Upload
sara-mays
View
223
Download
0
Embed Size (px)
Citation preview
7/28/2019 Brownley Chap One
1/32
Multi-objective DecisionAnalysis
7/28/2019 Brownley Chap One
2/32
7/28/2019 Brownley Chap One
3/32
Multi-objective DecisionAnalysis
Managing Trade-offs and Uncertainty
Clinton W. Brownley, Ph.D.
7/28/2019 Brownley Chap One
4/32
Multi-objective Decision Analysis: Managing Trade-offs and Uncertainty
Copyight Business Epet Pess, 2013.
All ights eseved. No pat of this publication may be epoduced,
stoed in a etieval system, o tansmitted in any fom o by any
meanselectonic, mechanical, photocopy, ecoding, o any othe
ecept fo bief quotations, not to eceed 400 wods, without the pio
pemission of the publishe.
Fist published in 2013 by
Business Epet Pess, LLC
222 East 46th Steet, New Yok, NY 10017www.businessepetpess.com
ISBN-13: 978-1-60649-452-3 (papeback)
ISBN-13: 978-1-60649-453-0 (e-book)
Business Epet Pess Quantitative Appoaches to Decision
Making Collection
Collection ISSN: 2163-9515 (pint)
Collection ISSN: 2163-9582 (electonic)
Cove and inteio design by Eete Pemedia Sevices Pivate Ltd.,
Chennai, India
Fist edition: 2013
10 9 8 7 6 5 4 3 2 1
Pinted in the United States of Ameica.
7/28/2019 Brownley Chap One
5/32
For my wife, Anushka, who insists goals can be achieved through
focus, enthusiasm, and perseverance
7/28/2019 Brownley Chap One
6/32
Abstract
Whether managing strategy, operations, or products, making the best
decision in a complex, uncertain business environment is challenging.One o the major diculties acing decision makers is that they oten
have multiple, competing objectives, which means trade-os will need
to be made. o urther complicate matters, uncertainty in the business
environment makes it hard to explicitly understand how dierent objec-
tives will impact potential outcomes. Fortunately, these problems can be
solved with a structured ramework or multiobjective decision analysis
that measures trade-os among objectives and incorporates uncertaintiesand risk preerences.
Tis book is designed to help decision makers by providing such an anal-
ysis ramework implemented as a simple spreadsheet tool. Tis ramework
helps structure the decision-making process by identiying what inorma-
tion is needed in order to make the decision, dening how that inormation
should be combined to make the decision, and, nally, providing quanti-
able evidence to clearly communicate and justiy the nal decision.Te process itsel involves minimal overhead and is perect or busy
proessionals who need a simple, structured process or making, track-
ing, and communicating decisions. With this process, decision making is
made more ecient by ocusing only on inormation and actors that are
well dened, measureable, and relevant to the decision at hand. Te clear
characterization o the decision required by the ramework ensures that a
decision can be traced and is consistent with the intended objectives and
organizational values. Using this structured decision-making ramework,
anyone can eectively and consistently make better decisions to gain a
competitive and strategic advantage.
Keywords
decision making, decision analysis, decision modeling, strategic deci-
sions, business decisions, how to decide, trade-os, multiobjective, values,
weights, value unctions, objectives, measures, alternatives, uncertainty,
probability, discrete, continuous, linear, exponential, expected value, utility,
expected utility, risk tolerance, certainty equivalents
7/28/2019 Brownley Chap One
7/32
Contents
Acknowledgments ix
Chapter 1 Introduction to Multiobjective Decision Analysis ..............1
Chapter 2 Structuring Objectives and Developing Alternatives ........19
Chapter 3 Value Functions and Preerence Weights ..........................45
Chapter 4 Uncertainty: Probability Distributions andExpected Value ................................................................73
Chapter 5 Uncertainty: Risk olerance and Expected Utility ............95
Chapter 6 Multiobjective Decision Analysis Under Uncertainty .....111
Chapter 7 Conclusion ....................................................................133
Notes 145
References 151
Index 157
7/28/2019 Brownley Chap One
8/32
7/28/2019 Brownley Chap One
9/32
Acknowledgments
A book like this cannot be written without help rom many people. Four
in particular played key roles. First and oremost, I want to thank my
wie, Anushka, or being patient and supportive during all o the nights
and weekends I spent writing. An insightul reviewer, she also suggested
many ways to make the book more consistent, clear, and concise. Second,
I want to thank Steven Nahmias or reviewing the manuscript and pro-
viding helpul suggestions. Tird, I want to recognize Cindy Durand orproviding excellent production assistance. She provided superb guidance
on gathering permissions, organizing the images and tables, and com-
piling the index. Finally, I owe a special debt to the collections editor,
Don Stengel, or shepherding this book rom start to nish, improving it
with his editorial direction, and providing a cheerul, proessional hand
throughout.
I also want to thank the sta o Business Expert Press, especially ScottIsenberg and David Parker, or their support and assistance with this pro-
ject. Scott provided seasoned advice on an assortment o topics and was
also tremendously helpul during the manuscript review process.
