Aiding Insight III

By Dennis Buede

This is the third survey of decision analysis software published by OR/MS Today since April 1993. There are a number of new packages since June 1994 (the last survey), but most of the packages are upgraded and improved ones. There remains a modest amount of turnover in the market of decision analysis software.

Decision analysis is the discipline of evaluating complex alternatives in terms of values (what we care about) and uncertainty (what we know and do not know). Experienced decision analysts and educators stress that the benefits of decision analysis are insight into how the defined alternatives differ from one another, and generating suggestions for new and improved alternatives. Too many critics stress the use of numbers to quantify subjective values and uncertainties without realizing the power of quantitative analysis for generating qualitative insight.

In order to create a decision analysis model, it is necessary to create the model structure and elicit the probabilities and values to populate the model for computation. Trained analysts are still providing most of the structuring support, although we continue to see advances in the software market for model structuring. Elicitation support involves queries about the values for probabilities, value functions for evaluating alternatives, value weights for measuring the trade-offs amongst objectives and risk preference.

Once the structure and numbers are in place, the analysis can begin. Much more is involved than computing the expected, weighted utility of each alternative. If we stopped there, decision makers would not get much insight. We have to examine the sensitivity of the expected, weighted utility to key probabilities, weights and risk preference parameters. As part of the sensitivity analysis, we can calculate the value of perfect information for uncertainties that have been explicitly modeled.

Additional quantitative comparisons include the direct comparison of weighted utility for two alternatives on all of the objectives and the comparison of all of the alternatives on any two selected objectives showing the Pareto optimality for these two objectives. In my experience, these last two comparisons have enabled substantial closure by the decision makers on the preferred alternative.

Survey questions
We have revised the questions that we sent to the software vendors this year, primarily because the move to graphical user interfaces is now nearly complete. In addition, de facto software standards have evolved. These questions are again easy to answer and were faxed to known decision analysis software vendors and many other software vendors who might have introduced decision analysis packages.

I reviewed the responses and eliminated those packages that were clearly not oriented to some part of decision analysis. This elimination is not a reflection of the value of these packages but simply my judgment that the package did not substantially deal with values or uncertainties associated with decision-making alternatives.

I want to stress that the information presented here is the best and most comprehensive that we could assemble in a short period of time; yet the information is not sufficiently complete to suggest that you, the reader, could make an informed choice of which package to select. We have provided a space for the vendor to describe concisely the most important aspects of the software.

Survey Results
There are 29 entries in this survey: 17 repeats and 12 new packages. These are the same percentages as last time. The PC with Windows-based operating system is still the dominant platform for software in this limited market.

The primary focus of the software vendors continues to be on analysis features rather than support for non-analysts in structuring the problem and a more complete set of elicitation support. These analysis features are useful to all users and critical for the insight that decision makers need.

However, the result is that there are relatively few of us that are qualified and comfortable in using these packages. These packages will never become as widespread as spreadsheet and database packages as long as the user has to be a highly educated analyst. For example, at the just finished Washington, D.C., INFORMS meeting, one educator lamented the lack of a software package that was sophisticated enough for a graduate level course on decision analysis, but easy enough to learn and use that it could fit into a semester level course.

The fastest growing segment of this software market is that oriented to group decision support. This segment demands the presence of a trained analyst, commands a substantially larger up-front and recurring monetary investment, and a substantial investment of the time of high-level members of the organization. The "glitz" of this market segment is high right now; time will tell about the return on investment as perceived by decision makers.

There is now some limited support for structuring the decision problem. Expert Choice has a bottom-up structuring facility that allows the user to enter pros and cons of alternatives and then organize these pros and cons into a hierarchy of objectives. Criterium and HIVIEW have a brainstorming feature that allows the user to enter objectives and graphically form them into a hierarchy.

There is still substantial room for improvement, yet the current set of packages afford the analyst with plenty of choices for selecting a powerful decision analysis tool.

Dennis Buede is an associate professor of systems engineering at the School of Information Technology and Engineering, George Mason University.

OR/MS Today copyright © 1997, 1998 by the Institute for Operations Research and the Management Sciences. All rights reserved.

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com

Web Site © Copyright 1997, 1998 by Lionheart Publishing, Inc. All rights reserved.
Web Design by Premier Web Designs, e-mail lionwebmaster@preweb.com