Lesson 04

Decision Making

The operations manager is a planner

and a decision maker in
environments of risk, uncertainty and
certainty.

04 - 1
 The Decision Process

Identify the Problem

Specify objectives and the
criteria for choosing a solution
Develop alternatives

Analyze and compare alternatives

Select “best” alternative (“best” depends on
whether we are considering costs or profits)
Implement chosen alternative
Monitor to ensure desired results are achieved

04 - 2
 Causes for Poor Decisions
There are many reasons for making poor decisions
. Mistakes - in the decision process (either logic in formulating
the problem, solution, or calculation errors)
. Bounded rationality - limitations from costs, human
abilities, time, technology and availability of information
. Sub-optimization - optimum solutions at the
departmental level may not be in the best interest
of the department rather than in the best interest
of the whole organization

04 - 3
 Decision Environments
There are 3 environments under which decisions are made
Certainty - Environment in which relevant parameters have known values (eg – Profit =
$5/unit, demand = 200 units. How much profit will you make?)
Risk - Environment in which probability estimates of possible future conditions are
evaluated (eg – Profit = $5/unit, demand has a probability of 50% for 100 units and 50%
for 200 units, How much profit will you make?)
Uncertainty - Environment in which it is impossible to assess the likelihood of various
future events (eg – Profit = $5/unit, demand is unknown. How much profit will you
make?)

04 - 4
 Decision Theory
Decision Theory represents a general quantitative approach to decision
making which is suitable for a wide range of operations management
decisions. Decisions are based on
. States of Nature - A set of possible future conditions that will
have a bearing on the results of the decision
. Alternatives – a list of considerations
. Payoff - A known result (could be good or bad) for
each alternative under each condition (state of nature)
. Likelihood – estimated probability of each future condition
. Decision criteria – a set of rules under which decisions are
made to select the best alternative

04 - 5
 Payoff Table
Payoff - A known result (could be good or bad) for each alternative
under each possible future condition. Good results are usually shown as
positive numbers. Bad results (costs or losses) are usually shown as
negative numbers.
Payoff Possible Future Demand
Alternative Low Moderate High
Small Facility 10 10 10
Medium Facility 7 12 12
Large Facility -4 2 16

04 - 6
 Regret (Opportunity Loss) Table
Regret (Opportunity Loss) – is based on the payoff table. The regret or
lost opportunity is the difference between the best alternative payoff and
each alternative payoff for each possible future demand.
Regret Possible Future Demand
Alternative Low Moderate High
Small Facility 0 2 6
Medium Facility 3 0 4
Large Facility 14 10 0
The best alternative payoff for low demand is 10. The regret for each
alternative is 0 (10 – 10) for a small facility, 3 (10 – 7) for a medium
facility, and 14 (10 - - 4) for a large facility.
04 - 7
 Decisions Under Certainty
Decisions under certainty (known future condition) are made by selecting
the alternative which has the best payoff.

Payoff Possible Future Demand

Alternative Low Moderate High
Small Facility 10 10 10
Medium Facility 7 12 12
Large Facility -4 2 16
If demand is low the best alternative is a Small facility
If demand is moderate the best alternative is a Medium facility
If demand is high the best alternative is a Large facility
04 - 8
04.1 Certainty

04 - 9
Decision Rule: For each state of nature, choose the alternative
which has the highest payoff .
Note: The alternative with the highest payoff is also the
alternative with the least regret.

For low demand: choose small facility

For moderate demand: choose medium facility

For high demand: choose large facility

04 - 10
 Decisions Under Uncertainty
Decisions under uncertainty (unknown future condition) are made by
selecting the alternative which has the best payoff based on one of the
following decision criteria.
maximin - “best worst” payoff establishes the minimum
outcome.
maximax - “best best” payoff establishes the best
possible outcome
Laplace - “best average” payoff establishes the average
payoff assuming each future condition is equally likely.
minimax regret (opportunity loss) - “best worst” regret
minimizes the difference between the realized
payoff and the best payoff for each future condition.
04 - 11
 Uncertainty - maximin
maximin - determine the worst payoff for each alternative across all
states of nature and select the alternative with the “ best worst” payoff

Payoff Possible Future Demand

Worst
Alternative Low Moderate High Payoff
Small Facility 10 10 10 10
Medium Facility 7 12 12 7
Large Facility -4 2 16 -4
The worst payoff for a small facility is 10, the worst payoff for a medium
facility is 7 and the worst payoff for a large facility is –4. The maximin
decision (“best worst”) is 10. Therefore, we would choose the small
facility alternative.
04 - 12
 Uncertainty - maximax
maximax - determine the best payoff for each alternative across all
states of nature and select the alternative with the “ best best” payoff