Finally, I want to acknowledge that Multi-objective Decision Analysis
refects an intellectual journey as well as a writing project. I rst 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 ortune to receive instruction, research experience, and inspiration
rom many distinguished decision scientists and operations researchers,
including Paul Fischbeck, Baruch Fischho, George Loewenstein, Don
Moore, Otto oby Davis, Robyn Dawes, Herbert Simon, M. Granger
Morgan, Jonathan Caulkins, Alred Blumstein, Michael rick, and
Michael Johnson.
Since then, I have enriched my understanding o decision analysis and
other eective techniques or strategic decision making by studying and
implementing the methodologies o many other exceptional decision
scientists, including Ward Edwards, Detlo von Wintereldt, Craig
Kirkwood, Howard Raia, Ronald Howard, Robert Schlaier, Ralph Keeney,
7/28/2019 Brownley Chap One
10/32
Robert Clemen, Robert Winkler, Reid Hastie, Harold Sox, Peter Moore,
Howard Tomas, John Hammond, John Magee, Robin Hogarth, Hillel
Einhorn, Rex Brown, Kenneth Hammond, Daniel Kahneman, Amos
versky, Richard Taler, Tomas Gilovich, Paul Slovic, Max Bazerman,
Scott Plous, Jerey Keisler, Paul Goodwin, George Wright, Jacob Ulvila,
David Hertz, Samuel Bodily, Myriam Hunink, Paul Glaziou, and Gerd
Gigerenzer.
Each o these individuals has contributed to my understanding o
decision analysis as an aid to judgment and decision making, and this
book benets enormously rom the theoretical and applications-based
advances they pioneered.
x ACKNOWLEDGMENTS
7/28/2019 Brownley Chap One
11/32
CHAPTER 1
Introduction toMultiobjective Decision
Analysis
Your best hope for a good decision outcome is a good decision process
J. Russo & P. Schoemaker
odays business environment is raught with complexity and uncertainty.
A variety o actors contribute to such complexitythe desire to achieve
multiple objectives at once, the wish to address the values and attitudes
toward risk o multiple stakeholders, the diculty o identiying suitablealternatives, the challenge o measuring intangibles, and the oten
impossibility o precisely predicting the uture consequences o alternatives.
While it would be magnicent i it were otherwise, complexity is inherent
to the business environment, so it cannot be avoided.
Such complexity makes it incredibly dicult to make important
decisions inormally in a deensible manner. And in todays high-stakes
business environment, managers need to be able to justiy and deendtheir decisions to a variety o impacted groups, including shareholders,
bosses, co-workers, the public, and themselves. Since inormal analysis is
likely to be insucient or most key business decisions, proessionals need
a ormal methodology and set o tools they can use to make and justiy
their decisions.
What Constitutes a Decision?
Since multiobjective decision analysis is a methodology that helps people
make inormed decisions, it is important to rst understand what a decision
is. A decision is an opportunity to make a choice between at least two
7/28/2019 Brownley Chap One
12/32
2 MULTI-OBJECTIVE DECISION ANALYSIS
dierent things.1 Tat is, or a decision to exist there must be at least two
alternatives. Not having any alternatives may be a problem, but its not
a decision problem. And, o course, the attractiveness o the alternatives
matters. It can be very rustrating to ace a decision situation in which there
seem to be nogoodalternatives.
Many important decisions involve alternatives that lead to dierent
consequences. Consider the decision o whether to build a new
manuacturing plant. Te consequences o building the new acility are
likely to be signicantly dierent rom the consequences o not build-
ing the acility. Tis makes the decision meaningul. I the potential
consequences are not dierent, that is, i the alternatives result in the same
consequences, then the choice between alternatives isnt very meaningul.
O course, one o the most challenging aspects o any decision situa-
tion is deciding on the actors that are important or evaluating the con-
sequences o the alternatives, that is, selecting the values, objectives, and
evaluation measures.2 Consider, once again, the decision o whether to
build a new manuacturing plant. What actors wouldyou use to evaluate
whether or not to build the acility? A ew actors you may consider areexpected revenue, market share, product mix, time to completion, and
cost. Which actors should you use to make the decision? Fortunately,
though sometimes rustratingly, there is no one-and-only-one correct
answer to this question. As the decision maker, you should include all
o the actors you consider important or 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 o the alternatives. Tat is, in many situations, we must
make a decision now without the benet o knowing with certainty what
will be the uture result or outcome o our decision. Consider, or one
nal time, the decision o whether to build a new manuacturing plant.As stated earlier, one o the actors we may use to evaluate the alternatives
is the amount o revenue to be generated by the new acility. Despite
our best eorts at estimation, orecasting, and simulation, the new acil-
ity doesnt even exist yet, so we cannot know with certainty the amount
7/28/2019 Brownley Chap One
13/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 3
o revenue that will be generated i we build the new acility. Since
quantiying and dealing explicitly with uncertainty is a tremendously
important step in any decision-making process, the procedures or doing
so are discussed in chapters our and ve o this book.
