Multi-objective Decision Analysis

Multi-objective Decision Analysis

Managing Trade-offs and Uncertainty
Clinton W. Brownley, Ph.D.

Multi-objective Decision Analysis: Managing Trade-offs and Uncertainty Copyright Business Expert Press, 2013. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any meanselectronic, mechanical, photocopy, recording, or any other except for brief quotations, not to exceed 400 words, without the prior permission of the publisher. First published in 2013 by Business Expert Press, LLC 222 East 46th Street, New York, NY 10017 www.businessexpertpress.com ISBN-13: 978-1-60649-452-3 (paperback) ISBN-13: 978-1-60649-453-0 (e-book) Business Expert Press Quantitative Approaches to Decision MakingCollection Collection ISSN: 2163-9515 (print) Collection ISSN: 2163-9582 (electronic) Cover and interior design by Exeter Premedia Services Private Ltd., Chennai, India First edition: 2013 10 9 8 7 6 5 4 3 2 1 Printed in the United States of America.

For my wife, Anushka, who insists goals can be achieved through focus, e nthusiasm, and perseverance

Abstract
Whether managing strategy, operations, or products, making the best decision in a complex, uncertain business environment is challenging. One of the major difficulties facing decision makers is that they often have multiple, competing objectives, which means trade-offs will need to be made. To further complicate matters, uncertainty in the business environment makes it hard to explicitly understand how different objectives will impact potential outcomes. Fortunately, these problems can be solved with a structured framework for multiobjective decision analysis that measures trade-offs among objectives and incorporates uncertainties and risk preferences. This book is designed to help decision makers by providing such an analysis framework implemented as a simple spreadsheet tool. This framework helps structure the decision-making process by identifying what information is needed in order to make the decision, defining how that information should be combined to make the decision, and, finally, providing quantifiable evidence to clearly communicate and justify the final decision. The process itself involves minimal overhead and is perfect for busy professionals who need a simple, structured process for making, tracking, and communicating decisions. With this process, decision making is made more efficient by focusing only on information and factors that are well defined, measureable, and relevant to the decision at hand. The clear characterization of the decision required by the framework ensures that a decision can be traced and is consistent with the intended objectives and organizational values. Using this structured decision-making framework, anyone can effectively and consistently make better decisions to gain a competitive and strategic advantage.

Keywords
decision making, decision analysis, decision modeling, strategic decisions, business decisions, how to decide, trade-offs, multiobjective, values, weights, value functions, objectives, measures, alternatives, uncertainty, probability, discrete, continuous, linear, exponential, expected value, utility, expected utility, risk tolerance, certainty equivalents

Contents
Acknowledgments ix Chapter 1 Introduction to M ultiobjective Decision Analysis1 Chapter 2 Structuring Objectives and Developing Alternatives19 Chapter 3 Value Functions and Preference Weights45 Chapter 4 Uncertainty: Probability Distributions and Expected Value73 Chapter 5 Uncertainty: Risk Tolerance and Expected Utility95 Chapter 6 Multiobjective Decision Analysis Under Uncertainty111 Chapter 7 Conclusion133 Notes145 References151 Index157

Acknowledgments
A book like this cannot be written without help from many people. Four in particular played key roles. First and foremost, I want to thank my wife, Anushka, for being patient and supportive during all of the nights and weekends I spent writing. An insightful reviewer, she also suggested many ways to make the book more consistent, clear, and concise. Second, I want to thank Steven Nahmias for reviewing the manuscript and providing helpful suggestions. Third, I want to recognize Cindy Durand for providing excellent production assistance. She provided superb guidance on gathering permissions, organizing the images and tables, and compiling the index. Finally, I owe a special debt to the collections editor, Don Stengel, for shepherding this book from start to finish, improving it with his editorial direction, and providing a cheerful, professional hand throughout. I also want to thank the staff of Business Expert Press, especially Scott Isenberg and David Parker, for their support and assistance with this project. Scott provided seasoned advice on an assortment of topics and was also tremendously helpful during the manuscript review process. Finally, I want to acknowledge that Multi-objective Decision Analysis reflects an intellectual journey as well as a writing project. I first became interested in decision analysis as an aid to judgment and decision making while I was a student at Carnegie Mellon University. At CMU, I had the great fortune to receive instruction, research experience, and inspiration from many distinguished decision scientists and operations researchers, including Paul Fischbeck, Baruch Fischhoff, George Loewenstein, Don Moore, Otto Toby Davis, Robyn Dawes, Herbert Simon, M. Granger Morgan, Jonathan Caulkins, Alfred Blumstein, Michael Trick, and Michael Johnson. Since then, I have enriched my understanding of decision analysis and other effective techniques for strategic decision making by studying and implementing the methodologies of many other exceptional decision scientists, including Ward Edwards, Detlof von Winterfeldt, Craig Kirkwood, Howard Raiffa, Ronald Howard, Robert Schlaifer, Ralph Keeney,

