Decision Tree
Application
Decision Trees
• Particularly useful when there are a series of decisions
and outcomes that lead to other decisions and outcomes
Decision nodes – represented by Chance nodes – represented by
squares circles
Select the Alternative with Calculate Expected Value based
Maximum value for profit/benefit or on values and probabilities of
Minimum value for cost outgoing branches
2
Procedure
1. Include all possible alternatives and states of nature –
including “doing nothing”
2. Enter payoffs at end of branch
3. Determine the expected value of each branch and
“prune” the tree to find the alternative with the best
expected value
3
Decision Tree Example1 (1 of 4)
Figure
Silicon, Inc., 1aDecision
semiconductorTree for Development
manufacturer, is of a New Product
investigating the possibility of producing and
marketing a microprocessor. Undertaking this project
will require either purchasing a sophisticated CAD
system or hiring and training several additional
engineers. The market for the product could be
either favorable or unfavorable. Silicon, Inc., of
course, has the option of not developing the new
product at all.
With favorable acceptance by the market,
sales would be 25,000 processors selling for $100
each. With unfavorable acceptance, sales would be
only 8,000 processors selling for $100 each. The
cost of CAD equipment is $500,000, but that of
hiring and training three new engineers is only
$375,000. However, manufacturing costs should
drop from $50 each when manufacturing without
CAD to $40 each when manufacturing with CAD.
The probability of favorable acceptance of
the new microprocessor is .40; the probability of
4
unfavorable acceptance is .60.
Decision Tree Example1 (2 of 4)
Figure 1 [Continued]
5
Decision Tree Example1 (3 of 4)
Figure 1 [Continued]
6
Decision Tree Example1 (4 of 4)
Figure 1 [Continued]
7
Example 2
• Ritz Products's materials manager, Tej Dhakar, must determine whether to make
or buy a new semiconductor for the wrist TV that the firm is about to produce.
One million units are expected to be produced over the life cycle. If the product is
made, start-up and production costs of the make decision total $1 million, with a
probability of .4 that the product will be satisfactory and a .6 probability that it will
not If the product is not satisfactory, the firm will have to reevaluate the decision.
If the decision is reevaluated, the choice will be whether to spend another $1
million to redesign the semiconductor or to purchase. Likelihood of success the
second time that the make decision is made is .9. If the second make decision
also fails, the firm must purchase. Regardless of when the purchase takes place,
Dhakar's best judgment of cost is that Ritz will pay $0.50 for each purchased
semiconductor plus $1 million in vendor development cost.
• a) Assuming that Ritz must have the semiconductor (stopping or doing without is
not a viable option), what is the best decision?
• b) What criteria did you use to make this decision?
• c) What is the worst that can happen to Ritz as a result of this particular decision?
What is the best that can happen? 8
If the product is made, start-up and production costs of the make decision total $1 million, with a probability of .4
that the product will be satisfactory and a .6 probability that it will not If the product is not satisfactory, the firm will
have to reevaluate the decision. If the decision is reevaluated, the choice will be whether to spend another $1
million to redesign the semiconductor or to purchase. Likelihood of success the second time that the make
decision is made is .9. If the second make decision also fails, the firm must purchase. Regardless of when the
purchase takes place, Dhakar's best judgment of cost is that Ritz will pay $0.50 for each purchased
semiconductor plus $1 million in vendor development cost
• a) Assuming that Ritz must have the semiconductor (stopping or doing without is not a
viable option), what is the best decision?
• b) What criteria did you use to make this decision?
• c) What is the worst that can happen to Ritz as a result of this particular decision?
What is the best that can happen?
Answer
• (a) The best decision would be to buy the semiconductors. This decision has an
expected payoff (cost) of $1,500,000.
• (b) Expected monetary value, minimum cost.
• (c) The worst that can happen is that Ritz fails at its attempt to make semiconductors,
ends up buying them instead, and spends $3,500,000. The best that can happen is
that they make the semiconductors and spend only $1,000,000.
10