Payoff Possible Future Demand

Best
Alternative Low Moderate High Payoff
Small Facility 10 10 10 10
Medium Facility 7 12 12 12
Large Facility -4 2 16 16
The best payoff for a small facility is 10, the best payoff for a medium
facility is 12 and the best payoff for a large facility is 16. The maximax
decision (“best best”) is 16. Therefore, we would choose the large
facility alternative.
04 - 13
 Uncertainty - Laplace
Laplace - determine the average payoff for each alternative across all
states of nature and select the alternative with the “ best average” payoff

Payoff Possible Future Demand

Average
Alternative Low Moderate High Payoff
Small Facility 10 10 10 10.00
Medium Facility 7 12 12 10.33
Large Facility -4 2 16 4.67
The average payoff for a small facility is 10, the avearge payoff for a
medium facility is 10.33 and the average payoff for a large facility is 4.67.
The Laplace decision (“best average”) is 10.33. Therefore, we would
choose the medium facility alternative.
04 - 14
 Uncertainty – minimax regret
minimax regret (opportunity loss) - determine the worst regret for each
alternative and select the alternative with the “best worst” regret

Regret Possible Future Demand

Worst
Alternative Low Moderate High Regret
Small Facility 0 2 6 6
Medium Facility 3 0 4 4
Large Facility 14 10 0 14
The worst regret for a small facility is 6, the worst regret for a medium
facility is 4 and the worst regret for a large facility is 14. The minimax
regret decision (“best worst”) is 4. Therefore, we would choose the
medium facility alternative.
04 - 15
04.2 Uncertainty

04 - 16
Maximin Decision Rule: Calculate the maximum payoff for each alternative then choose the
alternative with the maximum payoff: Choose Small Facility
Minimax Decision Rule: Calculate the minimum payoff for each alternative then choose the
alternative with the minimum payoff: Choose Large Facility
Laplace Decision Rule: Calculate the average payoff for each alternative then choose the
alternative with the maximum average payoff: Choose Medium Facility
Minimax Regret Decision Rule: Calculate the minimum regret for each alternative then choose
the alternative with the least minimum regret: Choose Medium Facility
04 - 17
 Decisions Under Risk
Decisions under risk involve evaluating outcomes where probabilities of
future conditions are estimated. The decision criteria under which the
decisions are made is called the Expected Monetary Value (EMV). The
EMV determines the “best expected” payoff across all states of nature.

EValternative   payoff * probability

state of nature

EMV  best ( EValternative 1 ,..., EValternative n )

04 - 18
 Decisions Under Risk
EValternative   payoff * probability
state of nature

Payoff Possible Future Demand

Probability 0.3 0.5 0.2
Expected
Alternative Low Moderate High Payoff
Small Facility 10 10 10 10.00
Medium Facility 7 12 12 10.50
Large Facility -4 2 16 3.00

EMV  best ( EValternative 1 ,..., EValternative n )

04 - 19
04.3 Risk

04 - 20
Decision Rule Payoff Table: Calculate the expected (weighted average using the probability as
weights) payoff for each alternative then choose the alternative with the maximum expected
payoff: Choose Medium Facility
Decision Rule Regret Table: Calculate the expected regret for each alternative then choose the
alternative with the least expected regret: Choose Medium Facility
Note: Whether you use the payoff or regret table, the decision is the same!

04 - 21
 Decision Trees
A Decision Tree is a schematic representation of the options available to
a decision maker. The tree shows the decision alternatives as well as
the possible future conditions (states of nature) for each alternative.
Decision trees use the following conventions:
. Square nodes represent decision points
. Round nodes represent chance events
. Decision tree is read from right to left
. Decision tree is analyzed from left to right

The decision criteria under which the decisions are made is called the
Expected Monetary Value (EMV). The EMV determines the “best
expected” payoff across all states of nature

04 - 22
 General Format of a Decision Tree
Decision Point
Chance Event
f n atu re 1 Payoff 1
o
State
A’ Payoff 2
Choose 1
State o
’ f nature
e A 2 2
oos
Ch Choose A
’2 Payoff 3

1 Initial Decision Payoff 4

A’
Choose 3
Ch

n ature 1 2
oo

o f
State
se

Choose A
A’ 2

’4 Payoff 5

State o
f nature Payoff 6
2

04 - 23
 Decisions Using Decision Trees
Example 5: A manager must decide on the size of a video arcade to
construct. The manager has narrowed the choices to two: Large or
Small. Information has been collected on payoffs, and a decision tree
has been constructed. Analyze the decision tree and determine which
initial alternative (build small or build large) should be chosen to
maximize the Expected Monetary Value (EMV).
The decision tree showing the alternatives and possible states of nature
are shown on the following slide.