x ACKNOWLEDGMENTS

obert Clemen, Robert Winkler, Reid Hastie, Harold Sox, Peter Moore, R Howard Thomas, John Hammond, John Magee, Robin Hogarth, H illel Einhorn, Rex Brown, Kenneth Hammond, Daniel Kahneman, Amos Tversky, Richard Thaler, Thomas Gilovich, Paul Slovic, Max Bazerman, Scott Plous, Jeffrey Keisler, Paul Goodwin, George Wright, Jacob Ulvila, David Hertz, Samuel Bodily, Myriam Hunink, Paul Glaziou, and Gerd Gigerenzer. Each of these individuals has contributed to my understanding of decision analysis as an aid to judgment and decision making, and this book benefits enormously from the theoretical and applications-based advances they pioneered.

CHAPTER 1

Introduction to Multiobjective Decision Analysis

Your best hope for a good decision outcome is a good decision process. J. Russo & P. Schoemaker Todays business environment is fraught with complexity and uncertainty. A variety of factors contribute to such complexitythe desire to achieve multiple objectives at once, the wish to address the values and a ttitudes toward risk of multiple stakeholders, the difficulty of identifying suitable alternatives, the challenge of measuring intangibles, and the often impossibility of precisely predicting the future consequences of alternatives. While it would be magnificent if it were otherwise, complexity is inherent to the business environment, so it cannot be avoided. Such complexity makes it incredibly difficult to make important decisions informally in a defensible manner. And in todays high-stakes business environment, managers need to be able to justify and defend their decisions to a variety of impacted groups, including shareholders, bosses, co-workers, the public, and themselves. Since informal analysis is likely to be insufficient for most key business decisions, professionals need a formal methodology and set of tools they can use to make and justify their decisions.

What Constitutes a Decision?

Since multiobjective decision analysis is a methodology that helps p eople make informed decisions, it is important to first understand what a decision is. A decision is an opportunity to make a choice between at least two

MULTI-OBJECTIVE DECISION ANALYSIS

ifferent things.1 That is, for a decision to exist there must be at least two d alternatives. Not having any alternatives may be a problem, but its not a decision problem. And, of course, the attractiveness of the alternatives matters. It can be very frustrating to face a decision situation in which there seem to be no good alternatives. Many important decisions involve alternatives that lead to d ifferent consequences. Consider the decision of whether to build a new manufacturing plant. The consequences of building the new facility are likely to be significantly different from the consequences of not building the facility. This makes the decision meaningful. If the potential consequences are not different, that is, if the alternatives result in the same consequences, then the choice between alternatives isnt very meaningful. Of course, one of the most challenging aspects of any decision situation is deciding on the factors that are important for evaluating the consequences of the alternatives, that is, selecting the values, objectives, and evaluation measures.2 Consider, once again, the decision of whether to build a new manufacturing plant. What factors would you use to evaluate whether or not to build the facility? A few factors you may consider are expected revenue, market share, product mix, time to completion, and cost. Which factors should you use to make the decision? Fortunately, though sometimes frustratingly, there is no one-and-only-one correct answer to this question. As the decision maker, you should include all of the factors you consider important for evaluating the alternatives in the given decision context. Since generating and structuring the values, objectives, and evaluation measures is a challenging, though incred ibly valuable, step in any decision-making process, these activities are discussed in more detail in the next chapter. Finally, many important decisions involve uncertainty about the consequences of the alternatives. That is, in many situations, we must make a decision now without the benefit of knowing with certainty what will be the future result or outcome of our decision. Consider, for one final time, the decision of whether to build a new manufacturing plant. As stated earlier, one of the factors we may use to evaluate the alternatives is the amount of revenue to be generated by the new facility. Despite our best efforts at estimation, forecasting, and simulation, the new facility doesnt even exist yet, so we cannot know with certainty the amount

Introduction to Multiobjective Decision Analysis 3

of revenue that will be generated if we build the new facility. Since quantifying and dealing explicitly with uncertainty is a tremendously important step in any decision-making process, the procedures for doing so are discussed in chapters four and five of this book.

What Are the Challenges of Decision Making?

