Decision Support Techniques

Decision Trees

Prof. Dr. Ir. Geert Waeyenbergh

Based on an example taken from ‘Operations Research’ by Wayne L. Winston

Case Colaco Chocola
Colaco currently has assets of $150,000 and wants to decide whether to market a new chocolate-flavored soda, Chocola.
Colaco has three alternatives:

 Alternative 1 Test market Chocola locally, then utilize the results of the market study to determine whether or not to
market Chocola nationally.
 Alternative 2 Immediately (without test marketing) market Chocola nationally.
 Alternative 3 Immediately (without test marketing) decide not to market Chocola nationally.

In the absence of a market study, Colaco believes that Chocola has a 55% chance of being a national success and a 45%
chance of being a national failure. If Chocola is a national success, Colaco’s asset position will increase by $300,000, and if
Chocola is a national failure, Colaco’s asset position will decrease by $100,000.

If Colaco performs a market study (at a cost of $30,000), there is a 60% chance that the study will yield favorable results
(referred to as a local success) and a 40% chance that the study will yield unfavorable results (referred to as a local failure). If
a local success is observed, there is an 85% chance that Chocola will be a national success. If a local failure is observed, there
is only a 10% chance that Chocola will be a national success.

If Colaco is risk-neutral (wants to maximize its expected final asset position), what strategy should the company follow?
To draw a decision tree that represents Colaco’s problem, we begin at the present and proceed toward future events and

decisions. The decision tree in Figure 7 is constructed with two kinds of forks: decision forks (denoted by □ ) and event
forks (denoted by ○).
A decision fork represents a point in time when Colaco has to make a decision. Each branch emanating from a decision fork
represents a possible decision. An example of a decision fork occurs when Colaco must determine whether or not to test
market Chocola.

An event fork is drawn when outside forces determine which of several random events will occur. Each branch of an event
fork represents a possible outcome, and the number on each branch represents the probability that the event will occur.

For example, if Colaco decides to test market Chocola, the company faces the following event fork when observing the results of the test market study:

A branch of a decision tree is a terminal branch if no forks emanate from the branch. Thus, the branches indicating National success and National failure are
terminal branches of Colaco’s decision tree. Since we are maximizing expected final asset position at each terminal branch, we must enter the final asset
position that will result if the path leading to the given terminal branch occurs. For example, the terminal branch National failure that follows Local failure leads
to a final asset position of 150,000 - 30,000 - 100,000 = $20,000. If we were maximizing expected revenues, we would enter revenues on each terminal branch.
To determine the decisions that will maximize Colaco’s expected final asset position, we work backward (sometimes called “folding back the tree”) from right
to left. (≈ Dynamic Programming)
 Colaco’s Decision Tree (Risk-Neutral)

Test market

Don’t test market

 Colaco’s Decision Tree (Risk-Neutral)

60% Local sucess

40% Local failure

Test market

Don’t test market

Market nationally

Don’t market nationally

 Colaco’s Decision Tree (Risk-Neutral)

Don’t market nationally 85%

National
success
Market nationally
60% Local sucess 15% National failure
Don’t
market
40% Local failure nationally
Test market 10% National success

Market nationally 90% National failure

55% National success

Don’t test market
45% National failure
Market nationally

Don’t market nationally

 Colaco’s Decision Tree (Risk-Neutral)

Don’t market nationally 85%

National
Market nationally success
60% Local sucess 15% National failure
Don’t
market
40% Local failure nationally
Test market 10% National success

Market nationally 90% National failure

55% National success

Don’t test market
45% National failure
Market nationally

Don’t market nationally

At each event fork, we calculate the expected final asset position and enter it in ○. At each decision fork, we denote by //
the decision that maximizes the expected final asset position and enter the expected final asset position associated with

that decision in □ . We continue working backward in this fashion until we reach the beginning of the tree. Then the
optimal sequence of decisions can be obtained by following the //.

We begin by determining the expected final asset positions for the following three event forks:
• Market nationally after Local success. Here we have an expected final asset position of:
85% x 420,000 + 15% x 20,000 = $360,000.
• Market nationally after Local failure. Here we have an expected final asset position of: See next 2 slides
10% x 420,000 + 90% x 20,000 = $60,000.
• Market nationally after Don’t test market. Here we have an expected final asset position of:
55% x 450,000 + 45% x 50,000 = $270,000.

