OFFICIAL (CLOSED) \ NON-SENSITIVE

Lesson 11
Decision Trees

E210 – Operations Planning

1
 OFFICIAL (CLOSED) \ NON-SENSITIVE

E210 Operations Planning Topic Tree

2
2
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Mini-Lecture (1)
Recall:
OFFICIAL (CLOSED) \ NON-SENSITIVE

Expected Value of Perfect Information (EVPI)

• Expected Value of Perfect Information (EVPI)
 EVPI is the maximum amount that should be paid to gain
information that would result in a better decision than those
decisions made without perfect information — W08
 It is under the ideal case that perfect information is available.

4
 OFFICIAL (CLOSED) \ NON-SENSITIVE

An Example: The Scenario

• Palma Food Products (PFP) is deciding whether to introduce a new line of chocolates into the
market. The company can conduct a market test or bypass the market test and introduce the
product nationwide. The cost of the market test is $100,000.

• If the company conducts the market test, it must wait and see the results before deciding
whether to introduce the chocolates. Past experience shows that, if the actual market is a
success, the probabilities of a positive and negative test result are 0.9, and 0.1, respectively.
On the other hand, if the actual market is a failure, the probabilities of a positive and negative
test result are 0.3, and 0.7, respectively.

• Alternatively, the company can skip the market test and introduce the chocolates anyway. If
the chocolates introduced are a success, the company estimates that it will realize an annual
profit of $1.5M; while if the chocolates fail, it will incur a loss of $600,000. The company
estimates that the probability of success for the chocolates is 0.50 if it is introduced without
the market test.

Actual Market Outcome Market Test Outcome

Outcome Actual Profit/Loss ($) Probability Outcome Probability
Positive 0.9
Success 1,500,000 0.5
Negative 0.1
Positive 0.3
Failure -600,000 0.5
Negative 0.7

• Using Decision-Tree analysis, determine whether the company should conduct the market
test, the value of the market test (EVSI), efficiency of sample information, and the overall best
course of action.
5
 OFFICIAL (CLOSED) \ NON-SENSITIVE
Post and Share: Skip Market Test -
Decision Tree
Assume that PFP is to skip the market test and introduce the chocolates.
Draw a decision tree and analyse whether PFP should introduce the
chocolates into the market. Calculate and specify the probabilities, payoffs
and EMVs in the decision tree clearly.

Construct your decision

tree and post to Padlet
 OFFICIAL (CLOSED) \ NON-SENSITIVE
Post and Share: Skip Market Test -
Decision Tree

Expected Monetary
Value (EMV) =
0.5 x (1,500,000) + 0.5 x (-600,000)
= $450,000

Probability of
‘Failure’

The best course of action:

Introduce the chocolates with an expected payoff of

$450,000.
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Mini-Lecture (2)
 OFFICIAL (CLOSED) \ NON-SENSITIVE
Post and Share: Skip Market Test -
Determining the EVPI

Market Outcome
EVPI = EVwithPI – EMV of the best alternative Alternatives Success Failure
50% 50%

EVwithPI = Introduce

EVPI = Not to Introduce

 OFFICIAL (CLOSED) \ NON-SENSITIVE
Post and Share: Skip Market Test -
Determining the EVPI

Market Outcome
EVPI = EVwithPI – EMV of the best alternative Alternatives Success Failure
50% 50%

EVwithPI = 1,500,000*0.5 + 0*0.5 = $750,000 Introduce 1,500,000 -600,000

EVPI = 750,000- 450,000 = $300,000 Not to Introduce 0 0

Better payoff Better payoff

when Market when Market
Outcome is a Outcome is a
‘Success’ ‘Failure’
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Mini-Lecture (3)
 OFFICIAL (CLOSED) \ NON-SENSITIVE
Post and Share: Not to Skip Market Test -
Decision Tree
If PFP decides not to bypass the market test, using Decision-Tree analysis,
determine whether the company should conduct the market test, the value
of the market test (EVSI), efficiency of sample information, and the overall
best course of action. The cost of the market test is $100,000.

• Construct your decision tree and post to Padlet.

• What are the probabilities in the decision tree that
you cannot determine?
 OFFICIAL (CLOSED) \ NON-SENSITIVE
Post and Share: Not to Skip Market Test -
Decision Tree

Conditional Probability: Given that Market

Test Outcome is Positive, the probability
Expected Monetary that Market Outcome is Success.
Value (EMV) =
0.6 x 975,000 + 0.4 x 0
= $585,000

Joint probability: the

probability that Market
Test Outcome is
Negative and Market
Outcome is Failure. It
Decision Value = is not within the scope
Max {450,000, 0) = $450,000 of this module.
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Mini-Lecture (4)
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Expected Value of Sample Information (EVSI)

• Most commonly, perfect information is not available.
However, additional information regarding future events
can be gained through tests or experiments, such as
market research, medical tests, sampling of physical
environments, etc.
• This additional information can help a decision maker more
accurately estimate probabilities of possible future events
and therefore make better decision. The value of this
additional information is often called the Expected Value
of Sample Information (EVSI).