04 - 24
 Decisions Using Decision Trees
First, determine the “best” payoff for each state of nature
$40
( .4)
an d $40
e m
D
Lo w t hing
o No
D e
ll High D Overtim $50
a emand
Sm (.6) 2
u ild Expand $55
B
1 ($10)
ing
Do Noth
.4)
Bu

nd ( 2
ema
ild

D
Low
La

Reduce P $50
r

rices
ge

High D
emand $70
( .6)
04 - 25
 Decisions Using Decision Trees
Next, determine the “expected” payoff for each alternative

EVsmall  40(.4)  55(.6)  16  33  49

EVl arg e  50(.4)  70(.6)  20  42  62

Finally, determine the EMV “best expected” payoff

EMV  best ( EVsmall , EVl arg e )  best (49,62)  62

04 - 26
 Expected Value of Perfect Information (EVPI)
The Expected Value of Perfect Information (EVPI) is the difference between the
expected payoff under certainty (known outcomes) and the expected payoff
under risk. It is a very useful decision criteria when a future condition is
pending.
For example: There is discussion regarding whether or not to build a Johns
Island Connector similar to the James Island Connector. If a bridge is built
connecting Johns Island, property values on the Johns Island will increase.
An investor may want to take an option on property on Johns Island contingent
on the bridge construction. If the bridge is eventually constructed the investor
can choose whether to exercise the option.
EVPI can help the investor determine how much greater the expected payoff
due to delaying a decision to buy property on Johns Island is versus making a
decision based on current risks. (i.e. the expected payoff above the expected
monetary value)
04 - 27
 Expected Value of Perfect Information (EVPI)
The Expected Value of Perfect Information (EVPI) can be calculated
using two different methods. Method 1 involves the “ expected” payoff.

EVPI = Expected value of = Expected payoff - Expected payoff

perfect information under certainty under risk

EMV  best ( EValternative 1 ,..., EValternative n )

EVcertainty   ( best payoff ) * probability
state of nature

04 - 28
 Expected Value of Perfect Information (EVPI)
The Expected Value of Perfect Information (EVPI) can be calculated
using two different methods. Method 2 involves the “ expected” regret.

EVPI = Expected value of = Expected regret

perfect information under risk

EMV  best ( EValternative 1 ,..., EValternative n )

04 - 29
 Decisions using EVPI – “expected” payoff
Example 6: Determine the Expected Value of Perfect Information (EVPI)
for the payoff table shown below using “expected” payoff.

Payoff Possible Future Demand

Probability 0.3 0.5 0.2
Alternative Low Moderate High
Small Facility 10 10 10
Medium Facility 7 12 12
Large Facility -4 2 16

04 - 30
First, determine the “maximum” payoff for each alternative across each state of
nature under certainty
Payoff Possible Future Demand
Probability 0.3 0.5 0.2
Alternative Low Moderate High
Small Facility 10 10 10
Medium Facility 7 12 12
Large Facility -4 2 16
Best 10 12 16

Then calculate the “expected” payoff (EV under certainty) using the probability
weights and the expected payoffs.
EVcertainty   ( best payoff ) * probability
state of nature

 10(.30)  12(.50)  16(.20)  12.2

04 - 31
Next, determine the “expected” payoff for each alternative across each state of
nature under risk

Payoff Possible Future Demand

Probability 0.3 0.5 0.2
Expected
Alternative Low Moderate High Payoff
Small Facility 10 10 10 10.00
Medium Facility 7 12 12 10.50
Large Facility -4 2 16 3.00

Then determine the “best” expected payoff (EMV)

EMV  best ( EVsmall , EVmedium , EVl arg e )

 best (10,10.5,3)  10.5
04 - 32
 Finally, determine EVPI using “expected” payoff.

EVPI = Expected value of = Expected payoff - Expected payoff

perfect information under certainty under risk
= 12.2 - 10.5
= 1.7

04 - 33
 Decisions using EVPI – “expected” regret
Example 6: Determine the Expected Value of Perfect Information (EVPI)
for the payoff table shown below using “expected” regret. Note: In the
case of regret “best” means least.

Payoff Possible Future Demand

Probability 0.3 0.5 0.2
Alternative Low Moderate High
Small Facility 10 10 10
Medium Facility 7 12 12
Large Facility -4 2 16

04 - 34
First, determine the regret table.

Regret Possible Future Demand

Probability 0.3 0.5 0.2
Alternative Low Moderate High
Small Facility 0 2 6
Medium Facility 3 0 4
Large Facility 14 10 0

04 - 35
Next, determine the “expected” regret for each alternative across each state of
nature under risk

Regret Possible Future Demand

Probability 0.3 0.5 0.2
Expected
Alternative Low Moderate High Regret
Small Facility 0 2 6 2.20
Medium Facility 3 0 4 1.70
Large Facility 14 10 0 9.20

Then, determine the EVPI = Expected Monetary Value (“least” regret).

