October 2000 Decision Analysis: Aiding Insight V Fifth biennial survey reveals significant improvements in visualization of model results and sensitivities By Daniel T. Maxwell This is the fifth biennial survey of decision analysis software published in OR/MS Today. Over this time period, continuously increasing computational power and advances in computer graphics have facilitated tremendous advances in the power and usability of these software tools. The response rate for this survey indicates that the number of software developers is stabilizing, as there were 33 packages compared with 32 entries previously. Most of the vendors have been around for multiple surveys, and the market for potential software users continues to grow. Decision analysis is a process, supported by quantitative modeling and structured elicitation techniques, designed to improve the quality of complex decisions. The intent of this process is to provide a decision-maker insight into their preferences, particularly as they relate to the multiple objectives they are attempting to simultaneously achieve, and the uncertainties that are resident in any decision situation. The four stages of the decision analysis process are: 1) problem structuring, 2) value elicitation, 3)probability elicitation, and 4) the exploration of computed results (sensitivity analysis). The structuring stage consists of the qualitative identification of a decision-maker's objectives, key variables that could influence outcomes, decision alternatives and their inter-relationships. The value and probability elicitation stages support the systematic quantification of a decision-maker's preferences (values) and the quantification of relevant uncertainties. Sensitivity analysis explores the robustness of the model, as well as the implications of a decision-maker's choice given different outcomes or preferences. The process is iterative. Therefore, decision analysis software should ideally support all four stages of the process and facilitate changes to models relatively easily. Historically, decision analytic software has focused on developing functionality that supports computing results and sensitivity analysis within a given model structure. Fortunately, the last couple of years have witnessed significant improvements in the usability of these tools for model structuring and quantitative elicitation. Additionally, group modeling tools continue to emerge that allow analysts to work with multiple stakeholders in collective groups and distributed locations. The Survey The survey was administered by Peter Horner and Jaime Tanner of the OR/MS Today staff. Questions were similar to those included in previous surveys. They are easy to answer, consisting largely of yes/ no and multiple choice questions. Two questions were added to address explicitly the functionality of the software for group elicitation. The survey was sent to known vendors of decision analysis software, as well as various other vendors that may have an interest in these types of products. The results presented here summarize the responses provided by the vendors. The information presented should be used to support initial screening of available decision analytic products. Readers are strongly encouraged to conduct more thorough research into the software packages and the techniques they implement before making a purchase decision. My experience is that software performance does not necessarily correlate to cost. That is, the more expensive packages may not be the most powerful. It is also my experience that not all practitioners of decision analysis agree on the validity of underlying computational algorithms in these software packages. Potential users (especially analysts) of any package should become familiar, and comfortable, with the supporting logic and algorithms as part of the software selection process. The Results Twenty vendors responded to the survey. Nine packages are new entries. Windows remains the predominant operating system. The number of packages supporting UNIX is increasing, and Web browser based packages are an emerging feature. One package, Mesa Vista, supports the Linux operating system. Two of the new offerings, "High Priority" and "On Balance," are from Krysalis Ltd., a long-standing producer of decision analysis software. In general, the software packages identified in the survey results fall into two classes. They emphasize the structuring and elicitation of complex value structures, or the elicitation and analysis of complex probability (uncertainty) models. Three packages — Criterium Decision Plus 3.0, Decision Tools Suite Professional 4.0. and EXSYS Corvid — provide graphical support for both value and probability elicitation. In the case of value modeling tools, the development focus continues to move toward group support and the use of Web-based applications to improve modeling convenience. Many of the packages are also developing interfaces to standard software, such as EXCEL and ACCESS. Additionally, there are niche packages that support specialized applications of decision analysis to a particular class of problems; Equity, by Krysalis, for resource allocation modeling is an example. The packages that emphasize uncertainty modeling are introducing features that facilitate "automated" model development. Both Hugin and Norsys indicate they have implemented features that allow the software to learn from data. Additionally, developers continue to implement algorithms that improve the computational efficiency of fully specified models. In general, all of the packages continue to make significant headway in improving the ability to visualize model results and sensitivities. It is my opinion that some of these tools have matured sufficiently in sophistication and ease of use to provide a suitable complement to a graduate course in decision analysis. There remains room for improvement, particularly in the area of problem structuring support. However, this list contains software products in many price ranges that could be a powerful component of an OR analyst's tool kit. Dan Maxwell is a Principal technical staff member of the MITRE Corporation ORSA Department in the Information Technology Division in McLean, Va. He has taught graduate level decision analysis at the George Mason University School of Information Technology and Engineering. He can be reached via e-mail at: dmaxwell@mitre.org. Be sure to look at the accompanying survey for more information. Table of Contents OR/MS Today Home Page OR/MS Today copyright © 2000 by the Institute for Operations Research and the Management Sciences. All rights reserved. Lionheart Publishing, Inc. 506 Roswell Street, Suite 220, Marietta, GA 30060, USA Phone: 770-431-0867 | Fax: 770-432-6969 E-mail: lpi@lionhrtpub.com URL: http://www.lionhrtpub.com Web Site © Copyright 1999, 2000 by Lionheart Publishing, Inc. All rights reserved.