What Are the Challenges of Decision Making?
Many important business decisions are extremely challenging to make.
Oten the challenge stems rom the complexity o the decision itsel or the
environment in which its being made.3 Since multiobjective decision anal-
ysis is meant to help decision makers deal with complexity systematically,
it is useul to describe the actors that contribute to decision complexity.
Keeney lists several actors that make decisions complex and challenging:4
Multiple objectivesOne actor that makes decisions complex and chal-
lenging is when there are multiple, conficting objectives. With multiple,
competing objectives, it isnt possible to maximize all o the objectives
with a single alternative.For example, suppose a business owner wants to purchase a commercial
real estate property rom a list o properties. Te business person would
certainly have and want to do as well as possible on several objectives,
including maximizing the total square ootage obtained and minimiz-
ing the purchase price. However, these two objectives are likely to be in
confict, or competing, with one another. Tat is, the property with the
greatest amount o square ootage is likely to have the highest price, ratherthan the lowest. Since a property with both the most square ootage 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 or trade-os among the objectives.5 An example o a noncompensatory
strategy would be i the business person chose to ignore purchase price and
then used the square ootage values to choose a property. Other choice strat-egies, such as multiobjective decision analysis, are compensatory, meaning
they do allow or trade-os among the objectives. With a compensatory
strategy, the business person would evaluate trade-os between the square
ootage and purchase price values in order to choose a property.
7/28/2019 Brownley Chap One
14/32
4 MULTI-OBJECTIVE DECISION ANALYSIS
Using a compensatory choice strategy such as multiobjective decision
analysis requires more mental eort and data than using a noncompensa-
tory strategy; however, doing so is worthwhile because it enables decision
makers to incorporate all o their objectives into the decision and make
value trade-os among their objectives.
Good alternativesSome o the most rustrating decision situations are
when it is dicult to identiygoodalternatives. Imagine a situation in
which your existing alternatives are ring one-third o your employees,
reducing everyones pay, or scaling back services. Even though these are
alternatives, they arent good, so it eels like there arent any alternatives.Te challenge in a situation like this is one o imagination and creativity.
It takes a lot o mental eort to think o alternatives that perorm well
on the decision objectives. Te challenge is compounded when there isnt
sucient time or brainstorming additional alternatives.
IntangiblesSome decisions are challenging because they involve
dicult-to-measure intangibles such as consumer satisaction, employee
morale, and product quality. Incorporating intangibles like these intoa decision can be dicult because it takes additional time and mental
eort to decide how to quantiy these actors. As will be discussed in
urther detail in the next chapter, to be meaningul, the denitions and
measurement scales used to quantiy these actors should be unambiguous,
comprehensive, direct, operational, and understandable.6
Long time horizonsFor many decisions, the eects o the decisions
occur over a long time period rather than immediately. Research and
development and policy decisions oten have this characteristicthe
R&D project or policy decision must be made today, but the decisions
consequences will not be known or months, years, or even decades.
Because o this lag between the decision and its eects, it can be very
dicult to understand, or even recognize, the underlying associations or
causal mechanisms between the alternatives and their consequences. Tis
makes it very dicult to predict the uture implications o the alternatives,and it also impacts the decision makers ability to learn rom experience.
Sequential nature o decisionsIn many cases, todays decisions aect
tomorrows decisions by altering the set o alternatives available in the uture
7/28/2019 Brownley Chap One
15/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 5
and by altering the attractiveness o those alternatives. For example, install-
ing a new inormation technology system throughout the organization
means that certain projects will be possible and others will not be possible.
Furthermore, the new system may make certain projects more easible or
attractive relative to other projects. Sequential decisions are challenging
because it can be dicult to analyze the impact o todays decision on
uture opportunities and decisions and incorporate those eects into the
analysis o todays decision beore having to make the decision.
Interdisciplinary substanceMany important business decisions require
inormation rom several areas o expertise. For example, launching a newproduct line may require inormation about legal matters, inormation
technology, operations, marketing, nance, and sales. Many executives
are not qualied in all o these areas, even though they are tasked with
making decisions based on inormation rom these areas; thereore, exec-
utives receive inormation rom proessionals who have expertise in each
o these areas. In these situations, the interdisciplinary substance o the
decisions is challenging because it aects question raming, inormationexchange, interpretation, coordination, and understanding.
Uncertainty and riskUncertainty about the uture is oten one o
the greatest challenges to decision making. Should I expand my product
line? Will the demand be there? What are the chances o demand alling?
Much o the doubt, ear, and anxiety people eel when making impor-
tant decisions stems rom not being able to precisely predict the uture
consequences o the decisions. While it is usually impossible to preciselypredict the uture, oten there is inormation that can be used to quantiy
the uncertainty in the decision. Tis inormation may come rom his-
torical data, expert judgments, or even personal judgments. Quantiying
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 requently have dierent levels o
risk, the likelihood o loss i the uture turns out unavorably. Generally,
low-risk alternatives are associated with lower returns; whereas high-risk
alternatives are associated with higher returns. For example, investing
7/28/2019 Brownley Chap One
16/32
6 MULTI-OBJECTIVE DECISION ANALYSIS
in a U.S. reasury bond carries one level o risk and return and invest-
ing in a corporate bond carries another, higher level o risk and return.
