Decision Tree

Visual/pictorial display of the problem

Can be used instead of a table to show alternatives,
outcomes, and payoffs
Chronologically depicts the sequence of actions and
outcomes as they unfold
Helpful when sequence of decisions are to be made (Multi-
Stage Decision Making)

Constructing the Decision Tree

Decision node: A decision to be made at that point
represented by a square
Event (chance outcome) node: A random event occurs at that
point
represented by a circle
 Thompson Lumber Co. Example
1. Decision: Whether to expand the product line by
manufacturing and marketing a new product,
backyard storage sheds
2. Alternatives:
Build a large plant to manufacture sheds
Build a small plant
Do nothing (not developing the new product
line)
3. Outcomes: Demand for sheds will be high,
moderate, or low
 Payoffs

Outcomes (Demand)
Alternatives High Moderate Low
Large plant 200,000 100,000 -120,000
Small plant 90,000 50,000 -20,000
No plant 0 0 0
Decision Tree for Thompson Lumber
High Demand $200,000

Moderate Demand
b $100,000

Low Demand -$120,000

High Demand $90,000

Small Plant Moderate Demand

a c $50,000

Low Demand -$20,000

All Demands
d $0
 Folding Back a Decision Tree
For identifying the best decision using the tree
Work from right to left
Calculate the expected payoff/ monetary value at
each event node
Choose the best alternative at each decision node
(based on expected payoff/ EMV)
 EMV
Outcomes (Demand)
Alternatives High Moderate Low EMV
Large plant 200,000 100,000 -120,000 86,000
Small plant 90,000 50,000 -20,000 48,000
No plant 0 0 0 0
p1= 0.3 p2= 0.5 p3= 0.2

Choose Large Plant

 Decision Tree for Thompson Lumber
High Demand (0.30)
$200,000
$86,000
Moderate Demand (0.50)
b $100,000

Low Demand (0.20) -$120,000

High Demand (0.30) $90,000
$86,000 $48,000
Small Plant Moderate Demand (0.50)
a c $50,000

Low Demand (0.20) -$20,000

$0
All Demands
d $0

Node b: 200000 0.3 + 100000 0.5 - 120000 0.2 = $86000

Node c: 90000 0.3 + 50000 0.5 - 20000 0.2 = $48000
 Decision Trees
Multistage Decision-Making Problems

Multistage problems involve a sequence of several

decisions and outcomes

It is possible for a decision to be immediately

followed by another decision

Decision trees are best for showing the sequential

arrangement
A finance manager is considering drilling a well. In the past, only 70%
of wells drilled were successful at 20 metres depth in that area.
Moreover, on finding no water at 20 metres, some persons in that area
drilled it further up to 25 metres but only 20% struck water at that level.
The prevailing cost of drilling is Rs 500 per metre. The finance manager
estimated that in case he does not get water in his own well, he will have
to pay Rs 15,000 to buy water from outside for the same period of
getting water from the well.

The following decisions are considered:

a) Do not drill any well
b) Drill up to 20 meters, and
c) If no water is found at 20 meters, drill further upto 25 meters

Draw a decision tree and determine the finance managers optimal

strategy
The Oil India Corporation is considering whether to go for an offshore
drilling contract to be awarded in Bombay High. If they bid, value would
be Rs 600 million with 65% chance of gaining the contract. Corporation
may set up a new drilling operation or move already existing operation,
which has proved successful, to new site. The probability of success and
expected returns are as follows:
New Drilling operation Existing Operation
Outcome Probability Expected Probability Expected
Revenue Revenue
(Rs millions) (Rs millions)

Success .75 800 .85 700

Failure .25 200 .15 350
If they do not bid or lose contract, they can use Rs 600 mill to modernize
their operations. This would result in a return of either 5% or 8% on the
sum invested with probabilities 0.45 and 0.55 respectively.
a) Construct a decision tree
b) Recommend whether they should bid the contract
 Decision Making with Experimentation

Decision makers have prior probability assessments

for various states of nature.

However, one may want to seek additional

information about states of nature.

New information can be used to revise the prior

probabilities.
 Experimentation is the process of collecting
additional information in order to obtain a better
estimate of probabilities for states of nature.
-product testing, raw material sampling, market survey

Better the estimate better the decision.

