You are on page 1of 5

DECISION THEORY Possible Future Demand

Alternatives Low Moderate High


Decision theory represents a general
approach to decision making. It is suitable for a Small facility $10 $10 $10
wide range of operations management Medium facility 7 12 12
decisions. Among them are capacity planning, Large facility (4) 2 16
product and service design, equipment
selection, and location planning. Decisions that
lend themselves to a decision theory approach The payoffs are shown in the body of
tend to be characterized by these elements: the table. In this instance, the payoffs are in
terms of present values, which represent
1. A set of possible future conditions exists equivalent current dollar values of expected
that will have a bearing on the results of the future income less costs. This is a convenient
decision. measure because it places all alternatives on a
2. A list of alternatives for the manager to comparable basis. If a small facility is built, the
choose from. payoff will be the same for all three possible
3. A known payoff for each alternative under states of nature. For a medium facility, low
each possible future condition. demand will have a present value of $7 million,
whereas both moderate and high demand will
In order to use this approach, a decision have present values of $12 million. A large
maker would employ this process: facility will have a loss of $4 million if demand
is low, a present value of $2 million if demand
1. Identify the possible future conditions (e.g., is moderate, and a present value of $16 million
demand will be low, medium, or high; the if demand is high.
number of contracts awarded will be one,
two or three; the competitor will or will not The problem for the decision maker is to
introduce a new product). These are called select one of the alternatives, taking the
states of nature. present value into account.
2. Develop a list of possible alternatives, one
of which may be to do nothing. Evaluation of the alternatives differs
3. Determine or estimate the payoff according to the degree of certainty associated
associated with each alternative for every with the possible future conditions. There are
possible future condition. three possibilities to consider: complete
4. If possible, estimate the likelihood of each certainty, risk, and uncertainty.
possible future condition.
5. Evaluate alternatives according to some
decision criterion (e.g., maximize expected Decision Making under Certainty
profit), and select the best alternative.
When it is known for certain which of the
The information for a decision is often possible future conditions will actually happen,
summarized in a payoff table, which shows the the decision is usually relatively
expected payoffs for each alternative under the straightforward: Simply choose the alternative
various possible states of nature. These tables that has the best payoff under the state of
are helpful in choosing among alternatives nature.
because they facilitate comparison of
alternatives. Consider the following payoff Example 1. Determine the best alternative in
table, which illustrates a capacity planning the preceding payoff table for each of the
problem. following cases: It is known with certainty that
demand will be: (a) low, (b) moderate, (c) high.
Solution. Choose the alternative with the Example 2. Referring to the preceding payoff
highest payoff. Thus, if we know that demand table, determine which alternative would be
will be low, we would elect to build a small chosen under each of these strategies:
facility and realize a payoff of $10 million. If we
know demand will be moderate, a medium a. Maximin
factory would yield the highest payoff of $12 b. Maximax
million. For high demand, a large facility would c. Laplace
provide the highest payoff of $16 million.
Solution.
Although complete certainty is rare in a. Using maximin, the worst payoffs for the
such situations, this kind of exercise provides alternatives are:
some perspective on the analysis. Moreover, in
some instances, there may be an opportunity Small facility: $10 million
to consider allocation of funds to research Medium facility: 7 million
efforts, which may reduce or remove some Large facility: – 4 million
uncertainty surrounding the states of nature. Hence, since $10 million is the best, choose to
build the small facility using the maximin
Decision Making under Uncertainty strategy.

At the opposite extreme is complete b. Using maximax, the best payoffs are:
uncertainty: no information is available on how
likely the various states of nature are. Under Small facility: $10 million
Medium facility: 12 million
those conditions, four possible decision criteria
Large facility: 16 million
are maximin, maximax, Laplace, and minimax
regret. These approaches can be defined as The best overall payoff is the $16 million in the
follows: third row. Hence, the maximax criterion leads
to building a large facility.
Maximin – Choose the alternative with the
best of the worst possible payoffs. The c. For the Laplace criterion, first find the row
maximin approach is essentially a pessimistic totals, and then divide each of those
one because it takes into account only the amounts by the number of states of nature.
worst possible outcome for each alternative. (three in this case). Thus, we have:
The actual outcome may not be as bad as that, Small facility: ($10 + $10 + $10)/3 = $10.00 million
but this approach establishes a “guaranteed Medium facility: ($7 + $12 + $12)/3 = $10.33 million
minimum.” Large facility: (– $4 + $2 + $16)/3 = $4.67 million