People dier in their tolerance or risk, so dierent alternatives appeal to
dierent people. Te challenge is in acknowledging and incorporating
that sentiment into the decision-making process to ensure the decision is
consistent with the decision makers risk preerences.8
Value trade-ofsMany important decisions involve multiple objectives,
and part o the challenge o making decisions with multiple objectives
is in expressing the value trade-os among the objectives. Tese value
trade-os indicate which objectives are relatively more important to thedecision maker. For example, imagine a decision in which two objectives
are decrease production costs and increase product quality. Does the
decision maker preer these objectives equally, or is one objective relatively
more important than the other? How much more important is it? Depend-
ing on the circumstances, it can be quite challenging to recognize and think
deeply about these preerences.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 decision makers values.
Multiple decision makersOten a group o people, not just an
individual, is responsible or making a particular decision. In these
situations, it is requently necessary to reach a consensus, or at least
a majority, opinion in order to make the decision. Tese situations
are complicated because people have dierent types o training and
background, levels o understanding, motivations, and opinions. Evenwhen a great deal o actual inormation is available, important decisions
usually involve values, judgments, and trade-os that cant be resolved with
additional actual inormation. In these circumstances, it is important to
be explicit and clear about all o the acts, values, judgments, and trade-
os being used so that everyone has the inormation and understanding
needed to make a well-inormed decision.10
What Is Multiobjective Decision Analysis?
In general, multiobjective decision analysis (MODA) is a structured
approach to making inormed decisions. More specically, it is a philosophy,
7/28/2019 Brownley Chap One
17/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 7
methodology, and collection o systematic procedures or evaluating decision
alternatives in the ace o multiple, conficting objectives and uncertainty.11
According to MODA, a decision maker should choose the best
alternative based on an evaluation o two actors: (1) the likelihood o the
possible consequences or each alternative and (2) the decision makers
preerences or the possible consequences or each alternative.12
As a methodology, the key term in the phrase MODA is the word analysis,
which reers to decomposing something into separable components. In this
case, it reers to breaking down the complex decision problem into a set o
smaller, and hopeully, more manageable problems. Tese smaller problems
involve the assessment o the decision makers objectives, alternatives, and
preerences, as well as his or her judgments about relevant uncertainties.
Ater the decision maker has assessed each o these smaller problems sepa-
rately, MODA provides a ormal mechanism that the decision maker can
use to combine all o the inormation to identiy the preerred alternative.13
Te procedures o a MODA will be outlined later in this chapter, and will be
described in greater detail in the remaining chapters o this book.
One o the distinctive eatures o this methodology is that it separatesthe analysis o uncertainty rom the analysis o preerences (i.e., values or
utilities). Analysis o uncertainty reers to an assessment o the likelihood
o the potential consequences; whereas analysis o preerences reers to
an assessment o the attractivenesso the potential consequences to the
decision maker.
Uncertainty analysis relies on probability theory, subjective
probability judgments, as well as historical and experimental data toassess the likelihood o experiencing the various uncertain consequences
or outcomes.14 Conducting an uncertainty analysis is signicantly bet-
ter than using vague terms like airly sure and pretty condent to
describe the degree o uncertainty, because those terms mean dierent
things to dierent people. By clariying and making explicit the degree
o uncertainty, an uncertainty analysis ensures everyone uses the same
denition o uncertainty and understands the degree o uncertainty beingexpressed. An uncertainty analysis also improves upon the use o simple
summary measures, such as the mean, because it enables the decision
maker to see and understand the whole distribution and likelihood o
potential consequences.
7/28/2019 Brownley Chap One
18/32
8 MULTI-OBJECTIVE DECISION ANALYSIS
Preerence (i.e., value or utility) analysis relies on a decision makers
unique set o values and preerences to assess how attractive the various
consequences or outcomes are to the decision maker.15 Tis analysis has
two components. Te rst component consists o speciying a preerence
ordering over the range o outcomes or a single evaluation criterion. For
example, i one criterion is revenue, measured in dollars, and the potential
outcomes or revenue range rom $100 million to $200 million, then the
decision maker must speciy a value unction that indicates the value o
any dollar amount in this range to the decision maker.
Te second component consists o speciying a preerence ordering
over all o the evaluation measures; that is, making trade-os among the
measures. For example, i 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 preerred to time to completion,
given their ranges, and be able to speciy the degree o that preerence.
Terminology
Up to this point, many terms have been used to describe the components
o decisions and MODA. Some o the terms that have been used are
values, objectives, evaluation measures, and goals.