Many important business decisions are extremely challenging to make. Often the challenge stems from the complexity of the decision itself or the environment in which its being made.3 Since multiobjective decision analysis is meant to help decision makers deal with complexity systematically, it is useful to describe the factors that contribute to decision complexity. Keeney lists several factors that make decisions complex and challenging:4 Multiple objectivesOne factor that makes decisions complex and challenging is when there are multiple, conflicting objectives. With multiple, competing objectives, it isnt possible to maximize all of the objectives with a single alternative. For example, suppose a business owner wants to purchase a c ommercial real estate property from a list of properties. The business person would certainly have and want to do as well as possible on several objectives, including maximizing the total square footage obtained and minimizing the purchase price. However, these two objectives are likely to be in conflict, or competing, with one another. That is, the property with the greatest amount of square footage is likely to have the highest price, rather than the lowest. Since a property with both the most square footage and the lowest price is unlikely to exist, the business person will need to decide on the choice strategy he or she will use to evaluate the alternatives. Some choice strategies are noncompensatory, meaning they do not allow for trade-offs among the objectives.5 An example of a n oncompensatory strategy would be if the business person chose to ignore purchase price and then used the square footage values to choose a p roperty. Other choice strategies, such as multiobjective decision analysis, are compensatory, meaning they do allow for trade-offs among the objectives. With a compensatory strategy, the business person would evaluate trade-offs between the square footage and purchase price values in order to choose a property.

MULTI-OBJECTIVE DECISION ANALYSIS

Using a compensatory choice strategy such as multiobjective decision analysis requires more mental effort and data than using a noncompensatory strategy; however, doing so is worthwhile because it enables decision makers to incorporate all of their objectives into the decision and make value trade-offs among their objectives. Good alternativesSome of the most frustrating decision situations are when it is difficult to identify good alternatives. Imagine a situation in which your existing alternatives are firing one-third of your employees, reducing everyones pay, or scaling back services. Even though these are alternatives, they arent good, so it feels like there arent any alternatives. The challenge in a situation like this is one of imagination and creativity. It takes a lot of mental effort to think of alternatives that perform well on the decision objectives. The challenge is compounded when there isnt sufficient time for brainstorming additional alternatives. IntangiblesSome decisions are challenging because they involve difficult-to-measure intangibles such as consumer satisfaction, employee morale, and product quality. Incorporating intangibles like these into a decision can be difficult because it takes additional time and mental effort to decide how to quantify these factors. As will be discussed in further detail in the next chapter, to be meaningful, the definitions and measurement scales used to quantify these factors should be u nambiguous, 6 comprehensive, direct, operational, and understandable. Long time horizonsFor many decisions, the effects of the decisions occur over a long time period rather than immediately. Research and development and policy decisions often have this characteristicthe R&D project or policy decision must be made today, but the decisions consequences will not be known for months, years, or even decades. Because of this lag between the decision and its effects, it can be very difficult to understand, or even recognize, the underlying associations or causal mechanisms between the alternatives and their consequences. This makes it very difficult to predict the future implications of the alternatives, and it also impacts the decision makers ability to learn from experience. Sequential nature of decisionsIn many cases, todays decisions affect tomorrows decisions by altering the set of alternatives available in the future

Introduction to Multiobjective Decision Analysis 5

and by altering the attractiveness of those alternatives. For example, installing a new information technology system throughout the o rganization means that certain projects will be possible and others will not be possible. Furthermore, the new system may make certain projects more feasible or attractive relative to other projects. Sequential decisions are challenging because it can be difficult to analyze the impact of todays decision on future opportunities and decisions and incorporate those effects into the analysis of todays decision before having to make the decision. Interdisciplinary substanceMany important business decisions require information from several areas of expertise. For example, launching a new product line may require information about legal matters, information technology, operations, marketing, finance, and sales. Many executives are not qualified in all of these areas, even though they are tasked with making decisions based on information from these areas; therefore, executives receive information from professionals who have expertise in each of these areas. In these situations, the interdisciplinary substance of the decisions is challenging because it affects question framing, information exchange, interpretation, coordination, and understanding. Uncertainty and riskUncertainty about the future is often one of the greatest challenges to decision making. Should I expand my product line? Will the demand be there? What are the chances of demand falling? Much of the doubt, fear, and anxiety people feel when making important decisions stems from not being able to precisely predict the future consequences of the decisions. While it is usually impossible to precisely predict the future, often there is information that can be used to quantify the uncertainty in the decision. This information may come from historical data, expert judgments, or even personal judgments. Quantifying uncertainty enables decision makers to communicate their judgments about it clearly and to incorporate the uncertainty systematically into their decision-making process.7 Attitude toward riskAlternatives frequently have different levels of risk, the likelihood of loss if the future turns out unfavorably. Generally, low-risk alternatives are associated with lower returns; whereas high-risk alternatives are associated with higher returns. For example, investing

