Inventory Models with

deterministic demand
 Introduction
• Most industries have to deal with inventories. E.g., shelf and warehouse
(back room) inventory in retail, and raw material, work in process and
finished product inventory in manufacturing.
• Proper inventory management is crucial for the success of any business.
• E.g., holding too much inventory involves “tying” large capitals, while
holding little inventory leads to lost revenue and dissatisfied customers.
• To get an idea of the importance of inventory management, note that the
capital invested in U.S. inventories was estimated to be $1 trillion in 1991.
• In an inventory system, inventory is depleted by customer demand and is
replenished from orders to suppliers.
– The key questions in inventory management are
– When to order?
– How much to order?
 Type of Inventories
• Raw Material: Resources required in the
production or processing activity of the Firm
• Components: Components correspond to
items that have not yet reached completion in
the production process.
• Work in process: Inventory waiting in the
system for processing or being processed.
• Finished goods: These are the final products of
the production process
 Reason for holding inventories
• Economies of scale: Producing products in larger
quantities will lead to cheaper cost per item
• Uncertainty in demand
• Speculation: If the value of an item is expected to
increase, it is more economical to buy large
quantities now and store the items for future use.
• Smoothing: Changes in demand patterns (for
example seasonality). Producing and storing
inventory in anticipation if peak demand can help
to alleviate the disruptions caused by changing
production rates and workforce levels
 Characteristics of Inventory Systems
• Demand:
-Constant vs variable
-Known vs random: Most models assume that the
demand rate is constant
• Lead time:
-Defined as the amount of time that elapsed from
the instant that an order is placed until arrives. If
items are produced internally, lead time represents
the amount of time required to produced a batch of
items
 Characteristics of Inventory Systems
• Review time: This defines how often inventory is
reviewed. There are two basic variations. Under
continuous review, the inventory is monitored
continuously. Under periodic review, the
inventory is checked periodically (e.g. once per
day, per week, etc.)
• Example of periodic review: a grocery store in
which physical stock taking is required to
determine the current level of on-hand inventory
 Characteristics of Inventory Systems
• Excess demand: Demand that cannot be filled
immediately from stock
• Two common assumptions:
-Demand is backordered (held over to be
satisfied at a future time)
-Demand is lost
 Characteristics of Inventory Systems
• Other factors that consume inventory: E.g.,
perishable inventory (e.g. in case of food
inventory), obsolescence (e.g. clothing and
electronics inventory)
• Number of items: Most inventory models
consider a single item. Other models consider
multiple items competing over demand or
space.
 Characteristics of Inventory Systems
• Planning Horizon: this specifies the time
interval the model considers. It can be a single
period, a finite number of periods, or an
infinite horizon
 Inventory Costs
• The two key questions of when and how much to
order are commonly answered based on a cost
minimization criteria.
• There are three types of costs in an inventory system:
– The cost to place an order.
– The cost of holding inventory.
– The penalty cost for unmet demand.
 Holding Cost
• The holding cost or the inventory carrying cost is the cost
associated with storage of inventory until it is sold or
used.
-Cost of providing the physical space to store the items
-Insurance, taxes
-the cost of capital tied up in inventory
• The holding cost is computed as a function of the
amount of inventory on hand.
• It is assessed based on the average on-hand inventory or
based on inventory level at the end of a certain period.
• It is calculated per unit held in inventory per unit time.
 Ordering Cost
• The holding cost includes all those costs that are proportional to the amount
of inventory on hand, whereas the order cost depends on the amount of
inventory that is ordered or produced
• The order cost has 2 components: a fixed cost K incurred independent of the
size of the order and a variable cost incurred on a per unit basis

