purvey to smaller retailers or restaurants.
Therefore, it is considered that only one wholesaler is supplied per city. This wholesaler makes
resales independently. As a result, no further storage costs are incurred as no further DC is
required. Transportation to the wholesalers and handling is currently being handled entirely by
a logistics service provider as BERLIN BREWERY does not yet have the necessary capacity
and occupancy rate for profitable shipment. This service provider picks up the goods in the
brewery, stores them in their own warehouse, and launches directly to the distributor as needed.
The current financial performance is presented in Table 2.

$
KPIs
346,991
Transportation cost
996,450
Other cost
905,117
Inbound processing cost

23
 348,895
Outbound processing cost
2,282,011
Inventory purchases
390,133
Production

Revenue 15,684,866

10,407,351
Profit
Table 2: Current cost structure

3.3 Greenfield Analysis (GFA)

Now we conduct a GFA. The aim of this GFA is to determine optimal DC locations in the SC
subject to minimum total transportation costs.

Creating an ALX model

Step 1. Create scenario BR GFA Advanced.

Step 2. Enter data into tables Customers, DCs and Factories, Demand, Unit Conversions,
and Products.

Performing experiments

Step 1. Go to GFA Experiment and run it for “Number of sites = 7”.

Step 2. Analyze the results using statistics “Flows” and “New Sites”:

a) What are the optimal coordinates of the new DCs?

b) What is the maximum distance from an optimal DC location to a customer?
c) What are the total costs of the SC?
d) Compare the data in statistics “Flows” and table “Demand”. Do we satisfy all customer
demand from the optimal DC locations?
e) Which costs, relevant for choosing an optimal DC location, were not considered in this
GFA analysis?

Step 3. Go to GFA Experiment and run it for “Number of sites = 2”.

Step 4. Analyze the results using statistics “Flows” and “New Sites”:

a) What are the total costs of the SC?

b) Compare the results with 7 and 2 DCs in terms of costs and responsiveness.

24
3.4 Network Optimization (NO)

The NO offers the possibility of optimizing an existing SC according to maximum possible

profit. In this case, the solutions of the GFA will be taken into account to optimize the SC.
Having checked the suggested GFA sites for DCs, the SC manager of BERLIN BREWERY
analysed those locations further regarding additional factors such as availability of warehouses
to rent, construction costs for building new warehouses, fixed costs, infrastructure, future de-
mand forecasts, etc. As a result, some of the GFA suggested locations are moved (Figure 6).

Figure 6: Alternative DCs locations

The NO goal is to find the SC design with the highest possible profit. To define the NO problem
from a mathematical perspective, several parameters must be input as data. Each of the DCs
has an inventory capacity of min. 5,800 beer crates and max. 11,600 beer crates as well as a
one-week inventory range. The brewery can stock 10,000 crates at maximum and should carry
an inventory of at least 5,000 crates. Customers and their demands remain the same as in the
GFA. To avoid confusion, the DCs are now marked as red icons. Green icons are added to
symbolize the suppliers of beer ingredients. These suppliers are located in Nuremberg, Koblenz,
and Dresden.

Table 3 contains the costs of the sites. These numbers have been adjusted to the income ratios
of each country. Two of them, Berlin and Bochum, are in Germany, and the overall prices in
Berlin are cheaper than in western Germany.

25
 Other costs Carrying Outbound Inbound ship- Transporta-
Sites / costs in ($)
per day costs per day shipment pro- ment pro- tion costs per
beer pallet,
per beer pallet cessing costs cessing costs from DCs to
per beer pallet per beer pallet customers and
from Brewery
to DCs
120 1.00 0.07
DC Austria 0.20 0.66
70 0.5 0.035
DC Spain 0.10 0.33
310 1.55 0.108
DC Sweden 0.40 1.02
120 1.07 0.07
DC Bochum 0.20 0.68
146 1.46 0.102
DC Switzerland 0.20 0.96
90 0.5 0.05
DC Italy 0.20 0.33
100 1.00 0.07
DC Berlin 0.20 0.66
2,630 40.00 (ingre- 0.07
Factory Berlin 0.005 0.66 (beer)
dients)
(beer)
Table 3: Cost structure per site

Production costs at the brewery are $1.46. Revenue is $60.00.

Creating an ALX model

Step 1. Create scenario BR NO Advanced.