We may now evaluate three decision forks:

1. Decision after Local success. Market nationally yields a larger expected final asset position than Don’t market nationally,
so we // Market nationally and enter an expected final asset position of $360,000.
2. Decision after Local failure. Don’t market nationally yields a larger expected final asset position than Market nationally,
so we // Don’t market nationally and enter an expected final asset position of $120,000.
3. Decision for Don’t test market. Market nationally yields a larger expected final asset position than Don’t market
nationally, so we // Market nationally and enter an expected final asset position of $270,000.
 Colaco’s Decision Tree (Risk-Neutral)

Don’t market nationally 85%

National
Market nationally success
60% Local sucess 15% National failure
Don’t
market
40% Local failure nationally
Test market 10% National success

Market nationally 90% National failure

55% National success

Don’t test market
45% National failure
Market nationally

Don’t market nationally

Colaco’s Decision Tree (Risk-Neutral)
At each event fork, we calculate the expected final asset position and enter it in ○. At each decision fork, we denote by //
the decision that maximizes the expected final asset position and enter the expected final asset position associated with

that decision in □ . We continue working backward in this fashion until we reach the beginning of the tree. Then the
optimal sequence of decisions can be obtained by following the //.

We begin by determining the expected final asset positions for the following three event forks:
• Market nationally after Local success. Here we have an expected final asset position of:
85% x 420,000 + 15% x 20,000 = $360,000.
• Market nationally after Local failure. Here we have an expected final asset position of:
10% x 420,000 + 90% x 20,000 = $60,000.
• Market nationally after Don’t test market. Here we have an expected final asset position of:
55% x 450,000 + 45% x 50,000 = $270,000.

We may now evaluate three decision forks:

1. Decision after Local success. Market nationally yields a larger expected final asset position than Don’t market nationally,
so we // Market nationally and enter an expected final asset position of $360,000.
2. Decision after Local failure. Don’t market nationally yields a larger expected final asset position than Market nationally,
so we // Don’t market nationally and enter an expected final asset position of $120,000.
3. Decision for Don’t test market. Market nationally yields a larger expected final asset position than Don’t market
nationally, so we // Market nationally and enter an expected final asset position of $270,000. See next slide
Colaco’s Decision Tree (Risk-Neutral)
We now must evaluate the event fork emanating from the Test market decision.

This event fork yields an expected final asset position of 60% x 360,000 + 40% x 120,000 = $264,000, which is entered in ○.
All that remains is to determine the correct decision at the decision fork Test market versus Don’t test market.

We have found that:

• Test market yields an expected final asset position of $264,000.
• Don’t test market yields an expected final asset position of $270,000.

Thus, we // Don’t test market and enter $270,000 in □.

We have now reached the beginning of the tree and have found that Colaco’s optimal decision is Don’t test market and then
Market nationally.

This strategy will yield an expected final asset position of $270,000. Observe that the decision tree also tells us that if we had
test marketed and then acted optimally (Market nationally after Local success and Don’t market nationally after Local
failure), we would have obtained an expected final asset position of $264,000.
Colaco’s Decision Tree (Risk-Neutral)
Incorporating Risk Aversion into Decision Tree Analysis
Note that Colaco’s optimal strategy yields a 45% chance that the company will
end up with a relatively small final asset position of $50,000.

On the other hand, the strategy of test marketing and acting optimally on the
results of the test market study yields only a 60% x 15% = 9% chance that
Colaco’s asset position will be below $100,000. (Why?)

Thus, if Colaco is a risk-averse decision maker, the strategy of immediately

marketing nationally may not reflect the company’s preference.

To illustrate how risk aversion may be incorporated into decision tree analysis,
suppose that Colaco has the risk-averse utility function u(x) in the figure on
the left. (How do we know that this utility function exhibits risk aversion?)

To determine Colaco’s optimal decisions (that is, the decisions that maximize
expected utility), simply replace each final asset position x0 with its utility
(final asset position)
u(x0). Then at each event fork, compute the expected utility of Colaco’s final
asset position, and at each decision fork, choose the branch having the largest
expected utility.
Colaco’s Utility Function
We find from figure Colaco’s Utility Function that u($450,000) = 1, u($420,000) = .99, u($150,000) = .48, u($120,000) = .40,
u($50,000) = .19, and u($20,000) = 0.

Substituting these values into the decision tree yields the new decision tree below. We compute the expected utility at the
following three event forks:
1 Market nationally after Local success. Here we have an expected utility of .85(.99) + .15(0) = .8415.
2 Market nationally after Local failure. Here we have an expected utility of .10(.99) + .90(0) = .099.
3 Market nationally after Don’t test market. Here we have an expected utility of .55(1) + .45(.19) = .6355.

We may now evaluate three decision forks:

1 Decision after Local success. Market nationally yields a larger expected utility than Don’t market nationally, so for this fork we Market nationally and enter an expected utility of .8415.
2 Decision after Local failure. Don’t market nationally yields a larger expected utility than Market nationally, so for this fork we Don’t market nationally and enter an expected utility of .40.
3 Decision for Don’t test market. Market nationally yields a larger expected utility than Don’t market nationally, so for this fork we Market nationally and enter an expected utility of .6355.