MULTI-OBJECTIVE DECISION ANALYSIS

in a U.S. Treasury bond carries one level of risk and return and investing in a corporate bond carries another, higher level of risk and return. People differ in their tolerance for risk, so different alternatives appeal to different people. The challenge is in acknowledging and incorporating that sentiment into the decision-making process to ensure the decision is consistent with the decision makers risk preferences.8 Value trade-offsMany important decisions involve multiple objectives, and part of the challenge of making decisions with multiple objectives is in expressing the value trade-offs among the objectives. These value trade-offs indicate which objectives are relatively more important to the decision maker. For example, imagine a decision in which two objectives are decrease production costs and increase product quality. Does the decision maker prefer these objectives equally, or is one objective relatively more important than the other? How much more important is it? Depending on the circumstances, it can be quite challenging to recognize and think deeply about these preferences.9 At the same time, incorporating them into the decision-making process is tremendously valuable because it ensures the decision is based on, and consistent with, the d ecision makers values. Multiple decision makersOften a group of people, not just an individual, is responsible for making a particular decision. In these situations, it is frequently necessary to reach a consensus, or at least a majority, opinion in order to make the decision. These situations are complicated because people have different types of training and background, levels of understanding, motivations, and opinions. Even when a great deal of factual information is available, important decisions usually involve values, judgments, and trade-offs that cant be resolved with additional factual information. In these circumstances, it is important to be explicit and clear about all of the facts, values, judgments, and tradeoffs being used so that everyone has the information and u nderstanding 10 needed to make a well-informed decision.

What Is Multiobjective Decision Analysis?

In general, multiobjective decision analysis (MODA) is a structured approach to making informed decisions. More specifically, it is a philosophy,

Introduction to Multiobjective Decision Analysis 7

methodology, and collection of systematic procedures for evaluating decision alternatives in the face of multiple, conflicting objectives and uncertainty.11 According to MODA, a decision maker should choose the best alternative based on an evaluation of two factors: (1) the likelihood of the possible consequences for each alternative and (2) the decision makers preferences for the possible consequences for each alternative.12 As a methodology, the key term in the phrase MODA is the word analysis, which refers to decomposing something into separable components. In this case, it refers to breaking down the complex decision problem into a set of smaller, and hopefully, more manageable problems. These smaller problems involve the assessment of the decision makers objectives, alternatives, and preferences, as well as his or her judgments about relevant uncertainties. After the decision maker has assessed each of these smaller problems separately, MODA provides a formal m echanism that the decision maker can use to combine all of the information to identify the preferred alternative.13 The procedures of a MODA will be outlined later in this chapter, and will be described in greater detail in the remaining chapters of this book. One of the distinctive features of this methodology is that it separates the analysis of uncertainty from the analysis of preferences (i.e., values or utilities). Analysis of uncertainty refers to an assessment of the likelihood of the potential consequences; whereas analysis of preferences refers to an assessment of the attractiveness of the potential consequences to the decision maker. Uncertainty analysis relies on probability theory, subjective probability judgments, as well as historical and experimental data to assess the likelihood of experiencing the various uncertain consequences or outcomes.14 Conducting an uncertainty analysis is significantly better than using vague terms like fairly sure and pretty confident to describe the degree of uncertainty, because those terms mean different things to different people. By clarifying and making explicit the degree of uncertainty, an uncertainty analysis ensures everyone uses the same definition of uncertainty and understands the degree of uncertainty being expressed. An uncertainty analysis also improves upon the use of simple summary measures, such as the mean, because it enables the decision maker to see and understand the whole distribution and likelihood of potential consequences.

MULTI-OBJECTIVE DECISION ANALYSIS

Preference (i.e., value or utility) analysis relies on a decision makers unique set of values and preferences to assess how attractive the various consequences or outcomes are to the decision maker.15 This analysis has two components. The first component consists of specifying a preference ordering over the range of outcomes for a single evaluation criterion. For example, if one criterion is revenue, measured in dollars, and the potential outcomes for revenue range from $100 million to $200 million, then the decision maker must specify a value function that indicates the value of any dollar amount in this range to the decision maker. The second component consists of specifying a preference ordering over all of the evaluation measures; that is, making trade-offs among the measures. For example, if one evaluation measure is cost and another measure is time to completion, then the decision maker must think about whether cost is more, equally, or less preferred to time to completion, given their ranges, and be able to specify the degree of that preference.

Terminology Up to this point, many terms have been used to describe the components of decisions and MODA. Some of the terms that have been used are values, objectives, evaluation measures, and goals. There are no universal definitions of the terms values, objectives, evaluation measures, or goals. For example, some authors have referred to a specific concept as an objective while others have referred to the same concept as a goal.16 For clarity of communication, this section describes how these terms are used in this book. For example, in this book, the words objective and goal refer to two different concepts. The definitions used in this book are consistent with those presented by Keeney and Raiffa11 and Kirkwood19: Values are the areas of concern, considerations, or matters the decision maker thinks are important enough to be taken into account when evaluating alternatives. For example, values for a company considering alternative ways of rolling out an initiative may be ease of implementation, company image, and profit.