0 if 𝑥 = 0
𝑂𝐶(𝑥) = ൝
𝐾 + 𝑐𝑥 if 𝑥 > 0

• Costs included in K typically include book-keeping costs associated with

processing an order, fixed transportation costs, and order handling costs.
 Penalty Cost
• The penalty cost, also known as shortage cost, is the cost
of not being able to meet customer demand when it
occurs.
• The penalty cost depends on whether the excess demand
is backordered (also known as backlogged) or lost.
• In a backorder demand situation, a customer not finding
any inventory on hand is willing to wait until an order
arrives to receive his request.
• In a lost demand situation, a customer not finding any
inventory on hand, will not wait (and perhaps look for his
request elsewhere).
• p is used to denote and is charged on a per unit basis
 Estimating Inventory costs
• The ordering cost is usually easy to estimate
• Holding cost is somewhat easy to estimate based on
the firm cost of capital, which is found as a weighted
average of the interest rates from different sources
of funding.
• The penalty cost is the most difficult to estimate.
How do you measure “loss of customer goodwill”?
• Because difficulty in its estimation, sometimes a
service level constraint is used instead of a penalty
cost
 The classic Economic Ordering
Quantity Model (EOQ)
• Consider a facility (e.g., a retailer or a warehouse) that faces a constant demand
for a single item at a rate D per unit time
• Placing an order of size Q > 0 to a supplier costs K+cQ dollars, where K and c are
the fixed and variable order costs.
• Holding inventory incurs a unit cost h ($/unit/unit time) which is proportional to
average inventory level.
• The lead time until an order is received is zero.
• No shortages are allowed.
• The planning horizon is infinite.
• The objective is to determiner the ordering policy that minimizes the cost per unit
time (e.g. annual cost).
• When to order in such a situation? When the inventory level reaches 0 (saves on
holding cost).
 The classic Economic Ordering
Inventory

Quantity Model (EOQ)

Q D
1

Time
T

• Let Q be the order size

• Whenever the inventory level reaches zero, order Q.
• Now we can determine the optimal value of Q that
minimizes cost per unit time.
Inventory

Q D
1

Time
T

• T = Q/D.
• The cost per ordering cycle is the sum of holding and ordering cost.
• Let I(t) be the inventory level at time t. The holding cost per cycle is
𝑇

න ℎ𝐼(𝑡)𝑑𝑡 = ℎ × (area under inventory level) = ℎ𝑄 2 /(2𝐷) .

0

• The ordering cost per cycle is K + cQ.

• Then, the cost per ordering cycle is 𝐶𝑇 (𝑄) = 𝐾 + 𝑐𝑄 + ℎ𝑄2 /(2𝐷)
• The cost per unit time is 𝐶𝑇 (𝑄) 𝐾 + 𝑐𝑄 + ℎ𝑄2 /(2𝐷) 𝐷 𝑄
𝐶𝑈 (𝑄) = = =𝐾 + 𝑐𝐷 + ℎ .
𝑇 𝑄/𝐷 𝑄 2
Inventory