Tere are no universal denitions o the terms values, objectives,
evaluation measures, or goals. For example, some authors have reerred to
a specic concept as an objective while others have reerred to the same
concept as a goal.16 For clarity o communication, this section describeshow these terms are used in this book. For example, in this book, the
words objective and goal reer to two dierent concepts. Te denitions
used in this book are consistent with those presented by Keeney and
Raia11 and Kirkwood19:
Values are the areas o concern, considerations, or matters the
decision maker thinks are important enough to be taken intoaccount when evaluating alternatives. For example, values or a
company considering alternative ways o rolling out an initiative
may be ease o implementation, company image, and prot.
7/28/2019 Brownley Chap One
19/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 9
Objectives augment values by speciying the preerred direction
o movement. Tus, a company considering alternative ways
o rolling out an initiative would nd an alternative that is
easierto implement moredesirable. Similarly, an initiative that
increases company image or increases prot is more desirable.
Evaluation measures (a.k.a. criteria or attributes) are metrics
or scales or quantiying an objective and assessing the extent
to which it is achieved. For example, a company may use
annual prot in dollars as the evaluation measure or the
objective increase prot.
Goals are thresholds or evaluation measures that alternatives
either do or do not achieve. For example, a company might
have a goal o implementing an initiative within eight
weeks. For a given alternative, this goal may or may not be
achievable.
Figure 1.1 shows a diagram o the MODA process. Te components
o the process are described in greater detail as ollows.
1. Understand the decision contextTe rst step, or component, in
the iterative process is to understand, or rame, the decision context.
Understanding and raming the decision context includes identiying
the decision maker(s); gathering dierent perspectives on the decision
situation; assessing initial reerence points, opinions, and assumptions;
and determining the time rame or making the decision.17
2. Identiy the objectivesTe second step involves identiying and
structuring the values and objectives the decision maker intends
to use to assess the alternatives, as well as the associated evalua-
tion measures. Tis includes brainstorming relevant objectives,
separating undamental rom means objectives, and developing
evaluation measures to quantiy the objectives.18 Because o the need
to be creative and think deeply about the decision and the relevantobjectives and evaluation measures, this step is oten one o the most
challenging. When done well, it is also one o the most valuable steps
in the decision-making process.
7/28/2019 Brownley Chap One
20/32
10 MULTI-OBJECTIVE DECISION ANALYSIS
3. Model preerencesTe third step involves modeling the decision
makers preerences over the evaluation measures. Tis includes
speciying value or utility unctions or each o the evaluation
measures (value unctions i there is no uncertainty; utility unctionsi the decision involves uncertainty) and expressing preerence
weights or each o the evaluation measures.19
4. Identiy the alternativesTe ourth step involves identiying rel-
evant alternatives, given the particular decision context. In practice,
Understand decision context
Identify alternatives
Model preferences
Score alternatives
Conduct sensitivity analyses
Does anycomponent need
to be refined?
Choose best alternative
No
Yes
Identify objectives
Learn from experience
Model uncertainty
Identify best alternative
Assess results
Figure 1.1. The multiobjective decision analysis (MODA) process.
7/28/2019 Brownley Chap One
21/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 11
identiying the evaluation measures and the alternatives is an iterative
process; however, identiying alternatives is listed as the ourth step
in the process, rather than being included in the second step, to
emphasize that the decision makers ocus should be on speciying
values and objectives. Only by understanding ones objectives can
one hope to act in a directed ashion to achieve them.
5. Score the alternativesTe th step is to score the alternatives. Tis
involves assessing the consequences o each alternative with respect to
each o the objectives and assigning a score to each alternativeevaluation
measure pair that refects the degree to which the consequences o each
alternative achieve each o the associated objectives.
6. Model uncertaintyTe sixth step involves modeling uncertainty. I
any o an alternatives consequences are uncertain, then it is important
to assess the probabilities associated with the uncertain consequences.
A ew sources o probabilities include historical data, data rom analo-
gous situations, simulations and other stochastic analyses, and expert
judgments.20 By enabling a decision maker to quantiy the uncer-
tainty in the decision, this step o the MODA process also adds sig-nicant value over and above inormal decision-making processes.
7. Conduct sensitivity analysesAter the proceeding steps deter-
mine the initially preerred alternative, the seventh step involves
conducting sensitivity analyses to test the robustness o the preerred
alternative to the models inputs. By changing many o the models
inputs, including the value unctions over the evaluation measures,
the preerence weights associated with the evaluation measures, theprobabilities associated with the uncertain consequences, and even
the scores given or each alternativeevaluation measure pair, the
decision maker can quickly assess the sensitivity o the preerred
alternative to the models inputs.21
8. Choose the best alternativeTe 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 deci-sion maker maximizes the value o the decision, given the uncertainties
involved and the decision makers preerences or the dierent objectives.
9.Assess results and learn rom experienceTe nal two steps are
assessing the results o the decision and learning rom experience. Tese
7/28/2019 Brownley Chap One
22/32
12 MULTI-OBJECTIVE DECISION ANALYSIS
steps involve reviewing the analysis that led to the nal decision and
action(s) to understand the reasons or selecting a particular course
o action. Once the outcomes, or consequences, o interest have been
realized, it is possible to evaluate the relationship between the analy-
sis, the decision, and the results to draw lessons rom the experience
that can be applied to uture decisions.
