1. Managers are decision makers. By making decisions, they exert an influence over the resources of the organization. Hence, decision making is a fundamental part of management. Decision making consists of these steps:

(1) Specify objectives and decision criteria

(2) Develop a list of alternatives

(3) Analyze alternatives

(4) Select the best alternative

(5) Implement the alternative

(6) Monitor results

2. A key tool in decision making is the use of models, which are abstractions of reality. Models are valuable to decision makers because they enable decision makers to focus on a relatively few important aspects of a decision, they provide the ability to manipulate variables, and they increase understanding of a problem.

3. Poor decisions can occur for a variety of reasons, including faulty or inadequate information, failure to consider key alternatives, and skipping or rushing through some of the steps listed above. Managers also face limitations on decision making caused by costs, human abilities, time, technology, and the availability of information; these factors are known as bounded rationality.

4. Decision theory is a general approach to decision making that can be used for decisions that involve a set of possible future conditions (states of natures) that will affect the payoff to the organization, a list of alternatives from which to choose, and an estimated payoff for each alternative under each possible future condition. It can be very helpful to organize this information-nation into a payoff table, listing the alternatives down the left side of the table, the possible future conditions across the top of the table, and the payoffs in the body of the table.

5. If estimated probabilities for events are available, the environment is one of risk. A typical approach in such cases is the following:

1. Compute the expected payoff for each alternative using the probabilities.
2. The alternative with the best expected payoff is the optimal decision.
3. Additional information over and above that specified can sometimes be obtained. The expected value of perfect information (EVPI) is the difference between the expectedvalue of the best payoff each event and the expected value of the payoffs for the best (single) alternative; an alternative definition of EVPI is expected regret.

6. If probabilities for events are unknown, the environment is one of uncertainty, and one or more of four decision criteria can help us decide which alternative is best:

maximin (pessimistic): the alternative with the best value of the worst payoffs.
maximax (optimistic): the alternative with the best payoff.
Laplace (equal weights): since the actual probabilities for events are unknown, for each alternative compute the expected value of the payoff, assuming every event has the same probability weight; the event with the highest expected payoff is the one selected.
minimax regret (cautious): the terms "regret" and "opportunity loss" mean the same thing: given an event E, for every alternative A. compute the difference between each payoff and the best payoff obtainable for any alternative with E, that difference is called regret. Calculate the regrets for every (A, E) pair and organize them into a matrix having the same dimensions as the payoff matrix. Select the alternative with the lowest value of the highest regrets, the minimax regret.

7. Because the four uncertainty criteria all have different psychological bases, varying from extreme pessimism about the outcome to extreme outcomes, it would be most unusual if all of them agreed on what the right decision Is. As a test for the reasonableness of the results obtained from this analysis, the decision maker should select the criterion that comes closest to representing the decision maker's own psychological makeup, and compare results with one or more of the other four criteria.

8. In this discussion, payoff is considered to be "positive" profit or revenue. The same decision process can be applied to costs, which are considered to be negative, including events, alternatives and payoffs.

9. In the special case that there are just two states of nature, say El and E2,. sensitivity studies can be performed graphically to find ranges of probability values for optimal decision making, without assuming that the probability distribution is known. The type of graph is shown in Example 8 of Chapter 5S. The horizontal axis is p, the probability of event E2, and (1 - p) is the probability of El; the vertical axes show the payoffs. At the left hand edge p = 0.0, denoting that the probability of E2 is 0.0 and the probability of El is 1.0; at the right-hand edge, the probabilities are reversed, 1.0 for E2 and 0.0 for E1. For each decision alternative draw a straight line from the point representing the of (A, El) at p = 0.0 to the point denoting the payoff of (A, E2) at p = 1.0. Every point on that line payoff is the expected value of alternative A for the respective p. The solid line at the top of the graph shows the highest expected payoff for every value of p, and the decision alternative associated with that payoff.