Introduction to Multiobjective Decision Analysis 9

Objectives augment values by specifying the preferred direction of movement. Thus, a company considering alternative ways of rolling out an initiative would find an alternative that is easier to implement more desirable. Similarly, an initiative that increases company image or increases profit is more desirable. Evaluation measures (a.k.a. criteria or attributes) are metrics or scales for quantifying an objective and assessing the extent to which it is achieved. For example, a company may use annual profit in dollars as the evaluation measure for the objective increase profit. Goals are thresholds for evaluation measures that a lternatives either do or do not achieve. For example, a company might have a goal of implementing an initiative within eight weeks. For a given alternative, this goal may or may not be achievable. Figure 1.1 shows a diagram of the MODA process. The components of the process are described in greater detail as follows. 1. Understand the decision contextThe first step, or component, in the iterative process is to understand, or frame, the decision context. Understanding and framing the decision context includes identifying the decision maker(s); gathering different perspectives on the decision situation; assessing initial reference points, opinions, and a ssumptions; and determining the time frame for making the decision.17 2. Identify the objectivesThe second step involves identifying and structuring the values and objectives the decision maker intends to use to assess the alternatives, as well as the associated evaluation measures. This includes brainstorming relevant objectives, separating fundamental from means objectives, and developing evaluation measures to quantify the objectives.18 Because of the need to be creative and think deeply about the decision and the relevant objectives and evaluation measures, this step is often one of the most challenging. When done well, it is also one of the most valuable steps in the decision-making process.

10

MULTI-OBJECTIVE DECISION ANALYSIS Understand decision context

Identify objectives

Model preferences

Identify alternatives

Score alternatives

Model uncertainty

Identify best alternative

Conduct sensitivity analyses

Does any component need to be refined? No Choose best alternative

Yes

Assess results

Learn from experience

Figure 1.1. The multiobjective decision analysis (MODA) process.

3. Model preferencesThe third step involves modeling the decision makers preferences over the evaluation measures. This includes specifying value or utility functions for each of the evaluation measures (value functions if there is no uncertainty; utility functions if the decision involves uncertainty) and expressing preference weights for each of the evaluation measures.19 4. Identify the alternativesThe fourth step involves identifying relevant alternatives, given the particular decision context. In practice,

Introduction to Multiobjective Decision Analysis 11

identifying the evaluation measures and the alternatives is an iterative process; however, identifying alternatives is listed as the fourth step in the process, rather than being included in the second step, to emphasize that the decision makers focus should be on specifying values and objectives. Only by understanding ones objectives can one hope to act in a directed fashion to achieve them. 5. Score the alternativesThe fifth step is to score the a lternatives. This involves assessing the consequences of each alternative with respect to each of the objectives and assigning a score to each alternativeevaluation measure pair that reflects the degree to which the c onsequences of each alternative achieve each of the associated objectives. 6. Model uncertaintyThe sixth step involves modeling uncertainty. If any of an alternatives consequences are uncertain, then it is important to assess the probabilities associated with the uncertain consequences. A few sources of probabilities include historical data, data from analogous situations, simulations and other stochastic analyses, and expert judgments.20 By enabling a decision maker to quantify the uncertainty in the decision, this step of the MODA process also adds significant value over and above informal decision-making processes. 7. Conduct sensitivity analysesAfter the proceeding steps determine the initially preferred alternative, the seventh step involves conducting sensitivity analyses to test the robustness of the preferred alternative to the models inputs. By changing many of the models inputs, including the value functions over the evaluation measures, the preference weights associated with the evaluation measures, the probabilities associated with the uncertain consequences, and even the scores given for each alternativeevaluation measure pair, the decision maker can quickly assess the sensitivity of the preferred alternative to the models inputs.21 8. Choose the best alternativeThe eighth step is simply selecting the alternative with the highest overall weighted value or utility. By selecting the alternative with the highest overall weighted value or utility, the decision maker maximizes the value of the decision, given the uncertainties involved and the decision makers preferences for the different objectives. 9. Assess results and learn from experienceThe final two steps are assessing the results of the decision and learning from experience. These

12

MULTI-OBJECTIVE DECISION ANALYSIS

steps involve reviewing the analysis that led to the final d ecision and action(s) to understand the reasons for selecting a particular course of action. Once the outcomes, or consequences, of interest have been realized, it is possible to evaluate the relationship between the analysis, the decision, and the results to draw lessons from the experience that can be applied to future decisions.  e preceding section has presented the MODA process as a r Th elatively linear process with a series of steps; however, it is important to remember that the process includes several feedback loops and is actu ally very iterative.22 For example, this book emphasizes specifying and structuring the objectives before considering the alternatives, but it is entirely possible to identify alternatives first or to do both in parallel. In practice, it is common to move among the steps relatively fluidly and iteratively as new ideas emerge, decision makers refine aspects of the decision, and additional information is collected. While subsequent chapters discuss the steps of the process in turn, keep in mind that the MODA methodology is actually an iterative process.