Te preceding section has presented the MODA process as a relatively
linear process with a series o steps; however, it is important to
remember that the process includes several eedback loops and is actu-
ally very iterative.22 For example, this book emphasizes speciying andstructuring the objectives beore considering the alternatives, but it is
entirely possible to identiy alternatives rst or to do both in parallel.
In practice, it is common to move among the steps relatively fu-
idly and iteratively as new ideas emerge, decision makers rene aspects
o the decision, and additional inormation is collected. While subse-
quent chapters discuss the steps o the process in turn, keep in mind
that the MODA methodology is actually an iterative process.
Prescriptive, not Descriptive
Te process described earlier appears to be common sense; however, ew
people ollow the process systematically, even or important decisions.
In this sense, the process is not descriptive, because it does not describe
how people usually make dicult decisions under uncertainty. Te pro-cess 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 orth an approach that anyone can use to think deeply
and systematically about real, complex problems in an uncertain world.23
In 1772, Joseph Priestly aced an important decision and asked
Benjamin Franklin or advice. Rather than recommend whathe shoulddo, Benjamin Franklin advised Joseph Priestly on howhe 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 o the analysis.
7/28/2019 Brownley Chap One
23/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 13
o Joseph Priestley
London, September 19, 1772
Dear Sir,
In the aair o so much importance to you, wherein you ask my
advice, I cannot or want o sucient premises, advise you what
to determine, but i you please I will tell you how.
When these dicult cases occur, they are dicult chiefy
because while we have them under consideration all the reasons
pro and con are not present to the mind at the same time; butsometimes one set present themselves, and at other times another,
the rst being out o sight. Hence the various purposes or inclina-
tions that alternately prevail, and the uncertainty that perplexes us.
o get over this, my way is, to divide hal a sheet o paper by
a line into two columns, writing over the one pro, and over the
other con. Ten during three or our days consideration I put down
under the dierent heads short hints o the dierent motives that atdierent times occur to me or or against the measure. When I have
thus got them all together in one view, I endeavour to estimate their
respective weights; and where I nd two, one on each side, that
seem equal, I strike them both out: I I nd a reason pro equal to
some two reasons con, I strike out the three. I I judge some two rea-
sons con equal to some three reasons pro, I strike out the ve; and
thus proceeding I nd at length where the balance lies; and i ater
a day or two o urther consideration nothing new that is o impor-
tance occurs on either side, I come to a determination accordingly.
And tho the weight o reasons cannot be taken with the
precision o algebraic quantities, yet when each is thus considered
separately and comparatively, and the whole lies beore me, I think
I can judge better, and am less likely to take a rash step; and in act
I have ound great advantage rom this kind o equation, in what
may be called moral or prudential algebra.
Wishing sincerely that you may determine or the best, I am
ever, my dear riend,
Yours most aectionately
B. Franklin
7/28/2019 Brownley Chap One
24/32
14 MULTI-OBJECTIVE DECISION ANALYSIS
As Ben Franklins decision-making process illustrates, decision makers
can adjust the level o detail with which they analyze a decision problem
based on the complexity and importance o the decision. At the same
time, its important to remember that or complex, consequential deci-
sions, ollowing 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 o the advantages o using MODA to
make decisions is that it promotes clear thinking. Without this meth-
odology, 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 o the
decision situation and can use their time more eectively to identiy the
best course o action.25
Increases comprehension and insightsIn act, the methodology
can lead to comprehension and insights that wouldnt be apparent rom
an intuitive, gut reaction decision-making process. For example, the
brainstorming and creativity the methodology requires can result in the
generation o better evaluation measures and alternatives.26 Moreover,
the methodology may show that there is a dierence between the alterna-
tive the decision maker intuitively preers and the alternative that should
be preerred based on the analysis. When this is the case, the decision
maker can explore the reasons or the discrepancy to achieve an even
better grasp o the decision situation and reasons or preerring one
alternative to the others.
Explains decision rationaleIn addition, because the methodology
orces a decision maker to be clear and explicit about the data and
judgments used to make a decision, it is possible to reer back to the
analysis and model inputs to understand why the decision maker chose a
particular course o action.27 Te act that this methodology is traceable
means that it can be used to support and deend the rationale used to make
7/28/2019 Brownley Chap One
25/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 15
a decision. Tis can be valuable when the decision needs to be justied to
others, including bosses, co-workers, regulators, and the general public.
Facilitates communication and understandingAnother advantage oMODA is that it acilitates communication and understanding among
multiple stakeholders.28 Even i there is initial disagreement among the
stakeholders, this methodology can be used to elucidate each persons
position so everyone gains a greater understanding o the issues involved
and the reasons or the confict. It may even show that, despite dierent
stances on an issue, the issue is not worth debating because it does not
aect which alternative should be chosen.