Step 2. Enter data in tables “Facility Expenses”, “Paths”, “Product Flows”, and “Production”.
Explain the data in the aforementioned tables. The data in these tables should correspond to
Table 3.

Performing experiments

Step 1. Go to NO Experiment and run it.

Step 2. Analyze the results using statistics “Optimization Results”, “Flow Details”, “Produc-
tion Flows”, “Demand”, and “Overall Stats”:

a) What is the most profitable SC design?

b) Is demand for all customers satisfied?
c) What is total profit of the most profitable SC?

26
 d) Compare the optimal SC design as computed in the NO and the initial SC design
(factory and DC in Germany) in terms of profit.
e) What other costs should be considered when redesigning the SC according to NO
results?
f) What other factors, apart from costs, should be considered when redesigning the SC
according to the results of the NO?

Given the threat of DC disruptions and the expansion plans with new customers, BERLIN
BREWERY decides to redesign the SC by adding a second DC in Spain, which means that the
fifth-best alternative is chosen out of 10 best SC designs displayed in “Optimization results”.
The difference in profit between the best and the fifth-best SC design is very small but the SC
design with 2 DCs offers more flexibility and resilience.

3.5 Simulation

Simulation can be used in many ways. It promotes understanding of how the SC will react in
the event of disruptions, such as outages and/or external influences. To run different simula-
tions, the outcome of the NO has been used. In the case of BERLIN BREWERY, a two-month
disruption has been simulated to see what happens if only one DC is kept or if a second is rented
in Spain. Differences in, for example, the service level and profit could be decisive. For the
simulation, assumptions must be made to make the model more realistic. The sourcing policies
are defined as follow: for the DCs, the program should choose the closest dynamic source. In
this case, there is only one factory and the customers should choose the fastest dynamic source
to receive orders as soon as possible.

In simulation, we extend our analysis by adding the following features:

- We transit from flows (as in NO) to orders. i.e., the customer demand is no more con-
sidered as an aggregated flow during a period but it is now generated as orders in certain
intervals, e.g., 10 crates per day.
- We introduce inventory control to manage ordering processes.
- We introduce sourcing policy (e.g., single vs. multiple sourcing) to manage replenish-
ment processes
- We introduce shipment control (LTL/FTL) to manage shipment processes.

First, we simulate the SC with two DCs in Germany and Spain subject to the following data
(Table 4).

27
 Object Inventory Policy Sourcing Policy Transportation
Policy
Min Max
Factory Germany (beer) 2000 5000 Closest dynamic LTL
Factory Germany (ingredients) On demand LTL
DC Germany 1800 11600 Closest dynamic LTL
DC Spain 1800 11600 Closest dynamic LTL
Customers Closest dynamic LTL
Table 4. Parameters for simulation model

To evaluate the simulation results, we consider six KPIs according to the needs of BERLIN
BREWERY:

(1) Financial KPIs, such as profit, revenue and costs

(2) Service level by products, which is calculated as (number of outgoing orders / number
of placed orders), where an outgoing order is an order that is not dropped
(3) Available inventory including backlog at DCs.

With all of the parameters described, we now run the simulations for a period of one year from
May 1, 2021 until April 30, 2022.

Creating an ALX model

Step 1. Create scenario BR SIM Advanced+.

Step 2. Enter data from Table 4 in tables “Inventory”, “Sourcing”, “Paths”, and “Shipping”.
Explain the data in the aforementioned tables. The data in these tables should correspond to
Table 4.

Performing experiments

Step 1. Go to Simulation Experiment and run it.

Step 2. Create a KPI Dashboard and analyze the results using the KPI Dashboard “Revenue,
Profit, Costs”, “Service Level”, “Average Inventory including Backlog”.

a) What are the profit, revenue, and costs of the SC? Does this result match with the
NO results? Explain.
b) Is demand for all customers satisfied? Explain.

28
 c) What is your judgment on the inventory dynamics in the SC? Explain the change in
inventory dynamics in the second part of the simulation period.

3.6. Risk analysis: Two-month disruption at one of the DCs

This analysis simulates a two-month disruption of the DC in Berlin from August1st to Septem-
ber 30th for two different network design scenarios, i.e., with two DCs and with a single DC in
Berlin.

Creating an ALX model

Step 1. Create scenarios BR SIM Advanced Disruption Risk 2 DC and BR SIM Advanced
Disruption Risk 1 DC.

Step 2. Enter data about the disruption in table “Events”.