Prescriptive, not Descriptive The process described earlier appears to be common sense; however, few people follow the process systematically, even for important decisions. In this sense, the process is not descriptive, because it does not describe how people usually make difficult decisions under uncertainty. The process is also not normative, because it is not an idealized theory about how super-rational beings with unbounded memory and intellect should make decisions under uncertainty. Instead, the process is prescriptive, because it sets forth an approach that anyone can use to think deeply and s ystematically about real, complex problems in an uncertain world.23 In 1772, Joseph Priestly faced an important decision and asked Benjamin Franklin for advice. Rather than recommend what he should do, Benjamin Franklin advised Joseph Priestly on how he should go about making the decision.24 As you read Benjamin Franklins letter, think about the similarities between his approach and the one discussed in this book, including listing evaluation measures, weighing them, and deciding based on the balance of the analysis.

Introduction to Multiobjective Decision Analysis 13

To Joseph Priestley London, September 19, 1772 Dear Sir, In the affair of so much importance to you, wherein you ask my advice, I cannot for want of sufficient premises, advise you what to determine, but if you please I will tell you how. When these difficult cases occur, they are difficult chiefly because while we have them under consideration all the reasons pro and con are not present to the mind at the same time; but sometimes one set present themselves, and at other times another, the first being out of sight. Hence the various purposes or inclinations that alternately prevail, and the uncertainty that perplexes us. To get over this, my way is, to divide half a sheet of paper by a line into two columns, writing over the one pro, and over the other con. Then during three or four days consideration I put down under the different heads short hints of the different motives that at different times occur to me for or against the m easure. When I have thus got them all together in one view, I endeavour to e stimate their respective weights; and where I find two, one on each side, that seem equal, I strike them both out: If I find a reason pro equal to some two reasons con, I strike out the three. If I judge some two reasons con equal to some three reasons pro, I strike out the five; and thus proceeding I find at length where the balance lies; and if after a day or two of further consideration nothing new that is of importance occurs on either side, I come to a determination accordingly. And tho the weight of reasons cannot be taken with the precision of algebraic quantities, yet when each is thus c onsidered separately and comparatively, and the whole lies before me, Ithink I can judge better, and am less likely to take a rash step; and in fact I have found great advantage from this kind of equation, in what may be called moral or prudential algebra. Wishing sincerely that you may determine for the best, I am ever, my dear friend, Yours most affectionately B. Franklin

14

MULTI-OBJECTIVE DECISION ANALYSIS

As Ben Franklins decision-making process illustrates, decision makers can adjust the level of detail with which they analyze a decision problem based on the complexity and importance of the decision. At the same time, its important to remember that for complex, consequential decisions, following a structured decision-making process systematically can result in better decisions, communication, and implementation, and, ultimately, better results.

Why Is Multiobjective Decision Analysis Important?

Promotes clear thinkingOne of the advantages of using MODA to make decisions is that it promotes clear thinking. Without this methodology, the thought process can remain muddledobjectives entangled with alternatives entangled with uncertainties entangled with values. By decomposing decisions into these separate components and addressing them individually, decision makers gain a greater understanding of the decision situation and can use their time more effectively to identify the best course of action.25 Increases comprehension and insightsIn fact, the methodology can lead to comprehension and insights that wouldnt be apparent from an intuitive, gut reaction decision-making process. For example, the brainstorming and creativity the methodology requires can result in the generation of better evaluation measures and alternatives.26 Moreover, themethodology may show that there is a difference between the alternative the decision maker intuitively prefers and the alternative that should be preferred based on the analysis. When this is the case, the d ecision maker can explore the reasons for the discrepancy to achieve an even better grasp of the decision situation and reasons for preferring one alternative to the others. Explains decision rationaleIn addition, because the m ethodology forces a decision maker to be clear and explicit about the data and judgments used to make a decision, it is possible to refer back to the analysis and model inputs to understand why the decision maker chose a particular course of action.27 The fact that this methodology is t raceable means that it can be used to support and defend the rationale used to make

Introduction to Multiobjective Decision Analysis 15

a decision. This can be valuable when the decision needs to be j ustified to others, including bosses, co-workers, regulators, and the g eneral public. Facilitates communication and understandingAnother advantage of MODA is that it facilitates communication and understanding among multiple stakeholders.28 Even if there is initial disagreement among the stakeholders, this methodology can be used to elucidate each persons position so everyone gains a greater understanding of the issues involved and the reasons for the conflict. It may even show that, despite different stances on an issue, the issue is not worth debating because it does not affect which alternative should be chosen. Enhances commitment to actionFinally, one additional advantage of MODA is that, by enabling multiple stakeholders to participate in the decision-making process, it helps them develop a shared understanding of the decision situation. When there are initial disagreements, a shared understanding of a situation usually decreases the differences of opinion among the people addressing the problem, or at least the degree or strength of their views.29 These characteristics of the methodology, encouraging involvement in the process and promoting a shared understanding of the decision, increase the likelihood that a group will be committed to the preferred course of action.