Q D
1

Time
T

• The cost per unit time is

𝐶𝑇 (𝑄) 𝐾 + 𝑐𝑄 + ℎ𝑄 2 /(2𝐷) 𝐷 𝑄
𝐶𝑈 (𝑄) = = = 𝐾 + 𝑐𝐷 + ℎ .
𝑇 𝑄/𝐷 𝑄 2

• Differentiating implies that 𝜕𝐶𝑈 (𝑄) 𝐾𝐷 ℎ

=− 2 + ,
𝜕𝑄 𝑄 2
2
𝜕 𝐶𝑈 (𝑄) 2𝐾𝐷
= 3 >0.
𝜕𝑄2 𝑄

• Therefore, CU(Q) is convex, and the optimal order quantity that minimizes CU(Q), Q*, is found
by setting the first derivative equal to zero as
• The corresponding optimal (minimum cost), CU(Q*) is CU * = cD + 2 KDh .
 EOQ facts
• Note that Q* is independent of the variable cost c.
• The optimal order quantity, Q*, is increasing in the
ordering cost K. That is, high ordering cost implies
ordering less frequently in large quantities.
• In addition, Q*, is decreasing in the holding cost h.
That is, high holding cost implies ordering more
frequently in small quantities.
• In fact, Q* explicitly balances ordering and holding
cost, since for Q = Q* the holding and fixed ordering
costs per unit time are equal: hQ*/2 = KD/Q*
 Inclusion of Order Lead Time
• If an order requires a lead time L > 0 to arrive.
Then, the optimal policy is to order Q*, L time
units before the inventory level reaches zero.
• Specifically, the reorder inventory level is DL if
L < T*.
• If L > T*, the reorder level is R=D(fractional part of
L/T*).
 EOQ facts
• An important property of EOQ is that the optimal cost, CU(Q), is not too
sensitive to Q when Q is close to Q*.
• Specifically, a relative deviation of q from Q* by using an order quantity Q =
(1+q)Q* leads to a relative change of the cost per unit time (excluding variable
cost) of q2/ [2(1+q)].
• For example a deviation of 20% in Q* increase cost by 0.22/[2(1+0.2)] = 0.017,
by less than 2%.
• For example a deviation of 20% in Q* increase cost by 0.22/[2(1+0.2)] = 0.017,
by less than 2%.
 Example 1
• Consider a three-ohm resistor used in the assembly of an automated
processor of X-ray films. The demand for this item has been relatively
level over time at a rate of 2,400 units/year. The unit holding cost of the
resistor is 0.096 $/unit/year, and the fixed ordering cost is $3.2.
• What is the optimal order quantity and corresponding annual cost?
2 KD 2  $3.2  2400 units/year
Q* = = = 400 units.
h $0.096 unit/year
CU * = 2 KDh = 2  $3.2  2400 units/yr  $0.096 unit/year
= $38.4 / year

• Note that the optimal order quantity involves ordering 2-month supply of
the resistor every two months (since T* = Q*/D = 2/12 years = 2 months).
• Note also that the optimal order cost KD/Q* = $19.2 is equal to the
optimal holding hQ*/2 = $19.2.
 Example 1 (continued)
• Suppose that it takes a lead time of one week to receive an order. How
should the resistor be ordered?
• The optimal policy is to order 400 units when the inventory level reaches
DL = 2400/52 ≈ 46 units.
• Suppose the supplier will not deliver below a three-month supply of the
transistor? How will this affect the cost?
• This involves using a Q = 3/2 Q* instead of Q*, which will increase the
annual cost by 0.52/[2(1+0.5)] = 0.083 ≈ 8%.
D Q
• This can be verified by evaluating CU (Q) = K Q + h 2
• for Q = 600, which gives CU(600) = $41.6, which is (41.6 − 38.4)/38.4 =
0.083 = 8.3%
 Example 1 (continued)
• Finally, it is instructive to plot the annual ordering,
holding, and total cost on the same graph.
100

80

CU( Q)

HU( Q) 60

OCU( Q)
40

20

200 400 600 800

Q
 EOQ model with a finite production
rate (EPQ model)
• In certain systems, especially production systems, the whole order quantity is not
delivered at the same time.
• The batch is instead delivered continuously according to a certain production rate
P per unit time.
• The economic production quantity model (EPQ) is the EOQ model but with orders
delivered according to a rate P > D.
• Adopting an order-at-zero-inventory policy order quantities all equal to Q, the
inventory over time varies as follows

Q(1-D/P)
P-D D
1 1

Q/P Time
T=Q/D
 Q(1-D/P)
P-D D
1 1

Q/P Time
T=Q/D

𝐷
𝐾+𝑐𝑄 ℎ𝑄 1−𝑃 𝑇 𝐾𝐷 ℎ𝑄 𝐷
• The cost per unit time is 𝐶𝑈 𝑄 = + = + 1− =
𝑇 2𝑇 𝑄 2 𝑃
𝐾𝐷 ℎ′ 𝑄 𝐷
+ where ℎ′ = ℎ(1 − )
𝑄 2 𝑃

2𝐾𝐷
𝑄 ∗= , 𝐶𝑈 ∗= 2𝐾𝐷ℎ(1 − 𝐷/𝑃)
ℎ(1 − 𝐷/𝑃)

• The optimal order quantity for the EPQ model is less than that of the EOQ
model because inventory spends less time on hand in the EPQ model,
which reduces holding cost.
• As P →∞, Q* converges to the EOQ order quantity
 Example 2
• Suppose that the resistor in Example 1 is replenished according to a production
rate of 900 per month. What are the optimal order quantity and annual cost?
• In this case, P = 800 units/month = 9,600/year. Then, D/P = 1/4, and

