Single Period Inventory (or

Newsvendor) Model
Recap from I Year and Extension

In this course we will use the phrase Newsvendor Model

1

Newsvendor Trade-off
Need to take a gamble today for a Future Demand
Risks (due to Uncertainties)
– Order too Much
– Order too Little

1
 Short Life Cycle Product: The Principal Trade-Off

Ordering Too Much

– Inventory left over at Period End
– (Salvage) Inventory sold at Loss
Ordering Too Little
– Not all Demand is satisfied
– Losing out on Revenue

Please refer to Chapter 13 of Textbook for Details

or First Year text
3

Problem Parameters

Wholesale price: W; Selling Price: R

Salvage value: S (S<W)
Demand: D (random variable)
Retailer orders Q units (decision variable)
“Perishable Good”
Excess Demand is Lost
Only One Time Opportunity to Buy/Order
We ignore Inventory Carrying cost (Why?)
How much should the retailer order (that is
Q)?
4

2
 Newsvendor model - Properties
One-time decision
Current decisions only impact the “next period” but not
future periods
Applications in fashion/seasonal items, electronics,
booking capacity (such as from contract manufacturers,
transportation services, etc.), media (online and TV
ads)
Relevant costs
– Co: Unit cost of excess inventory (cost of overstocking)
– Cu: Unit cost of shortage (cost of understocking)
Objective: Maximize Profit
5

Problem Analysis
Determine Q, such that the Expected Profit is
Maximized

Cu
Pr( D  Q * ) 
Cu  C o

C u  R W ;C o  W  S

Cost of Having too Little Cost of Having too Much

(Understocking) (Excess Inventory or Overstocking)

3
 An Example
Demand: Normally distributed with a mean = 350
units and standard deviation of  = 150

Purchase price (W) = Rs.100

Retail price (R) = Rs. 250
Salvage value (S) = Rs. 80

How many units should be ordered?

An Example
Cu = 150
Co = 20
Critical ratio = Cu/(Cu+Co) = 0.8824
Pr(D Q* ) = 0.8824
Therefore, Q* = 528 units

What is Expected Sales (in units)?

4
 Performance Measures in Newsvendor
Problem

Suppose you have inventory of 528 units

(decision variable)

Demand turns out to be 575 units (realization of a

random variable)

Lost sales: ?? units

Excess Inventory: ?? units
Sales: ?? units

Performance Measures in Newsvendor

Problem

Suppose you have inventory of 528 units

(decision variable)

Demand turns out to be 575 units (realization of a

random variable)

Lost sales: 47 units

Excess Inventory: 0 units
Sales: 528 units

10

10

5
 Performance Measures in Newsvendor
Problem

Suppose you have inventory of 528 units

(decision variable)

Demand turns out to be 508 units (realization of a

random variable)

Lost sales: ?? units

Excess Inventory: ?? units
Sales: ?? units

11

11

Performance Measures in Newsvendor

Problem

Suppose you have inventory of 528 units

(decision variable)

Demand turns out to be 508 units (realization of a

random variable)

Lost sales: 0 units

Excess Inventory: 20 units
Sales: 508 units

12

12

6
 Performance Measures in Newsvendor
Problem

Lost sales = maximum (D-Q,0)

– will happen when D > Q
Leftover Inventory = maximum (Q-D,0)
– will happen when D < Q

𝐷, 𝐷 𝑄
𝑆𝑎𝑙𝑒𝑠
𝑄, 𝐷 𝑄

Sales = minimum (D,Q)

13

13

Performance Measures in Newsvendor Problem

Demand = Sales + Lost Sales

D = min(D,Q)+max(D-Q,0)

Q = Sales + Left Over Inventory

Q=min(D,Q)+max(Q-D,0)

Q: is what the decision maker has ordered (Decision Variable)

D: is Demand (which is Random)
Also Note: Q need not be the Optimal Decision; it is the
inventory decided by the decision maker (optimal or not)
14

