Professional Documents
Culture Documents
Optimization
SUPPLY CHAIN ANALYTICS
Everyday question ?
Before you Run out , get More !
A more realistic situation occurs when customer demand is uncertain. In this case, the decision on
when to place orders becomes more difficult. These more difficult problems require probabilistic
inventory models.
Deterministic vs. probabilistic
models
In deterministic models, all inputs to the problem, particularly
customer demand, are assumed to be known when the decisions are
made.
◦ In reality, a company must always forecast future demands with some type
of forecasting model.
◦ The outputs of this forecasting model might include a mean demand and a
standard deviation of demand.
◦ In deterministic models, however, only the mean is used, and any
information about the uncertainty, such as the standard deviation, is
ignored.
◦ This makes the resulting models simpler, but usually less realistic.
Probabilistic models use this information about uncertainty explicitly.
They are typically more difficult to analyze, but they tend to produce
better decisions, especially when the level of uncertainty is high.
External vs. internal demand
A second factor in inventory modeling is whether demand for the
product is generated externally or internally.
External demand (or independent demand) occurs when the company
that sells the product cannot directly control the extent or the timing of
customer demand.
In contrast, internal demand (or dependent demand) occurs in most
assembly and manufacturing processes.
A framework for Inventory Mgt
DEPENDENT
INDEPENDENT DEMAND
DEMAND
Kanban/
Reorder
constant
STATIONARY point- order
process in
DEMAND quantity
work
Distribution
Material
requirement
Non- STATIONARY requirement
Planning (
DEMAND Planning
DRP)
Except possibly for emergency orders, these are the only times when orders are placed.
Continuous review models can certainly be implemented, given today’s computerized access to inventory
levels in real time, and these models can result in lower annual costs than periodic review models.
However, when a company stocks many products (hundreds or even thousands), it is often more
convenient to use periodic review.
Single-product vs. multiple-
product models
A final factor in inventory modeling concerns the number of products
involved.
Models that consider only a single product are conceptually and
mathematically simpler, so we initially analyze single-product models.
However, most companies have many different products that must be
considered simultaneously.
If the company orders these items from a supplier, it may be wise to
synchronize the orders in some way to minimize ordering costs.
Types of costs in inventory
models
Ordering (setup) cost is the fixed cost incurred every time an order is placed
or a batch is produced, independent of the amount ordered or produced.
◦ This ordering cost includes the cost of paperwork and billing each time an order is
placed and could include other costs as well, such as paying a truck driver to
deliver the order to the company’s warehouse.
◦ If the product is produced rather than ordered, this cost can include the cost to
set up equipment.
The unit purchasing (or production) cost is the cost for each additional unit
purchased or produced (often referred to as the variable cost).
The holding (or carrying) cost is the cost that motivates the company to
keep less inventory on hand.
◦ This cost generally has two components, the financial holding cost and the
nonfinancial holding cost. The nonfinancial holding cost is usually the cost of
storing the product.
Types of costs in inventory
models continued
It is often important to measure the cost of running out of inventory.
This shortage (or penalty) cost is a difficult cost to measure.
◦ At one extreme, there are lost sales models, where any demands that occur
when inventory is zero are lost; these customers take their business
elsewhere.
◦ At the other extreme, there are complete backlogging models, where
demands that occur when inventory is zero are satisfied as soon as a new
order arrives.
◦ Both of these models—or any in between, called partial backlogging
models—have negative effects for the company.
Economic order quantity
(EOQ) models
We first examine a class of models called economic order quantity
(EOQ) models. These are the most basic of all the inventory planning
models.
◦ A company orders a single product from a supplier and sells this product to
its customers.
Economic order quantity
(EOQ) models continued
◦ Orders can be placed at any time (continuous review).
◦ There is a constant, known demand rate for the product, usually expressed in
units per year (annual demand).
◦ There is a constant, known lead time for delivery of the product from the
supplier.
◦ There is a fixed ordering cost each time the product is ordered, independent
of the size of the order.
◦ The price the company charges for the product is fixed.
◦ The annual holding cost is proportional to the average amount of inventory
on hand.
The basic EOQ model
The most basic EOQ model adds the following two assumptions.
◦ No stockouts are allowed; that is, the company never allows itself to run out
of inventory.