2 KD 2  3.2  2400
Q* = =  462 units.
h(1 − D / P) 0.096(1 − 1/ 4)
CU * = 2 KDh(1 − D / P) = 2  3.2  2400  0.096(1 − 1/ 4)
= $33.3/ year
EOQ model with shortages allowed
• Suppose now that in the EOQ model shortages are allowed and
backordered.
• Suppose that the unit backorder cost is p $/unit/unit time. This can be
seen as a waiting cost.
• The shortage cost per unit time is proportional to the average backorder
level (similar to the holding cost).
• With a policy that orders in equal amounts, of size Q, the problem is to
determine the optimal value of Q and of the backorder level S at the end
of an ordering cycle. Inventory

Q-S

S Time

T=Q/D
Inventory

Q-S

S Time

T=Q/D

• The average inventory level is {(Q − S ) [(Q − S ) / D]/ 2}/ T = (Q − S )2 / 2Q

• The average backorder level is {S  [S / D]/ 2}/ T = S 2 / 2Q
• Therefore, the cost per unit time is D (Q − S )2 S2
CU (Q, S ) = K + cD + h + .
Q 2Q 2Q
• It can be easily shown that CU(Q,S) is convex in S (for a fixed value of Q) , since
CU ( S , Q) h(Q − S )  S
=− + ,
S Q Q
 2CU (Q) h + 
=  0.
S 2 Q
• Then, the optimal value of S that minimizes cost for Q fixed, is obtained by setting
the first derivative equal to zero as S * = Q h .
 +h
Inventory

Q-S

S Time

T=Q/D

• Now the cost “at optimal S” is

D Q{1 − [h /( + h)]}2 Q[h /( + h)]2

CU (Q) = K + cD + h +
Q 2 2
D Q[ /( + h)]2 Q[h /( + h)]2
= K + cD + h +
Q 2 2
D Q h
= K + cD +
Q 2  +h
• Therefore, the optimal order quantity is found similar to the EOQ model as

2 KD 2 KD  + h
Q* = = .
 h /( + h) h  h 2 KD h
• In addition, the optimal backorder level is S* = Q * =
 +h   +h
Inventory

Q-S

S Time

T=Q/D

2 KD 2 KD  + h
Q* = = .
 h /( + h) h 

h 2 KD h
S* = Q * =
 +h   +h

CU * = cD + [ /( + h)] 2 KDh .

• Note that the optimal order quantity increases relative to the EOQ model order
quantity as a result of backorders.
• Note also that as p →∞, Q* converges to the EOQ order quantity.
 Multi item inventory Models
• So far we have only considered inventory models with a single product. A more
realistic assumption is that we must simultaneously manage the inventory levels of
a variety of products. It is easy to show that if there are no joint order costs or
capacity constraints, then a problem with multiple items, each facing constant
demand, can be handled by solving each item’s EOQ problem separately.
• In reality, however, when we deal with multiple items, we often must be
concerned with not exceeding the capacity of the warehouse. In other words,
there is a limit on the total volume of inventory that may be held in the warehouse
at any point in time, and this limit may not be exceeded.
• Suppose there are n items, Item i consumes 𝑤𝑖 per unit and the total space
available is 𝑊. Let 𝑄𝑖 be the order quantity for item i. Note that 𝑄𝑖 is the
maximum inventory of item i that will be in the warehouse at any time. Then the
warehouse capacity constraint can be written as σ𝑛𝑖=1 𝑤𝑖 𝑄𝑖 ≤ 𝑊
 Multi item inventory Models
• There are two possible cases to consider:
• If σ𝑛𝑖=1 𝑤𝑖 𝐸𝑂𝑄𝑖 ≤ 𝑊, then the optimal
∗
solution is 𝑄𝑖 = 𝐸𝑂𝑄𝑖
• If σ𝑛𝑖=1 𝑤𝑖 𝐸𝑂𝑄𝑖 > 𝑊, then the current
solution is no longer feasible.
 Multi item inventory Models
• In this case we need to find the solution for
this current problem:
𝑛 𝐾𝑖 𝐷𝑖 ℎ𝑖 𝑄𝑖
• min σ𝑖=1 +
𝑄𝑖 2
• Subject to
• σ𝑛𝑖=1 𝑤𝑖 𝑄𝑖 ≤ 𝑊
• In these types of problems we can construct the
lagrangian function:
𝐾𝑖 𝐷𝑖 ℎ𝑖 𝑄 𝑖
• 𝐿 𝑄1 , 𝑄2 , … , 𝑄𝑛 , 𝜆 = σ𝑛𝑖=1 𝑄𝑖
+ 2
+ 𝜆(σ𝑛𝑖=1 𝑤𝑖 𝑄𝑖 − 𝑊)
• The optimal solution will satisfy:
𝜕𝐿 𝜕𝐿
• 𝜕𝑄𝑖
= 0, 𝜕𝜆 = 0