Enhances commitment to actionFinally, one additional advantage o
MODA is that, by enabling multiple stakeholders to participate in the
decision-making process, it helps them develop a shared understanding
o the decision situation. When there are initial disagreements, a shared
understanding o a situation usually decreases the dierences o opinion
among the people addressing the problem, or at least the degree or strength
o their views.29 Tese characteristics o the methodology, encouraginginvolvement in the process and promoting a shared understanding o the
decision, increase the likelihood that a group will be committed to the
preerred course o action.
The Focus of This Book
Tis book is about making multiobjective decisions under uncertainty.It describes how to structure and solve these types o problems using
spreadsheets. Te techniques and examples used in this book cover the
use o multiple evaluation measures, discrete and continuous value and
utility unctions, preerence weights, and expected value and utility
calculations.
Given the ocus o this book, there is little emphasis on the development
o decision trees to represent decision situations; however, this structure
still underlies these decisions. For more inormation about decision trees,
please seeMaking Hard Decisions: An Introduction to Decision Analysisby
Robert Clemen or Decision Analysis for Management Judgment by Paul
Goodwin and George Wright.
7/28/2019 Brownley Chap One
26/32
16 MULTI-OBJECTIVE DECISION ANALYSIS
How to Read This Book
Tis book explains how to conduct a MODA under uncertainty
using spreadsheets. Chapter 2 presents procedures or developing andstructuring values, objectives, and evaluation measures. Te chapter also
discusses how to identiy alternatives and deal with situations in which
there are too many or too ew alternatives.
Chapter 3 presents procedures or speciying value unctions over each
o the evaluation measures. Piecewise linear and exponential unctions
are shown to handle discrete and continuous evaluation measures,
respectively. Te chapter also presents procedures or articulating thepreerence weights associated with evaluation measures.
Chapter 4 introduces the use o probability to quantiy uncer-
tainty, 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 o risk tolerance, explains how to
determine utility unctions, and shows how to use certainty equivalentsto identiy the preerred alternative.
Chapter 6 synthesizes all o 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 oering reerences to additional
resources on those topics.
Key Points
Proessionals need a systematic methodology and set o tools
they can use to make and justiy their decisions because
complexity and uncertainty make it incredibly dicult to
make important decisions inormally in a deensible manner.
Multiobjective decision analysis is a methodology and
collection o systematic procedures or evaluating decisions
in the ace o multiple, conficting objectives and uncertainty.
Te methodology involves breaking down a complex decision
7/28/2019 Brownley Chap One
27/32
INTrODUCTION TO MULTIOBJECTIVE DECISION ANALYSIS 17
problem into a set o smaller, and hopeully, more manageable
problems. Ater the decision maker has assessed each o these
smaller problems separately, multiobjective decision analysis
provides a ormal mechanism the decision maker can use
to combine all o the inormation to identiy the preerred
alternative.
Multiobjective decision analysis promotes clear thinking,
leads to comprehension and insights that wouldnt be
apparent rom an intuitive, gut reaction decision-making
process, creates an audit trail that can be used to support
and deend the rationale used to make a decision, acilitates
communication and understanding among multiple
stakeholders, and increases the likelihood that a group will be
committed to the preerred course o action by promoting a
shared understanding o the decision situation.
7/28/2019 Brownley Chap One
28/32
7/28/2019 Brownley Chap One
29/32
Index
AAlternatives, 70, 129
certainty equivalents and, 102108determining overall values or,
6371developing good, 3844ully eatured product, developing, 70
good, 4hal-eatured product, developing, 70identiying, with creative thinking,
3941strategy generation tables, 4041value-ocused thinking, 3940
innite number o, 133136low-eatured product, developing, 70reducing number o, 4142status quo, maintaining, 70
techniques or identiying, 39under uncertainty, developing,
4244
CCertainty equivalents
and alternatives, 102108calculating, 102118, 121130relationship between expected
utility and, 102and spreadsheet, 102, 104,
124123Certainty equivalent value, 121123,
130, 138139Challenges, o decision making
attitude toward risk, 56good alternatives, 4intangibles, 4interdisciplinary substance, 5
long time horizons, 4multiple decision makers, 6multiple objectives, 34sequential nature o decisions, 45uncertainty and risk, 5value trade-os, 6
Common business decisions, 84, 140Company image (CI), 5356
and increment, 5356Constructed evaluation measures,
3135developing, 3235
picture scale, 34
proxy evaluation measures, 3435scale with dened levels, 3233