◦ The unit cost of purchasing the product from the supplier is constant. In
particular, no quantity discounts are available.
the average level of the inventory at a typical point in time is (Q+0)/2 = Q/2,
For example, the annual holding cost, determined by the equation for annual financial holding cost, is
calculated in cell B18 with the formula
=Annual_interest_rate*Unit_purchasing_cost*Order_quantity/2.
Note that the only changing cell is the Order_quantity cell.
Example : Using Solver
We maximize annual profit with a single changing cell, the order
quantity cell.
There are no constraints other than nonnegativity of the order quantity.
Also, GRG Nonlinear Solver should be used. The reason is that the
decision variable Q appears in the denominator of the annual ordering
cost. This makes the model nonlinear.
Answer :
The Solver solution specifies that Machey’s should order about 194
cameras each time it orders. This will result in about 6 orders per year,
or about one order every 59 days.
annual ordering cost = annual financial holding cost
EOQ : scenario 1
Annual demand for propene tanker is 50 tanks per week and
cost is$12 per tanker . Order cost is 500 per order. The
inventory holding cost is $3 per tank per year. How often
should propene be ordered and in what quantity ?
Scene 1: If capital cost is 5%. Suppose the full tanks are kept at
gov storage facility that charges $0.2 per tank per month
whereas empty tanks are kept in outside ground which is free.
What is storage cost?
Scene 2: The public storage company is offering an option of
leasing a fixed storage area to store propene tanks. The cost is
$18 per sq km per Year. The lease is on a yearly basis. It is
estimated 1 sq km can hold 10 tanks. Can Tank company saves
money by leasing a fixed storage area? How big storage area is
needed ?
Scene 2:
Storage cost per tank for a year = ?
Storage space required = Q
Storage Cost will be = Q * w
=Q *1.8
Capital inventory holding ( capital) cost = (Q/2)*I
Total cost = cost of order + inventory holding + storage cost
Actual storage cost = 2*1.8 = 3.6 +I = 3.6+0.6 = 4.2
Case 2: Quantity discount
Background information
The accounting form of AJ Taylor buys USB thumb drives from a large
mail-order distributor. The firm uses approximately 5000 drives per year
at a fairly constant rate.
The distributor offers the following quantity discount.
◦ If fewer than 500 drives are ordered, the cost per drive is $30.
◦ If at least 500 but fewer than 800 drives are ordered, the cost per drive is
$28.
◦ If at least 800 drives are ordered, the cost per drive is $26.
The fixed cost of placing an order is $100. The company’s cost of capital
is 10% per year, and there is no storage cost. The firm wants to find the
optimal order quantity and the corresponding total annual cost.
Non linear :Using Solver
Also, because the quantity discounts lead to a nonsmooth objective, it is a good idea to use the
Multistart option,. Alternatively, Evolutionary Solver could be used, but it doesn’t appear to be
necessary.
GRG Nonlinear Solver with the Multistart option finds the optimal solution quickly.
Additional insight is provided by the graph of total annual cost versus order quantity shown
above.
Within any quantity discount region, the curve increases almost imperceptibly, but at
quantity discount breakpoints, it decreases abruptly. This is the nonsmooth
EOQ model with shortages
allowed
A key assumption in the basic EOQ model is that the company decides,
as a matter of policy, not to allow any shortages. This means that it is
possible to prevent shortages from occurring. However, it might be in
the company’s best interests to allow a few shortages if the penalty for
a shortage is not too large.
First, are shortages backlogged or are these demands lost?
And what about the penalty cost for a shortage?
Does the penalty relate only to the number of units short per year or
also to the amount of time the shortages last?
Whatever type of shortage cost is assumed, the practical difficulty is
then assessing a specific dollar value for this cost.
Example 3:Background
information
GMB is a mail-order distributor of audio CDs that sells approximately 50,000
CDs per year. Each CD is packaged in a jewel case that GMB buys from a
supplier. The fixed cost of placing an order for jewel cases is $200. GMB
pays $0.50 for each jewel case, and its cost of capital is 10%. The cost of
storing a jewel case for one year is $0.50.
GMB believes it can afford to run out of jewel cases from time to time,
reasoning that this will simply make the time between customer orders and
customer deliveries a bit longer. It knows that there is some cost of doing
this – impatient customers might take their business elsewhere – but it is
not sure what dollar amount p to attach to this cost. It decides to use a trial
value of p=$52, reasoning that this value implies a $1 penalty for each extra
week a customer has to wait because of a backlogged jewel case.