The revised probabilities are called posterior

probabilities.

Note: Experiment provides sample information (not

perfect information)
 Expanded Thompson
Lumber Example

Suppose they will first decide whether to pay $4000

to conduct a market survey
Survey results will be imperfect
Then they will decide whether to build a large plant,
small plant, or no plant
Then they will find out what the outcome and payoff
are
 Thompsons Complex Decision Tree
Using Sample Information
Thompson Lumber has two decisions to make, with the
second decision dependent upon the outcome of the
first
First, whether or not to conduct the market survey, at a
cost of $4,000, to help them decide which alternative to
pursue (large, small or no plant)
The survey does not provide perfect information
Then, to decide which type of plant to build
Note that the $4,000 cost will be subtracted from each of the
branches which corresponds to conducting a survey. The, $196,000
payoff was originally $200,000 and the $-4,000 was originally $0.
 To Test or Not to Test
Often, companies have the option to perform market
tests/surveys, usually at a price, to get additional
information prior to making decisions.
However, some interesting questions need to be
answered before this decision is made:
How will the test results be combined with prior
information?
How much should you be willing to pay to test?
The good news is that Bayes Theorem can be used to
combine the information, and we can use our decision
tree to find EVSI, the Expected Value of Sample
Information.
In order to perform these calculations, we first need to
know how reliable the potential test may be.
Estimating Probabilities using Bayesian Analysis

Allows probability values to be revised based on new

information (from a survey or test market)
Prior probabilities are the probability values before
new information
Revised probabilities are obtained by combining the
prior probabilities with the new information
 Baye's Theorem
Apply the Conditional probability formula:
P(A|B) = P(A and B) / P(B)
P(A|B) is the probability of event A occurring, given that event
B has occurred
When P(A|B) P(A), this means the probability of event A has
been revised based on the fact that event B has occurred

Recap of total probability:

Given an event B, and a set of mutually exclusive and
collectively exhaustive events (B1,B2,,Bn),
P(B) = P(B/B1).P(B1) + P(B/B2).P(B2) + .... + P(B/Bn).P(Bn)

Bayes Theorem
P(B1/B) = P(B/B1).P(B1)/[P(B/B1).P(B1) + P(B/B2).P(B2) + .... +
P(B/Bn).P(Bn)]

= P(B/B1).P(B1)/P(B)
 Prior Analysis
Outcomes (Demand)
Alternatives High Moderate Low EMV
Large plant 200,000 100,000 -120,000 86,000
Small plant 90,000 50,000 -20,000 48,000
No plant 0 0 0 0
p1= 0.3 p2= 0.5 p3= 0.2

EP= 86,000. Choose Large Plant

For computing EPPI
In case of high demand best strategy is large plant (200,000)
In case of moderate demand best strategy is large plant (100,000)
In case of low demand best strategy is no plant (0)
EPPI= .3*200,000+.5*100,000+.2*0 = 110,000
EVPI= 110,000-86000=24000
 Posterior Analysis
Assume that a marketing research firm is engaged for
providing additional information resulting in a more
informed decision.
This additional information may call for a revision of the
decision of manufacturing large plant or reinforce it.
The approach:
-revise prior probabilities of high, moderate and low
demand to calculate posterior probabilities
- use these revised (posterior) probabilities to reach a
decision.
The new information will not be cent percent correct.
 Market research firm provided the following probabilities based on
its track record of survey accuracy (market survey reliability in
predicting actual states of nature):

Market Survey Report

Event Positive (PS) Negative (NS)
High Demand (HD) .967 .033
Moderate Demand .533 .467
(MD)
Low Demand (LD) .067 .933

Positive result (PS): A significant number of individuals contacted

expressed interest in purchasing the product. i.e. A favorable
report indicating positive demand
Negative result (NS): Very few of the individuals contacted
expressed interest in purchasing the product. i.e. An unfavorable report
indicating low demand
 Probabilities given in the context of marketing firms past records
are conditional probabilities.
P(PS|HD) = 0.967 P(NS|HD) = 0.033
P(PS|MD) = 0.533 P(NS|MD) = 0.467
P(PS|LD) = 0.067 P(NS|LD) = 0.933

