Scribd
Upload
45 views5 pages

Stochastic Demand Models

Stochastic demand models use probability and statistics to model uncertain demand. There are three types: single period for one-time decisions; multiple period for periodic decisions; and continuous time for continuous decisions. The single period model uses the Newsvendor model to determine the optimal order quantity. The Newsvendor model calculates the costs of excess inventory and shortages to find the quantity that minimizes total costs. Stochastic models are solved using probability distributions like the normal distribution to represent uncertain demand.

Uploaded by

Nishant Kumar Gupta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
45 views5 pages

Stochastic Demand Models

Stochastic demand models use probability and statistics to model uncertain demand. There are three types: single period for one-time decisions; multiple period for periodic decisions; and continuous time for continuous decisions. The single period model uses the Newsvendor model to determine the optimal order quantity. The Newsvendor model calculates the costs of excess inventory and shortages to find the quantity that minimizes total costs. Stochastic models are solved using probability distributions like the normal distribution to represent uncertain demand.

Uploaded by

Nishant Kumar Gupta
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Stochastic Demand Models:

2009, Notes Created by Sil

What are Stochastic Models?  Mathematical models involving probability.


Probability Distribution is used to represent uncertain factors.
Stochastic Models are based on Expected Values (long-run average of all possible outcomes).

If Demand is known  Deterministic Case


If Demand is unknown/uncertain (demand is a random variable)  Stochastic Case

3 Types of Stochastic Models:


1) Single Period  One time decision (How much to order)
E.g.: Fashion goods, perishable goods, goods with short lifecycles, seasonal goods.

2) Multiple Period  Periodic decision (How much to order in each period)


E.g.: Goods with recurring demand but whose demand varies from period to period; Inventory systems
with periodic reviews.

3) Continuous Time  Continuous decision (Continuously deciding the order quantity)


E.g.: Goods with recurring demand but with variable inter-arrival time between customer orders;
Inventory systems with continuous reviews.

(1)Single Period Uses “Newsvendor Model”


The newsvendor problem has numerous applications for decision making in manufacturing and service industries as well
as decision making by individuals. It occurs whenever the amount needed of a given resource is random, a decision must
be made regarding the amount of the resource to have available prior to finding out how much is needed, and the
economic consequences of having “too much” and “too little” are known.

The Newsvendor Problem:


Service Level Probability of
to meet the Optimal satisfying demand
Amount to Order during the period
P= Probability
D= Random Demand
q= Capacity Qty

= Cost of ordering too little


= Selling Price (SP) – Purchase Price (PP) = Lost Profit = Shortage Cost = Opportunity Cost

= Cost of ordering too much


= Purchase Price (PP) – Disposal Price or Salvage Value (SV) = Loss per excess = Excess Cost

Optimal Qty to Stock or to Order:

Find the by looking at the Normal Distribution


Normal Distribution:
% from Service Level  then look for
the Z-value on the Z-table

X
Z
0 Z

Normal or Discreet?

Normal Distribution:
If the distribution of demand is known to be normal (Normal distribution) with mean c and standard deviation
c we use the condition (usually given):

The Normal Distribution graph looks something like that:

And then the Optimal Ordering Level is:


q = F-1(Q*, c, c) OR q

In Excel: q = NORMINV(q, c, c)

Discreet Distribution
If the distribution of demand is discreet (Discreet distribution), we use the following condition:

And then the Optimal Ordering Level is:


Look at the table where there are three columns:

Probability that Demand Qty will be Demand

Find the Cumulative frequency that is the closest to the


service level. The corresponding demand will be your
Optimal Order Quantity

The discreet distribution graph looks something like that:

The SUM of all the Frequencies should be 100% always!


Applications for DESC372:

Simple Newsvendor problem


Given:
SP=$2
PP=$1
SV=$0.25

What is the optimal number of newspapers one should buy?

Shortage Cost = SP – PP  $2 - $1 = $1
Excess Cost = PP – SV  $1 - $0.25 = $0.75

or 57.14%

So the optimal order quantity should meet 57.14% of demand.


(Check the cumulative frequency and its corresponding quantity)

The closest to 57.14% is 55%, so the Optimal


Qty to Order will be 22 newspapers.
Example 15-2: Revenue Management- Allocating capacity to multiple segments
Given:
A trucking company serves two segments of customers.
A. Willing to pay $3.50 per cubic foot but wants to commit to a shipment with only 24 hours notice.
B. Willing to pay $2.00 per cubic foot and is willing to commit to a shipment with up to one week notice.
With two weeks to go, demand for (A) is forecast to be normally distributed, with a mean 3,000 cubic feet and
a standard deviation of 1,000.

a) How much of the available capacity should be reserved for (A)?


b) How should the company change its decision if (A) is willing to pay $5 per cubic foot?

Rearrange given info


Revenue:
R(A)=$3.50/cubic foot
R(B)=$2.00/cubic foot

Mean( ) = 3000 cubic feet


Standard deviation( ) = 1000 feet

Risks:
1) Spoilage: Too much capacity available because we were reserving the space for the higher price buyers
2) Spill: You cannot accommodate the higher price buyers because you already sold your space to lower
price buyers.

Prob that Capacity reserved for Higher price buyers is to the actual demand for higher price buyers

In Excel 1 - 0.5714 0.5714


q =NORMINV(Q*, , )
=NORMINV(0.5714, 3000, 1000) = 3179.9396
But it has to be the other side so 1-Q*:
=NORMINV(1 - 0.5714, 3000, 1000) = 2820.0604

Manually -0.18

But it has to be the other side so the negative of z:


0.40
So 2820 cubic feet should be reserved for segment (A)

If (A) is willing to pay $5 instead of $3.50:

=NORMINV(1 - 0.40, 3000, 1000) = 3253.3471


OR 0.25

So 3250 cubic feet should be reserved for segment (A) is they are willing to pay $5 instead.
Example 15-5: Revenue Management- Overbooking
Given:
An apparel supplier is taking orders for dresses with a Christmas motif.
Production capacity available from the supplier is 5,000 dresses, and it makes $10 for each dress sold.
The supplier is currently taking orders from the retailers and must decide on how many orders to commit to at
this time. If it has orders that exceed capacity, it has to arrange for backup that results in a loss of $5 per
dress.
Retailers have been known to cancel their orders near the winter season as they have better visibility into
expected demand.

a) How many orders should the supplier accept if cancellations are normally distributed, with a mean of
800 and a standard deviation of 400?
b) How many orders should the supplier accept if cancellations are normally distributed, with a mean of
15% of the orders accepted and a coefficient of variation of 0.5?

SOLUTION:
a) How many should we overbook?

Prob that Cancellations are going to be the optimal overbooking level

In Excel 0.6667
O* =NORMINV(S*, , )
=NORMINV(0.6667, 800, 400) = 972.3276 dresses

Manually

0.43
So 972 dresses should be overbooked thus
Suppliers should accept the capacity (5,000) + Overbooking limit (972) = 5973units accepted

b) How many should be accepted with new mean and std deviation?

In Excel 0.6667
O* =NORMINV(S*, , )
=NORMINV(0.6667,0.15*(5973),0.075*(5973))=1088.9461

Manually
0.43
So 1088 dresses should be overbooked thus
Suppliers should accept the capacity (5,000) + Overbooking limit (1088) = 6088units accepted

You might also like