Maximax – Choose the alternative with the Because the medium facility has the highest
best possible payoff. The maximax approach average, it would be chosen under the Laplace
is an optimistic, “go for it” strategy; it does not criterion.
take into account any payoff other than the
best. Example 3. Determine which alternative would
be chosen using a minimax regret approach to
Laplace – Choose the alternative with the best the capacity planning program.
average payoff of any of the alternatives. The
Laplace approach treats the states of nature as Solution. The first step in this approach is to
equally likely. prepare a table of opportunity losses, or
regrets. To do this, subtract every payoff in
Minimax regret – Choose the alternative that each column from the best payoff in that
has the least of the worst regrets. This column. For instance, in the first column, the
approach seeks to minimize the difference best payoff is 10, so each of the three numbers
between the payoff that is realized and the best in that column must be subtracted from 10.
payoff for each state of nature.
Going down the column, the regrets will be 10 weighted by the probability for the relevant
– 10 = 0, 10 – 7 = 3, and 10 – (–4) = 14. In the state of nature. Thus, the approach is:
second column, the best payoff is 12.
Subtracting each payoff in that column from 12 Expected monetary value criterion (EMV) –
yields 2, 0, and 10. In the third column, 16 is Determine the expected payoff for each
the best payoff. The regrets are 6, 4, and 0. alternative, and choose the alternative that has
These results are summarized in a regret table: the best expected payoff.

Regrets (in $ millions) Example 4. Using the expected monetary


Alternatives Low Moderate High Worst value criterion, identify the best alternative for
the previous payoff table for these probabilities:
Small facility $ 0 $ 2 $ 6 $ 6
Medium facility 3 0 4 4 low = 0.30, moderate = 0.50, and high = 0.20.
Large facility 14 10 0 14
Solution. Find the expected value of each
alternative by multiplying the probability of
The second step is to identify the worst occurrence for each state of nature by the
regret for each alternative. For the first payoff for that state of nature and summing
alternative, the worst is 6; for the second, the them:
worst is 4; and for the third, the worst is 14.
EVsmall = 0.30($10) + 0.50($10) + 0.20($10) = $10
The best of these worst regrets would EVmedium = 0.30($7) + 0.50($12) + 0.20($12) = $10.5
be chosen using the minimax regret. The EVlarge= 0.30($–4) + 0.50($2) + 0.20($16) = $ 3
lowest regret is 4, which is for the medium
facility. Hence, that alternative would be Hence, choose the medium facility because it
chosen. has the highest expected value.

The main weakness of these The expected monetary value approach


approaches (except for Laplace) is that they do is most appropriate when a decision maker is
not take into account all the payoffs. Instead, neither risk averse nor risk seeking, but is
they focus on the worst or best, and so they neutral. Typically, well-established
lose some information. The weakness of organizations with numerous decisions of this
Laplace is that it treats all states of nature as nature tend to use expected value because it
equally likely. Still, for a given set of provides an indication of the long-run average
circumstances, each has certain merits that payoff.
can be helpful to a decision maker.
Decision Trees
Decision Making under Risk A decision tree is a schematic
representation of the alternatives available to a
Between the two extremes of certainty decision maker and their possible
and uncertainty lies the case of risk: The consequences. The term gets its name from
probability of occurrence for each state of the treelike appearance of the diagram.
nature is known. (Note that because the states Although tree diagrams can be used in place of
are mutually exclusive and collectively a payoff table, they are particularly useful for
exhaustive, these probabilities must add to analyzing situations that involve sequential
1.00). The widely used approach under such decisions. For instance, a manager may
circumstances is the expected monetary initially decide to build a small facility only to
value criterion. The expected value is discover that demand is much higher than
computed for each alternative, and the one anticipated. In this case, the manager may
with the highest expected value is selected. then be called upon to make subsequent
The expected value is the sum of the payoffs decision on whether to expand or build an
for an alternative where each payoff is additional facility.
A decision tree is composed of a payoff will be $40 (thousand). Similarly, if a
number of nodes that have branches small facility is built, demand turns out high,
emanating from them. Square nodes denote and a later decision is made to expand, the
decision points, and circular nodes denote payoff will be $55. The figures in parentheses
chance events. Read the tree from left to right. on branches leaving the chance nodes
Branches leaving square nodes represent indicate the probabilities of those states of
alternatives; branches leaving circular nodes nature. Hence, the probability of low demand is
represent chance events (i.e., the possible 0.4, and the probability of high demand is 0.6.
states of nature). Payoffs in parentheses indicate losses.