When true state of nature is high level of demand, the report shall
be favourable 96.7% of the time, and unfavourable 3.3% of the
time.
Survey seems to be reliable enough.
How do we find the revised (posterior) probabilities (changed
prior probabilities) when the survey result is given?
For example: P(HD|PS) = ?
 Case: When a favorable/positive survey report is given (PS)

Calculation of posterior probabilities

Prior Prob Cond. prob Joint probability Posterior Prob
P(Ei) P(PS/Ei) P(PS Ei) P(Ei/PS)
Event Ei = P(PS/Ei) x P(Ei)
E1=HD .3 .967 .2901 P(PS E1)/P(PS)
= .2901/.57
=.509
E2=MD .5 .533 .2665 .468
E3=LD .2 .067 .0134 .023

Note: P(PS)=P(PS E1 ) +P(PS E2)+P(PS E3)

= P(PS/ E1) x P(E1) + P(PS/ E2) x P(E2) + P(PS/ E3) x P(E3)
= .2901+.2665+.0134 = .57
Notice that the probability of HD increased from 0.30 to 0.509
given the positive survey result
Case: When unfavorable/ negative survey report is given (NS)

Calculation of posterior probabilities

Prior Prob Cond. prob Joint prob P(NS Ei) Posterior Prob
Event Ei P(Ei) P(NS/Ei) = P(NS/Ei) x P(Ei) P(Ei/NS)
E1=HD .3 .033 .0099 P(NS E1)/ P(NS)
= .0099/.43
=.023
E2=MD .5 .467 .2335 .543
E3=LD .2 .933 .1866 .434

Note: P(NS)=P(NS E1 ) +P(NS E2)+P(NS E3)

= P(NS/ E1) x P(E1) + P(NS/ E2) x P(E2) + P(NS/ E3) x P(E3)
= .0099+.2335+.1866= .43
 Thompson Lumber
Optimal Strategy
1. Conduct the survey

2. If the survey results are positive, then build the

large plant (EMV = $141,840)

If the survey results are negative, then build the

small plant (EMV = $16,540)
 Expected Value of
Sample Information (EVSI)
The Thompson Lumber survey provides sample
information (not perfect information)
What is the value of this sample information?
EVSI = (EMV with free sample information)
- (EMV w/o any information)

If sample information had been free

EMV (with free SI) = 87,961 + 4000 = $91,961

EVSI = 91,961 86,000 = $5,961 (worth of this sample

information)
 Expected net gain of experimenting
(ENGE)

ENGE = EVSI - cost of experiment

If ENGE > 0 conduct the experiment.
If ENGE < 0 don't conduct the experiment.
If ENGE = 0 technically, one is indifferent about
conducting the experiment.
 Computing EVPI
Payoffs in blue would be chosen based on perfect
information (knowing demand level)
Demand

Alternatives High Moderate Low

Large plant 200,000 100,000 -120,000
Small plant 90,000 50,000 -20,000
No plant 0 0 0

Probability 0.3 0.5 0.2

EPPI = $110,000
EVPI = EPPI EMV = $110,000 - $86,000 = $24,000 = EOL
 EVSI vs. EVPI
How close does the sample information come to
perfect information?

Efficiency of sample information = EVSI

EVPI

Thompson Lumber: 5961 / 24,000 = 0.248

Survey is 24.8% as efficient as perfect information

A large steel manufacturing company has three options with
regards to production:
(i) produce commercially (ii) build pilot plant (iii) stop producing
steel.
The management has estimated that their pilot plant, if built, has
0.8 chance of high yield and .2 chance of low yield. If the pilot
plant does show a high yield, management assigns a probability of
0.75 that the commercial plant will also have a high yield and a
probability of 0.25 of having a low yield. If the pilot plant shows a
low yield, there is only a 0.1 chance that the commercial plant will
show a high yield. Finally, managements best assessment of the
yield on a commercial-size plant without building a pilot plant first
has a 0.6 chance of high yield. A pilot plant will cost Rs. 3,00,000.
The profits earned under high and low yield conditions are Rs.
1,20,00,000 and Rs. -12,00,000 respectively. Find the optimum
decision for the company with the help of a decision tree.