THE SPECIALIST WITH A UNIVERSAL MIND

MULTICRITERIA DECISION MAKING

by Andrew Vazsonyi
University of San Francisco, McLaren School of Business

Most of us believe that when teaching decision sciences we must emphasize the interaction between the model and the decision maker. However, we find it difficult to implement this fundamental principle, because we may lose the student in the mathematical intricacies of an optimizing model and miss the essence of the problem of making decisions. I believe that we can teach multicriteria decision making using very simple models, and focus on many aspects of the essence of decision sciences. Moreover, we can avoid all computational and graphic difficulties by using spreadsheets.

Let me illustrate the point by presenting a practical case study used by Kepner and Tregoe: the problem of selecting a software vendor. I'll use the example as a coat hanger, on which to ``hang'' some basic concepts of management science.

The Vendor Selection Problem

The student must understand that management needs to be helped and stimulated to agree on a set of attributes to be used in the selection process. Suppose that decision conferencing leads to the list:

• Quality of software
• Hardware performance
• User support
• Cost for five years
• Vendor reputation
• Application programs
• Delivery reliability
• Growth potential

The next step is ranking the attributes in decreasing importance and assigning weights of importance. Each vendor is scored on each attribute, using a scale of 0 to 10. Then the utility of each vendor is calculated by using a linear utility function of the form

(w1*x1) + (w2*x2) + (w3*x3) + ... + (w8*x8),

where the x's are the numeric values of the attributes, and the w's, the weights, that is, the measures of the relative importance of the attributes. This is the only model used in the analysis. The central problem is to assess the eight parameters of the model. The student is not confused by mathematical complexities.

The Assessment Problem

The parameters must be assessed so that insight and consensus is reached by the decision makers. The task is accomplished by WHAT-IF dialogs where the impact of alternate values of the parameters is reviewed. The tool to accomplish the task is the numeric and graphic images presented by the computer screen. These include such information as

• The ranking of the vendors
• Horizontal bar charts with pair-wise comparisons (Exhibit 1)
• Tradeoff charts for comparison of vendors for any pair of attributes (Exhibit 2).

There are only a few alternatives, and there is no room for optimization techniques. Managers make pair-wise comparisons between alternatives that differ only on two attributes, using the concept of caeteris paribus. Introducing intermediate phantom alternatives builds the chain to compare real alternatives. The concept of dominance filters out inferior solutions, and the concept of Pareto efficient solutions identifies the solution space. Samples of other concepts of decision sciences to ``hang'' on the tree are: indifference, perturbations, robust vs. fragile solutions.

More Complex Models

I deliberately ignored discussing hierarchical and nonlinear models, and the Analytical Hierarchy process, because I wanted to stay away from any complicating mathematical features. However, the spreadsheet approach of using tradeoffs and WHAT-IF analysis can be extended to more complex models, without requiring sophisticated mathematics. The references below provide source material to develop more advanced models.

Conclusions

We ought to teach multiattribute decision making in the early part of our curriculum to empower the instructor to discuss important concepts of management science without being burdened by complex mathematical models. Such concepts include: satisficing versus optimization, normative versus descriptive models, decision support and facilitation, WHAT-IF and sensitivity analysis, tradeoff, utility, value and goal-oriented thinking, dominated and Pareto efficient solutions, robust versus fragile solutions, the principle of caeteris paribus, phantom alternatives, group decision making and decision conferencing.

Working with such quantitative models, carrying out calculations, and performing graphic analysis, including floating graphic representations, is made possible by spreadsheets systems.

References

Keeney, R.L., , Harvard University Press, 1992.

Keeney, R.L. and Raiffa, Howard, , John Wiley & Sons, 1976.

Keptner, C.H. and Tregoe, B.B., , Princeton Press, Princeton, NJ, 1981.

Saaty, T.L., , Wadsworth, 1982.

Dr. Andrew Vazsonyi
156 Oak Island Dr.
Santa Rosa, CA 95409
707-539-0272
fax: 707-537-1833