After the tree has been drawn, it is Analyze the decisions from right to left:
analyzed from right to left; that is, starting with
the last decision that might be made. For each 1. Determine which alternative would be
decision, choose the alternative that will yield selected for each possible second decision.
the greatest return (or the lowest cost). If For a small facility with high demand, there
chance events follow a decision, choose the are three choices: do nothing, work
alternative that has the highest expected overtime, and expand. Because expand
monetary value (or the lowest expected cost). has the highest payoff, you would choose it.
Indicate this by placing a double slash
Example 5. A manager must decide on the through each of the other alternatives.
size of a video arcade to construct. The Similarly, for large facility with low demand,
manager has narrowed the choices to two: there are two choices: do nothing and
large or small. Information has been collected reduce prices. You would reduce prices
on payoffs, and a decision tree has been because it has the higher expected value,
constructed. Analyze the decision tree and so a double slash is placed on the other
determine which initial alternative (build small branch.
or build large) should be chosen in order to
maximize expected monetary value.

$40 2. Determine the product of the chance


Low demand (0.4) probabilities and their respective payoffs for
the remaining branches:
Do nothing
$40

Overtime $50 Build small


High demand (0.6) Low demand 0.4($40) = $16
Build small High demand 0.6($55) = 33
Expand $55

Build large Do nothing


($10) Build large
Low demand 0.4($50) = 20
Low demand (0.4)
High demand 0.6($70) = 42
Reduce prices $50

High demand (0.6) 3. Determine the expected value of each initial


$70 alternative:

Solution. The dollar amounts at the branch Build small $16 + $33 = $49
ends indicate the estimated payoffs if the Build large $20 + $42 = $62
sequence of chance events and decisions that
is traced back to the initial decision occurs. For Hence, the choice should be to build the
example, if the decision is to build a small large facility because it has a larger expected
facility and it turns out that demand is low, the value than the small facility.
Expected Value of Perfect Information

In certain situations, it is possible to Example 6. Using the information from


ascertain which state of nature will actually Example 4, determine the expected value of
occur in the future. For instance, the choice of perfect information.
location for a restaurant may weigh heavily on
whether a new highway will be constructed or Solution. First, compute the expected payoff
whether a zoning permit will be issued. A under certainty. To do this, identify the best
decision maker may have probabilities for payoff under each state of nature. Then
these states of nature; however, it may be combine these by weighting each payoff by the
possible to delay a decision until it is clear probability of that state of nature and adding
which state of nature will occur. This might the amounts. Thus, the best payoff under low
involve taking an option to buy the land. If the demand is $10, the best under moderate
state of nature is favorable, the option can be demand is $12 and the best under high
exercised; if it is unfavorable, the option can be demand is $16. The expected payoff under
allowed to expire. The question to consider is certainty is, then:
whether the cost of the option will be less than
the expected gain due to delaying the decision 0.30($10) + 0.5($12) + 0.20($16) = $12.2
(i.e., the expected payoff above the expected
value). The expected gain is the expected The expected payoff under risk, as computed
value of perfect information, or EVPI. in Example 4, is $10.5. The EVPI is the
difference between these:
Expected value of perfect information – the
difference between the expected payoff with EVPI = $12.2 – $10.5 = $1.7
perfect information and the expected payoff
under risk. This figure indicates the upper limit on the
amount the decision maker should be willing to
Other possible ways of obtaining perfect spend to obtain perfect information in this case.
information depend somewhat on the nature of Thus, if the cost equals or exceeds this
the decision being made. Information about amount, the decision maker would be better off
consumer preferences might come from not spending additional money and simply
market research, additional information about a going with the alternative that has the highest
product could come from product testing, or expected payoff.
legal experts might be called on.

One way to determine the EVPI is to compute


the expected payoff under certainty and
subtract the expected payoff under risk. That
is,
Expected payoff Expected payoff
EVPI = under certainty – under risk

You might also like