The Focus of This Book

This book is about making multiobjective decisions under uncertainty. It describes how to structure and solve these types of problems using spreadsheets. The techniques and examples used in this book cover the use of multiple evaluation measures, discrete and continuous value and utility functions, preference weights, and expected value and utility calculations. Given the focus of this book, there is little emphasis on the development of decision trees to represent decision situations; however, this structure still underlies these decisions. For more information about decision trees, please see Making Hard Decisions: An Introduction to Decision Analysis by Robert Clemen or Decision Analysis for Management Judgment by Paul Goodwin and George Wright.

16

MULTI-OBJECTIVE DECISION ANALYSIS

How to Read This Book

This book explains how to conduct a MODA under uncertainty using spreadsheets. Chapter 2 presents procedures for developing and structuring values, objectives, and evaluation measures. The chapter also discusses how to identify alternatives and deal with situations in which there are too many or too few alternatives. Chapter 3 presents procedures for specifying value functions over each of the evaluation measures. Piecewise linear and exponential functions are shown to handle discrete and continuous evaluation measures, respectively. The chapter also presents procedures for articulating the preference weights associated with evaluation measures. Chapter 4 introduces the use of probability to quantify uncertainty, explains how to determine discrete and continuous probability distributions, and shows how to use those distributions to calculate expected values. Chapter 5 introduces the concept of risk tolerance, explains how to determine utility functions, and shows how to use certainty equivalents to identify the preferred alternative. Chapter 6 synthesizes all of the material explored in the preceding chapters by demonstrating how to conduct a multiobjective decision analysis under uncertainty using a spreadsheet. Finally, Chapter 7 concludes the book by presenting extensions to the methodology developed in this book and offering references to additional resources on those topics.

Key Points
Professionals need a systematic methodology and set of tools they can use to make and justify their decisions because complexity and uncertainty make it incredibly difficult to make important decisions informally in a defensible manner. Multiobjective decision analysis is a methodology and collection of systematic procedures for evaluating decisions in the face of multiple, conflicting objectives and uncertainty. The methodology involves breaking down a complex decision

Introduction to Multiobjective Decision Analysis 17

problem into a set of smaller, and hopefully, more manageable problems. After the decision maker has assessed each of these smaller problems separately, multiobjective decision analysis provides a formal mechanism the decision maker can use to combine all of the information to identify the preferred alternative. Multiobjective decision analysis promotes clear thinking, leads to comprehension and insights that wouldnt be apparent from an intuitive, gut reaction decision-making process, creates an audit trail that can be used to support and defend the rationale used to make a decision, facilitates communication and understanding among multiple stakeholders, and increases the likelihood that a group will be committed to the preferred course of action by promoting a shared understanding of the decision situation.

Index
A Alternatives, 70, 129 certainty equivalents and, 102108 determining overall values for, 6371 developing good, 3844 fully featured product, developing, 70 good, 4 half-featured product, developing, 70 identifying, with creative thinking, 3941 strategy generation tables, 4041 value-focused thinking, 3940 infinite number of, 133136 low-featured product, developing, 70 reducing number of, 4142 status quo, maintaining, 70 techniques for identifying, 39 under uncertainty, developing, 4244 C Certainty equivalents and alternatives, 102108 calculating, 102118, 121130 relationship between expected utility and, 102 and spreadsheet, 102, 104, 124123 Certainty equivalent value, 121123, 130, 138139 Challenges, of decision making attitude toward risk, 56 good alternatives, 4 intangibles, 4 interdisciplinary substance, 5 long time horizons, 4 multiple decision makers, 6 multiple objectives, 34 sequential nature of decisions, 45 uncertainty and risk, 5 value trade-offs, 6 Common business decisions, 84, 140 Company image (CI), 5356 and increment, 5356 Constructed evaluation measures, 3135 developing, 3235 picture scale, 34 proxy evaluation measures, 3435 scale with defined levels, 3233 weighted scale, 33 for perception of company image, 3132 Continuous probability, 8592 Cumulative distribution function (CDF), 8889 D Decision framing, 1920 Decision making challenges of, 36 Discrete Probability, 7885 E Eliciting probability, 7882 Evaluation measures, 9, 3038, 6769, 123, 128. See also Utility function constructed, 3135 cost, second, 67 customer service, third, 67 decreasing preference, 4951 desirable properties of, 3538 measurable, 3536 operational, 3637 understandable, 3738 natural, 3031 preference weights, 5153 proxy, 3435 single-dimensional value functions, 4651 importance of swinging, 62, 63 increasing preference, 4951, 58