𝜕𝐿 2𝐾𝑖 𝐷𝑖
• 𝜕𝑄𝑖
= 0 ⇒ 𝑄𝑖 = ℎ𝑖 +2𝜆𝑤𝑖
𝜕𝐿
• 𝜕𝜆
= 0 ⇒ σ𝑛𝑖=1 𝑤𝑖 𝑄𝑖 = 𝑊
𝑤
• Consider the case where ℎ 𝑖 = 𝑤/ℎ. The holding cost is proportional to the size of each
𝑖
item. In this case:
2𝐾𝑖 𝐷𝑖 2𝐾𝑖 𝐷𝑖 1 𝑊
• 𝑄𝑖 = ℎ𝑖 +2𝜆𝑤𝑖
= ℎ𝑖 1+2𝜆𝑤/ℎ
= 𝑚𝐸𝑂𝑄𝑖 , 𝑚 = σ𝑛
𝑖=1 𝑤𝑖 𝐸𝑂𝑄𝑖
 Example
• Three items are produced in a small fabrication
shop. The shop management has established the
requirement that the shop never have more than
$30,000 invested in inventory of these items at
one time. The management uses a 25% annual
interest rate charge to compute the holding cost.
The relevant cost and demand parameters are
given in the following table:
• item 1 2 3
Demand rate 1850 1150 800
Variable cost 50 350 85
Setup cost 100 150 50
 Solution
2×100×1850
• 𝐸𝑂𝑄1 = = 172, 𝐸𝑂𝑄2 =
0.25×50
63, 𝐸𝑂𝑄3 = 61
• 172 × 50 + 63 × 350 + 61 × 85 =
$35,835 > $30,000

• 𝑄1∗ = 172 × 0.8371 = 144, 𝑄2∗ = 52, 𝑄3∗ = 51

EOQ model with quantity discounts
• Suppose that in the EOQ model the supplier offers a price discount for large
quantities.
 c1 , if Q  Q
• Specifically, suppose the unit variable order cost is c (Q ) = 
c2 , if Q  Q
where c2 < c1 .
• In addition, since a major part of the holding cost is related to cost of capital
(which is proportional to variable cost) the holding cost will be of the form
 h , if Q  Q
h= 1
h2 , if Q  Q

• The cost per unit time is then given by

 D Q
 K + c1D + h1 , if Q  Q
C (Q ), if Q  Q  Q
1
2
C U (Q ) = 
U
=
C (Q ), if Q  Q K D + c D + h Q , if Q  Q
2
U
 Q 2 2
2
  D Q
 K + c1D + h1 , if Q  Q
C (Q ), if Q  Q  Q
1
2
C U (Q ) = 
U
=
C (Q ), if Q  Q K D + c D + h Q , if Q  Q
2
U
 Q 2 2
2

• Then, the optimal order quantity is either the minimum of C 1

U (Q) for Q  Q , Q ,
1
or the minimum of UC 2
(Q ) for Q  Q , Q , whichever gives the least cost.
2

Q1 = min(Q, 2 KD / h1 )