14

7
 Demand, Sales, Lost Sales, Leftover Inventory

D = min(D,Q)+max(D-Q,0)

Take Expectations on both sides:

𝐸 𝐷 𝐸 min 𝐷, 𝑄 𝐸 max 𝐷 𝑄, 0
𝐸 𝐷 𝜇 (where 𝜇 is the Mean Demand)

15

15

Demand, Sales, Lost Sales, Leftover Inventory

Similarly,
Q = Sales + Left Over Inventory

Q=E(min(D,Q))+E(max(Q-D,0))

Remember Q is a constant: a number (decision made by the

decision maker); therefore, E(Q) = Q

16

16

8
 Expected Profit 𝜋 𝑄

𝜋 𝑄 𝑅𝐸 𝑆𝑎𝑙𝑒𝑠 𝑆𝐸 𝐿𝑒𝑓𝑡𝑜𝑣𝑒𝑟 𝐼𝑛𝑣𝑒𝑛𝑡𝑜𝑟𝑦 𝑊𝑄

17

17

Expected Profit 𝜋 𝑄

𝜋 𝑄 𝑅𝐸 𝑆𝑎𝑙𝑒𝑠 𝑆𝐸 𝐿𝑒𝑓𝑡𝑜𝑣𝑒𝑟 𝐼𝑛𝑣𝑒𝑛𝑡𝑜𝑟𝑦 𝑊𝑄

𝑅𝐸 min 𝐷, 𝑄 𝑆𝐸 m𝑎𝑥 𝑄 𝐷, 0 𝑊𝑄

18

18

9
 Expected Profit in Newsvendor
Purchase Cost

Profit Π Q
𝑅𝐸 min 𝑄, 𝐷 𝑆𝐸 max 𝑄 𝐷, 0 𝑊𝑄

E(Sales) E(Revenue from Salvage)

E: Expectation Operator See Page 414-417 of Text

19

19

Expected Profit in Newsvendor

Profit Π Q 𝑅𝐸 min 𝑄, 𝐷 𝑆𝐸 max 𝑄 𝐷, 0 𝑊𝑄

𝑅 𝑊 𝐸 𝐷
𝐸 𝑅 𝑊 𝑀𝑎𝑥 𝐷 𝑄, 0 𝑊 𝑆 𝑀𝑎𝑥 𝑄 𝐷, 0

The transformation from the first to the second equation can be done
with some manipulation (proof not required for this course).

20

20

10
 Expected Profit in Newsvendor

Profit Π Q 𝑅𝐸 min 𝑄, 𝐷 𝑆𝐸 max 𝑄 𝐷, 0 𝑊𝑄

𝑅 𝑊 𝐸 𝐷
𝐸 𝑅 𝑊 𝑀𝑎𝑥 𝐷 𝑄, 0 𝑊 𝑆 𝑀𝑎𝑥 𝑄 𝐷, 0

Total Cost of Understocking Total Cost of Overstocking

21

21

Supply-Demand Mismatch Cost

• The mismatch cost (denoted as Cm) is the cost of

Overstocking (“too much” inventory) plus the
Understocking cost (“too little” inventory)
• Mismatch cost: Cm = Co* E(Leftover
Inventory)+Cu*E(Lost Sales)
• The maximum profit is the expected profit without
any mismatch costs, that is, every unit is sold and
there are no lost sales:
• Maximum Profit = (R-W)*E(Demand)
• When will you achieve Maximum Profit?
22