15
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Expected Value of Sample Information (EVSI)

• EVSI is calculated as:
EVSI = Expected payoff with sample (additional) information –
Expected payoff without sample (additional) information

16
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Expected Value of Sample Information (EVSI)

• In general, obtaining sample (additional) information
includes an associated cost. Is it worthwhile to pay for
the sample (additional) information?
When EVSI > cost of obtaining the sample information through
test or experiment, it is worthwhile to conduct the test or
experiment.
When EVSI ≤ cost of obtaining the sample information through
test or experiment, it is not worthwhile to conduct the test or
experiment.

17
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Efficiency of Sample Information

• One way to judge how much information is generated by a sample (via
test or experiment) is to compute the ratio of EVSI to EVPI times 100,
expressed as a percentage.
• This is known as the efficiency of sample information.

• For example, with perfect information, the expected payoff is $100,000;

with market research, the expected payoff is $80,000; without any
information regarding future events, the expected monetary value is
$50,000. What are the values for EVPI, EVSI, and efficiency of sample
information?
 EVPI = 100,000 – 50,000 = $50,000
 EVSI = 80,000 – 50,000 = $30,000
 Efficiency of sample information = (EVSI/EVPI)*100% = (30,000/50.000)
*100% = 60%.

18
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Post and Share: Not to Skip Market Test -

Determining EVSI and Best Course of Action
If PFP decides not to bypass the market test, using Decision-Tree analysis,
determine whether the company should conduct the market test, the value
of the market test (EVSI), efficiency of sample information, and the overall
best course of action. The cost of the market test is $100,000.

• Calculate EVSI and efficiency of sample information;

• Recommend the overall best course of action;
• Share your answers in Padlet
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Post and Share: Not to Skip Market Test

Determining EVSI and Best Course of Action
The Expected Value of Since EVSI ($135,000) > cost of conducting
The Expected Sample Information (EVSI): market test ($100,000), PFP should conduct
payoff with
EVSI = $585,000 – $450,000 Market Test before deciding whether to
‘Market Test’ introduce the chocolates into the market.
= $135,000

The Expected payoff

without ‘Market Test’

The best course of action:

 Conduct Market Test
 If Market Test Outcome is ‘Positive’, introduce the chocolates;
 Otherwise, do not introduce the chocolates
 The Expected Payoff is 585,000 – 100,000 = $485,000 where $100,000 is the cost of Market Test.
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Post and Share: Not to Skip Market Test

The Efficiency of Sample Information
• The Efficiency of Sample Information:
Efficiency of Sample Information
= (EVSI/EVPI)*100% = (135,000/300,000)*100% = 45%

EVSI EVPI

• The efficiency of sample information can range from 0

to 1
 The closer the value is to 1, the closer the sample information is to
being perfect.
 The closer the value is to 0, the less information there is in the
sample.
 Thus a value such as 45% is slightly below mid-range, meaning that
relative to perfect information, the information that could be gained
from the market test is at the moderate level.
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Learning Outcomes
• Construct a decision tree for a decision making problem with a sequence of
dependent decisions.
• Calculate payoff at each terminal node.
• Calculate EMV at each event node.
• Perform decision making at each decision node.
• Apply software DPL to construct a decision tree for a problem with a
sequence of dependent decisions.
• Conduct decision making for a problem with a sequence of dependent
decisions using decision tree in real-world situations.
• Interpret Expected Value of Sample Information and Efficiency of Sample
Information when sample (additional) information is used in decision making.
• Calculate Expected Value of Sample Information and Efficiency of Sample
Information.
• Conduct decision making for a problem with a sequence of dependent
decisions when sample (additional) information is used.

22
 OFFICIAL (CLOSED) \ NON-SENSITIVE

Overview of E210 Operation Planning Module

Decision
Operations LP – Integer Making using
Planning Programming Decision Tree
Overview
LP -
Transportation Network
Model Diagram and
Decision
Making with Min. Spanning
Dependent Tree
Decisions

LP formulation LP – Binary
Process Strategy and EXCEL Integer
& Capacity Solver with Programming Decision
Requirements Sensitivity Making using
analysis Utility Dissimilarity
Function Index and MST

Linear
Programming Decision
(LP) Graphical Making under
Method uncertainty
and risk

We are
here ! 23