Performing experiments

Step 1. Go to Simulation Experiment and run it.

Step 2. Create KPI Dashboards and analyze the results using the KPIs “Revenue, Profit,
Costs”, “Service Level”, “Average Inventory including Backlog”.

a) What are the profit, revenue, and costs of the SC for the two different network design
scenarios?
b) Is demand for all customers satisfied? Explain.
c) What is your judgment on the inventory dynamics in the SC? Explain the change in
inventory dynamics in the disruption period.

3.7 Comparison experiment

In order to simplify comparison of the two simulation scenarios subject to disruption with each
other and with the “ideal” scenario (no disruption), we run a Comparison experiment.

Creating a Comparison Experiment

Step 1. Open your scenario BR SIM Advanced and create a Comparison Experiment.

Step 2. Select scenarios BR SIM Advanced, BR SIM Advanced Disruption Risk 1 DC and
BR SIM Advanced Disruption Risk 2 DC.

29
Step 3. Create KPI dashboard. The KPIs for a comparative analysis of the three scenarios
should be “Available Inventory”, “Profit”, “Total Costs” and “Service Level”.

Performing experiments

Step 1. Run Comparison experiment.

Step 2. Analyze the results using the KPIs “Available Inventory”, “Profit”, “Total Costs” and
“Service Level”.

a) What are the profit, service level, average inventory and total costs in each of the
three scenarios? Explain.
b) What’s the relationship between the Simulation and Comparison experiments?

3.8 Validation using Variation

Rather than running the same simulation multiple times with different parameter values or com-
binations, the variation experiment allows multiple variations of the same simulation to be run
simultaneously. A variation experiment highlights how KPIs change depending on different
parameter values. This kind of sensitivity analysis can also be used to verify the validity of the
results of the simulation model.

A variation analysis will be now performed. The MIN parameter for the DCs will be varied
with a minimum reorder point of 900 and a maximum reorder point of 2300 in steps of 200.
The variation should be performed for a period of one year.

Creating a Variation Experiment

Step 1. Open scenario BR SIM Advanced and create Variation Experiment.

Step 2. Check the parameter we will vary (i.e., Inventory Policy: DC for product beer:
MIN).

Step 3. Create KPI dashboard. The KPIs for a variation analysis should be “Available inven-
tory including Backlog”, “Profit”, and “Service Level”.

Performing experiments

Step 1. Run Variation experiment.

30
Step 2. Analyze the results using the KPIs “Available Inventory including Backlog”, “Profit”,
and “Service Level”.

a) What are the profit, service levels and inventory for different reorder points?
b) Why do the KPIs change as the reorder point changes?
c) Is the simulation model sensitive?

3.9 Recommendations

Develop recommendations for BERLIN BREWERY management. Consider GFA, NO, Simu-
lation, Comparison, and Variation results. Which SC design would you recommend? Consider
the impact of inventory control and disruptions. Are there any other relevant factors that should
be taken into account besides the experiments performed?

3.10 References

Ivanov, D. (2021): Supply chain simulation and optimization with anyLogistix: Teaching notes.
Berlin School of Economics and Law, 5th Edition
Ivanov, D., Tsipoulanidis, A., Schönberger, J. (2021): Global Supply Chain and Operations
Management - A Decision-Oriented Introduction to the Creation of Value, Springer In-
ternational Publishing Switzerland, 3rd Edition
Maak, K., Haves, J., Stracke, S. (2011) Entwicklung und Zukunft der Brauwirtschaft in
Deutschland https://www.boeckler.de/pdf/p_edition_hbs_260.pdf (Dec 12, 2017)
o.V .(2017) Unser Reinheitsgebot - Brauer Bund. http://www.brauer-bund.de/ (Nov 22, 2017)
o.V. (2015) From Grains to Growlers: A Look at the Craft Beer Industry Supply Chain Mate-
rial Handling and Logistics http://www.mhlnews.com/global-supply-chain/grains-growl-
ers-look-craft-beer-industry-supply-chain-infographic (Dec 12, 2017)
o.V. (2016) Bierkonsum in Deutschland bis 2016 Statista. https://de.statista.com/statistik/da-
ten/studie/4628/umfrage/entwicklung-des-bierverbrauchs-pro-kopf-in-deutschland-seit-
2000/ (Nov 12,2017)
Plank, R. (2013) Die Anwendung der Kälte in der Lebensmittelindustrie. Springer-Verlag,
Heidelberg.

31