GMB wants to develop a spreadsheet model to find the optimal order
quantity, the optimal amount to backlog, and the optimal annual cost. It
also wants to see how sensitive these quantities are to the unit shortage
cost p.
EOQ model with shortages
allowed
As the basic EOQ model, the first step of the solution is
to develop the components of the total annual cost.
The key is again a saw-toothed graph
Discussion:
Now there are two decision variables; Q, the order quantity, and b, the
maximum amount backlogged.
Each cycle has length Q/D, the time to deplete Q units at demand rate
D. But now a cycle has two parts.
During time (Q-b)/D, there is positive inventory and demands are being
met on time.
During the last section of each cycle, of length b/D, the inventory is
negative, which means that shortages exist.
right after any order arrives, there is Q – b units in inventory.
Shortage allowed
The total cost function:
The annual setup cost = k ( number of orders in a year) = K D/Q
The annual purchase cost = cD.
The annual financial holding cost is again the interest rate times half of the
purchase cost of an order icQ/2.
The annual storage cost = unit storage cost x the average inventory when
inventory is positive (Q-b)/2, x amount of time during the cycle when
inventory is positive, (Q-b)/D X total orders in a year
Annual storage cost = s[(Q-b)/2](Q-b)/D](D/Q) = s(Q-b)2/(2Q)
Annual shortage cost = shortage cost per unit p x average amount short, b/2
x amount of time when inventory is negative ,b/D x number of cycles per
year D/q
= p x( b/2) x (b/D) x (D/q)
Example: shortage
Objective: To find the order quantity and the maximum shortage
allowed that minimize total annual cost, and to see how sensitive the
solution is to the unit shortage cost.
Three approaches
1. Each product manager orders his or her model independently
2. The product managers jointly order every product in each lot
3. Product managers order jointly but not every order contains every product;
that is, each lot contains a selected subset of the products
Putting all of this together, the total annual cost to the company is f (K)i
plus the annual operating cost from any of the previous models.
In addition to any previous decision variables, such as Q, K must be
chosen, subject to the constraint K K0.
The following example illustrates the procedure.
Example: Background
information
The CompServe Company stocks expensive laser printers. The annual
demand for this product is 300 units. The cost from CompServe’s
supplier is $1000 per printer, the cost of capital is 10%, and the storage
cost per year is $30. CompServe currently incurs a setup cost of $800
per order, but it believes that by streamlining its ordering and delivery
operations, it can reduce this value and thereby achieve smaller
inventory levels. Specifically, it estimates that each 10% reduction in
setup cost will require $1500 investment. However, preliminary analysis
shows that it is physically impossible to reduce the setup cost below
$50, regardless of the amount invested.
Should the company invest in setup cost reductions, and if so, how will
affect its ordering policy?
Objective: To check, in the context of the basic EOQ model, whether it is
cost-effective to make a one-time investment in setup cost reduction.
Example : Solution
We must first find the parameters a0 and a1 of the investment cost
function f(K) by using the information on the original setup cost, $800,
and the cost per 10% setup cost reduction, $1500.
Then we can express all annual costs in terms of the decision variables K
and Q and use the Solver to optimize.
Example : Developing the
model
We will list key steps here:
◦ Parameters of setup cost reduction function.
◦ Calculate the parameters a0 and a1 of the setup cost reduction function in
cells B13 and B14 using the procedure outlined previously. Specifically,
calculate the slope a1 with the formula
=Cost_reduction_in_setup_cost/LN(0.9)
◦ calculate a0 with the formula =-Slope*LN(Initial_setup_cost)
◦ This formula ensures that the cost of making no setup cost reduction is 0.
◦ Cost of reducing setup cost.
◦ Enter the one-time investment in setup cost reduction in cell B23 with the
formula =Intercept+Slope*LN(Setup_cost_after_reduction).
◦ Then enter the equivalent annual cost in cell B24 with the formula
=B23*Annual_interest_rate.
Example : Using Solver
The rest of the model is exactly like the basic EOQ model.
To setup the Solver, we identify annual cost as the objective to minimize, the SetupCost and OrderQuan
cells as the changing cells.
We constrain SetupCost to be less than or equal to InitSetupCost and greater then or equal to
MinSetupCost, and we select the AssumeNon-Negative option.