EMV  best ( EVsmall , EVmedium , EVl arg e )

 best ( 2.2,1.7,9.2)  1.7
04 - 36
04.4 EVPI

04 - 37
04 - 38
 Sensitivity Analysis
Sensitivity Analysis is a method that is used to determine the probability range
of a future condition (state of nature) over which an alternative is best. It can
also be used to determine the probability at which the alternatives are the
same.

Sensitivity analysis can be used when many states of nature and many
alternatives are involved. For purposes of this lecture, we will only discuss
situations where there are two states of nature providing an excellent visual
understanding of the concept.

The methodology of sensitivity analysis relies on the knowledge of equations,

the ability to plot equations and the ability to solve simultaneous equations.

04 - 39
 Sensitivity Analysis Example
Example 8: A manager is trying to determine which alternative warehouse site
will be best. His/her decision is based on whether or not a new bridge will be
built. The payoff table for the analysis is shown below. Using graphical
sensitivity analysis determine the probability for “no new bridge” where each
alternative is optimal.

-Payoff / (Loss) Table-

-S1- -S2-
-Alternatives- New Bridge No New Bridge
A 4 12
B 16 2
C 12 8

We begin by plotting the payoff for each alternative on the same graph.
04 - 40
 Sensitivity Analysis Example (PS1)
-S1- -S2- Let’s begin by plotting the payoff
-Alternatives- New Bridge No New Bridge
Sensitivity Analysis for Alternative A.
A14.00 4 12 Remember all we need are two
points to plot a line.
12.00

10.00
Alternative Payoff

What is the Payoff for A when

the Probability
8.00
for S1 (New
Bridge) is 0. In other words
what is the 6.00
payoff for No New
Bridge is built.
4.00

2.00 What is the Payoff for A when

the Probability for S1 (New
-
0 0.1 0.2 0.3 0.4 0.5 0.6
Bridge)
0.7
is0.8 1. In other
0.9
words
1

Probability (State of Nature 1) what is the Payoff if a New

Bridge is built.
04 - 41
 Sensitivity Analysis Example (PS1)
When we finish, we will have all
Sensitivity Analysis
3 payoffs on the same graph.
18.00

16.00

Best Payoff is different over

Alternative Payoff

14.00

the probability range.

12.00

10.00

8.00

6.00

4.00

2.00

A is Best C is Best B is Best

-
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

A B C #N/A #N/A Probability of (New Bridge) Occurring

04 - 42
 Sensitivity Analysis Example (PS1)
Now all we have to do is calculate the
Sensitivity Analysis
intersections using the simultaneous
equations.
18.00

16.00 C (green) and B (pink)

A (red) and C (green) intersect at .6.
Alternative Payoff

14.00
intersect at .3333.
12.00

10.00

8.00
We can conclude in terms of P(S1):
A is “best” when 0 <= P(S1) < .3333
6.00 C is “best” when .3333 < P(S1) < .6
4.00
B is “best” when .6 < P(S1) <= 1

2.00

A is Best C is Best B is Best

-
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

A B C #N/A #N/A Probability of (New Bridge) Occurring

04 - 43
 Sensitivity Analysis – Conclusion
Conclusion in terms of P(S1)
We can conclude in terms of P(S1):
A is “best” when 0 <= P(S1) < .3333
C is “best” when .3333 < P(S1) < .6
B is “best” when .6 < P(S1) <= 1

Why do we not include equal signs for the intersection points?

Now state your conclusion in terms of P(S2):

We can conclude in terms of P(S2):
A is “best” when .6666 <= P(S2) < 1
C is “best” when .4 < P(S2) < .6667
B is “best” when 0 < P(S2) <= .4

04 - 44
04.5 Sensitivity Analysis

04 - 45
Here, we have the option to choose whether we
want to perform the analysis on P(S1) or P(S2). In
this case, we choose P(S1) = P(New Bridge)

04 - 46
Choose the intersecting lines to see the intersection
points which are displayed on the table and the
graph.

04 - 47
Here, we have chosen the option to view the
analysis in terms of P(S2) = P(No New Bridge)

04 - 48
 Special Notes Regarding Templates
All of the examples shown in class showed payoff tables.
Cost tables are handled in a similar manner; however, the templates require
costs to be entered as negative numbers.
For Sensitivity Analysis, when costs are used, the line at the bottom rather than
the line at the top, represents the lowest cost.

04 - 49
 Homework
Read and understand all material in the chapter.

Discussion and Review Questions

Recreate and understand all classroom examples

Exercises on chapter web page

04 - 50