weighted scale, 33or perception o company image,
3132Continuous probability, 8592Cumulative distribution unction
(CDF), 8889
D
Decision raming, 1920Decision making
challenges o, 36Discrete Probability, 7885
EEliciting probability, 7882Evaluation measures, 9, 3038,
6769, 123, 128. See alsoUtility unction
constructed, 3135cost, second, 67customer service, third, 67decreasing preerence, 4951desirable properties o, 3538
measurable, 3536operational, 3637understandable, 3738natural, 3031
preerence weights, 5153proxy, 3435single-dimensional value
unctions, 4651importance o swinging, 62, 63increasing preerence, 4951, 58
7/28/2019 Brownley Chap One
30/32
158 INDEx
natural, 3031prot, rst, 67single-dimensional value unctions, 62
time-to-manuacture, nal, 6869Expected utility, 85, 9697, 99,102103, 116, 117, 119
Expected value, 115116and alternative, 104105alternative way to calculate, 91calculating, 8385and certainty equivalent, 122o manuacturing costs, 92probability distributions and,
7393and uncertainties,137138Exponential constant, 5761Exponential value unctions, 5051
determining, 5662Exponential utility unctions, 97103Extended Pearsonukey
Approximation, 85, 9092, 104
F
Framing, decision, 1920
GGoals, 9Good alternatives, 3844
HHedging, 43
I
Identiying objectives, 2125considering problems and
shortcomings, 22determining generic objectives,
2425determining strategic objectives, 25developing wish list, 21identiying alternatives, 21identiying goals, constraints, and
guidelines, 23predicting consequences, 2223
Incrementand company image (CI), 5356
Insuring, 44Intangibles, 4
Interdisciplinary substance, 5Interdependent uncertainties,
136139
LLong time horizons, 4
MManuacturing costs, 8692Multiple objectives, 34
and uncertainity, 111116Multiobjective decision analysis
(MODA)
dened, 67extension to, 133143importance o, 1415process o, 912
Multiple decision makers, 6
NNatural evaluation measures, 3031Net prot, 103110, 136138, 141Net present value (NPV), 141
Normalized exponential constants(nec), 59, 60
Normalized mid-value (nmv), 59, 60
OObjectives, 6, 9, 27, 128
brainstorming, 2025structuring, 2530
separating undamental rommeans objectives, 2627
stopping the structuring process,2930
structuring undamentalobjectives hierarchies, 2729
techniques or identiying, 2125considering problems and
shortcomings, 22determining generic objectives,
2425determining strategic objectives, 25developing wish list, 21identiying alternatives, 21identiying goals, constraints,
and guidelines, 23predicting consequences, 2223
7/28/2019 Brownley Chap One
31/32
INDEx 159
Overall certainty equivalent values, 130Overall Weighted Values, 7071
PPearsonukey Approximation,extended, 85, 9092, 104
Picture scale, 34Piecewise linear single-dimensional
value unctions, 4950determining, 5356
Portolio Decision Analysis, 135Power-additive utility unction,
116118
Preerence analysis, 78Preerence weights, 5153, 69,128129
determining, 6163Present value (PV) ormula, 140Probabilistic independence, 121Probability, 129
anchoring and adjustment, 7778assuming certainty, 76availability, 7677
continuous, determining, 8592determining discrete, 7885elicit, preparing to, 7980eliciting, 8082expected value, 8285misunderstanding, 75relying on heuristics, 76representativeness, 77verbal descriptions, using, 7576veriying, 8285
Product manager, 8690, 96Property o expected values, 122, 138Proxy evaluation measures, 3435
RRisk
averse, 9599, 101, 105, 108, 112,114, 116, 118, 122, 138
aversion, 9597, 100, 107108,111, 134135, 138
attitude toward, 56neutral, 9599, 105, 116, 118, 122seeking, 9698, 118tolerence, 98102uncertainty and, 5
Risk aversion, 9597, 100, 107108,111, 134135, 142
Risk-averse decision makers,
9698, 118Risk-neutral decision makers, 97, 98,105, 114
Risk-seeking decision makers,9698, 114
Risk sharing, 43Risk tolerence, 98102. See also
Utility unctiondecision makers, 98101determining multiobjective,
118121determining, 98102Risky alternative, 95, 9799,
103104, 108, 114118
SScale, with dened levels, 3233Scores. SeeLevelsScreening criteria,
4142
Sequencing, 43Sequential nature o decisions, 45Single-dimensional value unctions,
4651, 6162, 70, 111,117, 128
exponential, 5051piecewise linear, 4950
Spreadsheet, 123123and certainty equivalent,
122124
layoutmultiobjective decisionanalysis (no uncertainty), 64,114, 120
Strategic Decision Making, 136Strategy generation table,
4041Structuring objectives, 2530
separating undamental rom meansobjectives, 2627
stopping the structuring process,2930
structuring undamental objectiveshierarchies, 2729
Swinging evaluation measure,62, 63
7/28/2019 Brownley Chap One
32/32
160 INDEx
UUncertainty, 4244
hedging, 43
insuring, 44and risk, 5risk sharing, 43sequencing, 43
Uncertainty analysis, 7Utility
expected, 85, 9697, 99, 102103,116, 117, 119
unction, 9798Utility unction
determining, 9798exponential, 97103exponential multiobjective, 117118power-additive, 116118
VValue-ocused thinking,
3940
Value unctions, 6769,123, 128Values, 8
brainstorming,2025
Value trade-os, 6Veriying probability, 8285
WWeak Law o Large Numbers, 83
Weighted scale, 33Weighted single-dimensional certaintyequivalents, 129130
Weighted single-dimensional values, 70