Q 2 = max(Q , 2KD / h2 )
 Example
• Because of convenience in manufacturing and shipping, the supplier of an item offers a
2% discount on any order of 100 units or higher. Demand for this item is estimated at
416 units per year. The unit variable cost is $14.2, and the fixed cost is $1.50. For this
item storage costs are negligible, and the unit holding cost is 24% (the cost of capital) of
the unit variable cost.
• What are the optimal order quantity and annual cost for the item?
• Here, K = $1.5, 𝑐1 = $14.2, and ℎ1 = 0.24×14.2 = $3.41. In addition, 𝑐1 = 0.98×14.2 =
$13.92, and ℎ2 = 0.24×13.92 = $3.34.
Q1 = min(100, 2 1.5  416 / 3.41)  19
CU1 (Q1 ) = 2 1.5  416  3.41 + 14.2  416  $5,972

Q2 = max(100, 2 1.5  416 / 3.34) = 100

416 100
CU2 (Q2 ) = 1.5  + 13.92  416 + 3.34   $5,964
100 2

• The optimal order quantity is therefore Q* = 100, and the corresponding annual cost is
$5,964.
 Incremental Quantity Discounts
• Consider now the case of incremental Quantity
Discounts:
• Example: Trash bags cost 30 cents for quantities
of 500 and fewer, for quantities between 500 and
1000, the first 500 cost 30 cents and the
remaining cost 29 cents each; for quantities over
1000, the first 500 cost 30 cents each, the next
500 cost 29 cents each, and the remaining cost 28
cents each.
• K=8,D=600,h=20%
 Example (continued)
𝐾𝐷 ℎ𝑄
• 𝐶𝑈 𝑄 = 𝑄 + 𝑐𝐷 + 2 , where 𝑐 = 𝑎𝑣𝑒𝑟𝑎𝑔𝑒 𝑐𝑜𝑠𝑡 𝑝𝑒𝑟 𝑖𝑡𝑒𝑚, ℎ =
𝑎𝑣𝑒𝑟𝑎𝑔𝑒 ℎ𝑜𝑙𝑑𝑖𝑛𝑔 𝑐𝑜𝑠𝑡 𝑝𝑒𝑟 𝑖𝑡𝑒𝑚
• If 𝑄 ≤ 500, 𝑐1 = 0.3, ℎ1 = 0.2 × 0.3 = 0.06
0.3×500+0.29× 𝑄−500
• If 500 < 𝑄 ≤ 1000, 𝑐2 = 𝑄
= 0.29 + 5/𝑄, ℎ2 = 0.2 ×
5
0.29 + = 0.058 + 1/𝑄
𝑄
0.3×500+0.29×500+0.28(𝑄−1000) 15
• If 𝑄 ≥ 1000, 𝑐3 = 𝑄
= 0.28 + 𝑄
, ℎ3 = 0.2 × ቀ0.28 +
15 3
ቁ = + 0.056
𝑄 𝑄

600 𝑄
8× 𝑄
+ 0.3 × 600 + 0.06 × 2 , 𝑄 ≤ 500
1
600 5 0.058+𝑄 𝑄
• 𝐶𝑈 𝑄 = 8× 𝑄 + 0.29 + 𝑄 × 600 + , 500 < 𝑄 ≤ 1000
2
600 15 3
8× 𝑄 + 0.28 + 𝑄 × 600 + 𝑄
+ 0.056 𝑄/2, 𝑄 > 1000
 Example (continued)
600 𝑄
8× + 0.3 × 600 + 0.06 × , 𝑄 ≤ 500
𝑄 2
1
0.058+ 𝑄
• 𝐶𝑈 𝑄 = 8 × 600 + 0.29 + 5 × 600 + 𝑄
, 500 < 𝑄 ≤ 1000
𝑄 𝑄 2
600 15 3
8× + 0.28 + × 600 + + 0.056 𝑄/2, 𝑄 > 1000
𝑄 𝑄 𝑄
• Set derivative with respect to Q equal to 0:
• 𝑄1∗ = 400, 𝑄2∗ = 519, 𝑄3∗ = 702
• 𝑄3∗ < 1000 so it is not in the range, so omit it
• Compare 𝐶𝑈 1 𝑄1∗ = 204.0, 𝐶𝑈2 𝑄2∗ = 204.58
• Optimal Quantity is 𝑄 ∗ = 𝑄1∗ = 400 with cost 𝐶𝑈 1 𝑄1∗ = 204.0