158 Index

natural, 3031 profit, first, 67 single-dimensional value functions, 62 time-to-manufacture, final, 6869 Expected utility, 85, 9697, 99, 102103, 116, 117, 119 Expected value, 115116 and alternative, 104105 alternative way to calculate, 91 calculating, 8385 and certainty equivalent, 122 of manufacturing costs, 92 probability distributions and, 7393 and uncertainties,137138 Exponential constant, 5761 Exponential value functions, 5051 determining, 5662 Exponential utility functions, 97103 Extended PearsonTukey Approximation, 85, 9092, 104 F Framing, decision, 1920 G Goals, 9 Good alternatives, 3844 H Hedging, 43 I Identifying objectives, 2125 considering problems and shortcomings, 22 determining generic objectives, 2425 determining strategic objectives, 25 developing wish list, 21 identifying alternatives, 21 identifying goals, constraints, and guidelines, 23 predicting consequences, 2223 Increment and company image (CI), 5356 Insuring, 44 Intangibles, 4

Interdisciplinary substance, 5 Interdependent uncertainties, 136139 L Long time horizons, 4 M Manufacturing costs, 8692 Multiple objectives, 34 and uncertainity, 111116 Multiobjective decision analysis (MODA) defined, 67 extension to, 133143 importance of, 1415 process of, 912 Multiple decision makers, 6 N Natural evaluation measures, 3031 Net profit, 103110, 136138, 141 Net present value (NPV), 141 Normalized exponential constants (nec), 59, 60 Normalized mid-value (nmv), 59, 60 O Objectives, 6, 9, 27, 128 brainstorming, 2025 structuring, 2530 separating fundamental from means objectives, 2627 stopping the structuring process, 2930 structuring fundamental objectives hierarchies, 2729 techniques for identifying, 2125 considering problems and shortcomings, 22 determining generic objectives, 2425 determining strategic objectives, 25 developing wish list, 21 identifying alternatives, 21 identifying goals, constraints, and guidelines, 23 predicting consequences, 2223

 Index 159

Overall certainty equivalent values, 130 Overall Weighted Values, 7071 P PearsonTukey Approximation, extended, 85, 9092, 104 Picture scale, 34 Piecewise linear single-dimensional value functions, 4950 determining, 5356 Portfolio Decision Analysis, 135 Power-additive utility function, 116118 Preference analysis, 78 Preference weights, 5153, 69, 128129 determining, 6163 Present value (PV) formula, 140 Probabilistic independence, 121 Probability, 129 anchoring and adjustment, 7778 assuming certainty, 76 availability, 7677 continuous, determining, 8592 determining discrete, 7885 elicit, preparing to, 7980 eliciting, 8082 expected value, 8285 misunderstanding, 75 relying on heuristics, 76 representativeness, 77 verbal descriptions, using, 7576 verifying, 8285 Product manager, 8690, 96 Property of expected values, 122, 138 Proxy evaluation measures, 3435 R Risk averse, 9599, 101, 105, 108, 112, 114, 116, 118, 122, 138 aversion, 9597, 100, 107108, 111, 134135, 138 attitude toward, 56 neutral, 9599, 105, 116, 118, 122 seeking, 9698, 118 tolerence, 98102 uncertainty and, 5

Risk aversion, 9597, 100, 107108, 111, 134135, 142 Risk-averse decision makers, 9698, 118 Risk-neutral decision makers, 97, 98, 105, 114 Risk-seeking decision makers, 9698, 114 Risk sharing, 43 Risk tolerence, 98102. See also Utility function decision makers, 98101 determining multiobjective, 118121 determining, 98102 Risky alternative, 95, 9799, 103104, 108, 114118 S Scale, with defined levels, 3233 Scores. See Levels Screening criteria, 4142 Sequencing, 43 Sequential nature of decisions, 45 Single-dimensional value functions, 4651, 6162, 70, 111, 117, 128 exponential, 5051 piecewise linear, 4950 Spreadsheet, 123123 and certainty equivalent, 122124 layoutmultiobjective decision analysis (no uncertainty), 64, 114, 120 Strategic Decision Making, 136 Strategy generation table, 4041 Structuring objectives, 2530 separating fundamental from means objectives, 2627 stopping the structuring process, 2930 structuring fundamental objectives hierarchies, 2729 Swinging evaluation measure, 62, 63

160 Index

U Uncertainty, 4244 hedging, 43 insuring, 44 and risk, 5 risk sharing, 43 sequencing, 43 Uncertainty analysis, 7 Utility expected, 85, 9697, 99, 102103, 116, 117, 119 function, 9798 Utility function determining, 9798 exponential, 97103 exponential multiobjective, 117118 power-additive, 116118

V Value-focused thinking, 3940 Value functions, 6769, 123, 128 Values, 8 brainstorming, 2025 Value trade-offs, 6 Verifying probability, 8285 W Weak Law of Large Numbers, 83 Weighted scale, 33 Weighted single-dimensional certainty equivalents, 129130 Weighted single-dimensional values, 70