22

11
 Expected Profit in Newsvendor
Profit Π Q 𝑅𝐸 min 𝑄, 𝐷 𝑆𝐸 max 𝑄 𝐷, 0 𝑊𝑄

𝑅 𝑊 𝐸 𝐷
𝑅 𝑊 𝐸 𝑀𝑎𝑥 𝐷 𝑄, 0 𝑊 𝑆 𝐸 𝑀𝑎𝑥 𝑄 𝐷, 0

𝑅 𝑊 𝐸 𝐷 𝐶
where 𝐶 : 𝑖𝑠 𝑡ℎ𝑒 𝑆𝑢𝑝𝑝𝑙𝑦 𝐷𝑒𝑚𝑎𝑛𝑑 𝑀𝑖𝑠𝑚𝑎𝑡𝑐ℎ Cost

𝐶 𝑅 𝑊 𝐸 𝑀𝑎𝑥 𝐷 𝑄, 0 𝑊 𝑆 𝐸 𝑀𝑎𝑥 𝑄 𝐷, 0

23

23

Expected Profit in Newsvendor: One more Observation

Π 𝑄 𝑅 𝑊 𝐸 𝐷
𝐸 𝑅 𝑊 𝑀𝑎𝑥 𝐷 𝑄, 0 𝑊 𝑆 𝑀𝑎𝑥 𝑄 𝐷, 0
Π 𝑄 𝑅 𝑊 𝐸 𝐷 𝐶

Now 𝑀𝑎𝑥 Π 𝑄 𝑀𝑖𝑛 𝐶

Therefore Maximizing Profits is equivalent

to Minimizing Supply Demand Mismatch Costs

24

24

12
 If Demand is Normally Distributed

ELS  L( z )
Q * 
z


ELS: Expected Lost Sales

25

25

L(z) table

A copy is available on Moodle

26

26

13
 Leftover Inventory (or Overstock)
= Q – Demand, if Demand < Q
= 0, if Demand ≥ Q

Expected Sales    Expected Lost Sales

Expected Left Over Inventory  Q  Expected Sales

First compute ELS, then Expected Sales

and finally Expected Left-over Inventory

See Page 414-417 of Text

27

27

An Example
Cu = 150
Co = 20
Critical ratio = Cu/(Cu+Co) = 0.8824
Pr(D Q* ) = 0.8824 or Q* = 528 units
What is Expected Sales (in units)?
z = (528-350)/150 = 1.19
L(z) = 0.0561; 𝐸𝐿𝑆 𝜎𝐿 𝑧 8.42 𝑢𝑛𝑖𝑡𝑠
E(sales) = 350-8.42 =341.5 units

Note that Demand is Normally distributed with a

mean = 350 units and standard deviation of  = 150 28

28

14
 Impact of Improving Forecasts (Lower  )
Demand: Normally distributed with a mean = 350
units and standard deviation of  = 150
Purchase price = Rs.100
Retail price = Rs. 250
Salvage value = Rs. 80

How many units should be ordered as  changes?

29

29

Impact of Improving Forecasts

 Q* Expected Expected Expected
Overstock Understock Profit
150 528 186.7 8.6 47,469
120 492 149.3 6.9 48,476
90 457 112.0 5.2 49,482
60 421 74.7 3.5 50,488
30 386 37.3 1.7 51,494
0 350 0 0 52,500

Work out these numbers on your own to check the correctness

30

30

15
 When is the Mismatch Cost high?

Critical Ratio Mismatch cost as a percent of

the maximum profit increases
0.3 0.4 0.5 0.6 0.7 0.8 0.9 as
0.1 12% 10% 8% 6% 5% 3% 2%  the coefficient of
0.25 29% 24% 20% 16% 12% 9% 5%
variability (CV) of
0.4 46% 39% 32% 26% 20% 14% 8%
demand increases
CV 0.5 58% 48% 40% 32% 25% 17% 10%
𝑪𝒖
0.6 70% 58% 48% 39% 30% 21% 12%  the critical ratio
𝑪𝒖 𝑪𝒐
0.75 87% 72% 60% 48% 37% 26% 15%
decreases
0.8 93% 77% 64% 52% 40% 28% 16%

31

31

Performance Measures in Newsvendor Problem

• For any order quantity, Q (optimum or not)
– Expected fill rate
» The fraction of demand that is satisfied (from
Inventory)
=
– In-stock probability
» Probability that entire demand is satisfied Pr 𝐷 𝑄

32

32

16