We now must evaluate the event fork emanating from the Test market decision.
This event fork yields an expected utility of .60(.8415) + .40(.40) = .6649, which is entered in . All that remains is to determine the correct decision at the
decision fork Test market versus Don’t test market. We know that Test market yields an expected utility of .6649, and Don’t test market yields an expected
utility of .6355, so we Test market and enter an expected utility of .6649 in . We have now reached the beginning of the tree and have found that Colaco’s
optimal decision is to begin by test marketing. If a local success is observed, then Colaco should market Chocola nationally; if a local failure is observed, then
Colaco should not market Chocola nationally. This optimal strategy yields only a .60(.15) .09 chance that Colaco will have a final asset position of less than
$100,000. This reflects the risk-averse nature of the utility function in figure Colaco’s Utility Function.

Also, we see from the figure that u($226,000) = .665. Since Colaco views the current situation as having an expected utility of .6649, this means that the
company considers the current situation equivalent to a certain asset position of $226,000. Thus, if somebody offered to pay more than 226,000 - 150,000 =
$76,000 to buy the rights to Chocola, Colaco should take the offer. This is because receiving more than $76,000 for the rights to Chocola would bring Colaco’s
asset position to more than 150,000 + 76,000 = $226,000, and this situation has a higher expected utility than .665.
Colaco’s Decision Tree (Risk-Averse)
Expected Value of Sample Information
Decision trees can be used to measure the value of sample or test market information.

To illustrate how this is done, we again assume that Colaco is risk-neutral. What is the value of the information that would be
obtained by test marketing Chocola?

We begin by determining Colaco’s expected final asset position if the company acts optimally and the test market study is
costless. We call this expected final asset position Colaco’s Expected Value With Sample Information (EVWSI). From Colaco’s
Risk-Neutral Decision Tree, we see that if we Test market and then act optimally, we will now have an expected final asset
position of 264,000 + 30,000 = $294,000. Since $294,000 is larger than the expected asset position of the Don’t test market
branch ($270,000), we find that EVWSI = $294,000.

We next determine the largest expected final asset position that Colaco would obtain if the test market study were not
available. We call this the Expected Value With Original Information (EVWOI). From the Don’t test market branch of Colaco’s
Risk-Neutral Decision Tree, we find EVWOI = $270,000. Now the expected value of the test market information, referred to as
Expected Value of Sample Information (EVSI), is defined to be EVSI = EVWSI - EVWOI.

In the Colaco example, EVSI is the most that Colaco can pay for the test market information and still be at least as well off as
without the test market information. Thus, for the Colaco example, EVSI = 294,000 - 270,000 = $24,000. Since the cost of the
test market study ($30,000) exceeds EVSI, Colaco should not (as we already know) conduct the test market study.
 Colaco’s Decision Tree (Risk-Neutral) - Expected Value of Sample Information

+ 30,000

?
294,000
Expected Value of Perfect Information
We can modify the analysis used to determine EVSI to find the value of perfect information. By perfect information we mean
that all uncertain events that can affect Colaco’s final asset position still occur with the given probabilities (so there is still a
55% chance of Chocola being a national success and a 45% chance that Chocola will be a national failure), but Colaco finds
out whether Chocola is a national success or a national failure before making the decision to market Chocola nationally or
not. This information can then be used to determine Colaco’s optimal marketing strategy. Thus, Expected Value With Perfect
Information (EVWPI) is found by drawing a decision tree in which the decision maker has perfect information about which
state has occurred before making a decision. Then the Expected Value of Perfect Information (EVPI) is given by EVPI = EVWPI
- EVWOI.

LARGER ON NEXT SLIDE

270,000

For the Colaco example, we find from the figure above that EVWPI = $315,000. Then EVPI = 315,000 - 270,000 = $45,000.
Thus, a perfect (one that was always correct) test marketing study would be worth $45,000. EVPI is a useful upper bound on
the value of sample or test market information; that is, no sample or test market information (no matter how good) can be
worth more than $45,000.
Expected Value with Perfect Information (EVWPI) for Colaco
Extra exercise
An art dealer’s client is willing to buy the painting Sunplant at $50,000. The dealer can buy the painting today for $40,000 or
can wait a day and buy the painting tomorrow (if it has not been sold) for $30,000. The dealer may also wait another day and
buy the painting (if it is still available) for $26,000. At the end of the third day, the painting will no longer be available for sale.
Each day, there is a 60% probability that the painting will be sold. What strategy maximizes the dealer’s expected profit?
Extra exercise
An art dealer’s client is willing to buy the painting Sunplant at $50,000. The dealer can buy the painting today for $40,000 or
can wait a day and buy the painting tomorrow (if it has not been sold) for $30,000. The dealer may also wait another day and
buy the painting (if it is still available) for $26,000. At the end of the third day, the painting will no longer be available for sale.
Each day, there is a 60% probability that the painting will be sold. What strategy maximizes the dealer’s expected profit?

Solution

The decision tree for this example is given in the figure above.

The key to drawing this decision tree is that each day, the dealer must choose between buying the painting and waiting
another day. Of course, waiting might mean that the dealer may never be able to buy the painting.

As we see from the decision tree, the dealer should buy the painting on the first day.