Decision Analysis - UPR


Decision analysis is a systematic approach to decision making that allows managers to tackle problems where uncertainty figures as a prominent factor. A normative model is developed to represent the decision problem, facilitate logical analysis, and produce a recommended course of action. The technique is most useful in managerial situations where risk is significant. The resultingformal model is capable of generating optimal strategies for multi-stage decision problems that involve a variety of contingencies.

Proficient decision making is the hallmark of competent managers. Anyone can make bad decisions. All four of the basic managerial functions —planning, organizing, leading, and controlling— require skillful decision making. Some seasoned managers prefer to rely on experience and intuition when dealing with strategically important decisions. They argue, in essence, that executive decisions are not amenable to quantitative approaches due to the “qualitative” nature of the problems involved. Quantification, it is said, may be useful in lower-level problems where decisions can be programmed. But, goes the doctrine, executive-level decision making does not conform to “rigid math.” (Do note thatquantitative  is not the opposite of qualitative ; see Terms below.)

Hallowed tradition notwithstanding, the view adopted here is that the classical managerial doctrinementioned above is at best a mythical holdover from the days when generals commanded their troops while mounted on handsome white steeds, and at worst just plain old nonsense. Modern military strategists are among the most assiduous users of systems analysis methods. Mother Nature also has something to say. The human brain comprises two hemispheres that are specialized in function: the right hemisphere is usually the “intuitive” side while the left is generally the “analytical” half. But the hemispheres always work in complementary fashion, sharing information constantly. There is no fixed organizational hierarchy insofar as the operation of the hemispheres is concerned. The brain functions as a unit, and both intellectual dimensions are continuously brought into play. The right hemisphere does not "look down" on its left counterpart; they work in unison —synergistically— always.

The position presented in this Website is that formal decision modeling contributes positively to managerial decision making by cross-checking and enhancing intuition with a logically robust model. Intuition and experience in turn contribute to the formal modeling process by providing important insights and knowledge that can then be incorporated into the models. The resulting synthesis yields better results than the individual parts by themselves. In addition, modeling produces a concrete representation of mental conceptions that is very useful for organizational communication in the implementation phase. Finally, the model itself can serve as a plan of action as well as an instrument for project control.

The outcome of a decision problem depends on two things: the  action alternative  chosen by the decision maker and the uncontrollable  state of nature  which happens to occur. Outcomes may be favorable or unfavorable to the decision maker. The payoff matrix is a way of showing the value to the decision maker of a set of possible outcomes.

States of nature, being uncontrollable, require probabilities to denote the likelihood of their occurrence. Except for the Laplace Criterion, elementary decision theory models do not make use of probabilities. Consequently, they are not practical for rational decision making.

Example: ACME Road Runner Traps
ACME International is considering building an overseas manufacturing plant to produce a new, high-tech version of their world-famous Road Runner Trap. Forecasted demand for roadrunner traps is highly uncertain, depending on two interrelated factors that are difficult to assess: roadrunner population and coyote population. ACME's CEO, Wolfgang E. Cactus, believes demand for their product in the next five years could be high, medium or low. If demand were to be high and ACME builds a large plant, they expect to make $15 million in (net present value) profits. If demand were medium, ACME would clear only $3 million with a large plant. But if demand were low, ACME would incur a $6 million loss. Were ACME to build a small plant, however, Cactus is confident they would avoid losses altogether but stand to make much less profit: $3 million, $2 million and $1 million for high, medium and low demands, respectively. ACME's VP for strategic analysis, Goldie Lockes, has proposed building a mid-size plant she calls "just right." With this plant, she estimates profits would be $9 million on a high-demand scenario and $4 million under medium demand. Low demand would produce an estimated loss of only $2 million. Cactus and Lockes are in agreement about these estimates as well as the planning horizon of five years.

The initial model - always include the "do nothing" alternative.

If an action alternative X leads to payoffs at least equal to the payoffs produced by another action alternative Y  for every state of nature in Sand also generates a higher payoff for at least one state of nature, X  is said to dominate Y  and alternative Y  should be eliminated from the model.

Eliminate all dominated alternatives from the model


The decision maker ranks order the events in terms of their likelihood of occurrence.

Assign an arbitrary weight to any one event.

The decision maker specifies the proportionate weights for the remaining events.

Weighting is done by referencing any event which already has been assigned a weight.

Normalize the weights to obtain the probability distribution.

Place the probabilities in the appropriate columns in the matrix model ...

... or under the appropriate branches of the decision tree model.


Make a Free Website with Yola.