10. Putting the Structure in Decision Making

8/7/2019 10. Putting the Structure in Decision Making

1/768 Te n C l a s s i c s f r o m S u p p l y C h a i n M a n a g e m e n t R e v i e w www.scmr.com

INITIATIVE BALANCE PROCESS PRACTICES ENGAGEMENT

By William B. Lee and Errol Wirasinghe

William B. Lee (wbleephd@

ymail.com) is a supply chain

educator and consultant and

co-author of the Ld

Effc S C

Trsforo . Errol

Wirasinghe ([email protected]) is a consultant on

decision-making and author of

T Ar of Mk Dcsos.

Many supply chain man-agers make the criticalerror o equating mak-ing decisions with solv-ing problems. Decisionmaking is not problem

solving. Decision making is specifc tothe person. Problem solving is specifcto the problem. Our work suggeststhat problem solvers are not necessar-ily the best decision-makers, and per-haps vice versa.

We have all heard jokes like, to asurgeon, cutting can fx anything orto a chiropractor, manipulation canfx anything or to a psychologist,everythings mental. These jokes are,o course, not air to those proession-

als, particularly the good ones. Butdont we have the same sort o phe-nomenon with our supply chain deci-sion making? Some will advocate thatlean can fx everything or that every-one needs to do Six Sigma. Again, wedont want to criticize the advocateso those techniques, just as we dontwant to disparage surgeons or chiro-practors or psychologists. Its just thatwe need to take a longer, more analyti-cal look at our supply chain decisions

and not simply jump in with the latestbuzz word. We believe one o the rea-sons companies have so much troublewith their supply chains is a lack o astructured decision-making process.

Many surveys have shown thatan overwhelming majority o pro-essionals indicate that they rely oncommon-sense and gut-eel inter-pretation o data and subsequentdecision making. They are inconsis-tent in how they approach decisions,

and yet they requently institute rulesin an attempt to be more consistent.They also systematically distort cer-tain pieces or types o inormation.The result o these decision-makingapproaches typically is that the samesituation presented repetitively yieldsdierent decisions. Furthermore,these decisions rarely are anywhereclose to being optimal.

This is pretty sad. Clearly, we arenot going to do much to fx this situ-ation with this one article, but maybewe can help. Many good books havebeen written on decision making; Twoo the best are David C. SkinnersIntroduction to Decision Analysis1 andErrol Wirasinghes The Art of Making

Decisions.2

Our objective is instead tohighlight the role o decision-makingprocesses, analysis, and models insupply chains.

Specifcally, we oer seven stepsto a more structured approach to sup-ply chain decision-making:

1. Frame and describe the situa-tion about which a decision is to bemade.

2. Defne the objective(s) o thedecision and the criteria that defne

the objectives.3. Extract obligatory criteria.4. Creatively identiy decision

options that meet all obligatory criteria.5. Gather inormation on decision

alternatives, and develop the judg-ment table.

6. Assign weights to the obligatorycriteria.

7. Rank alternatives.To illustrate the use o such mod-

els, we like the old quote: All models

Putting the Structure in


2/7www.scmr.com Te n C l a s s i c s f r o m S u p p l y C h a i n M a n a g e m e n t R e v i e w 69

DECISION MAkINGSupply chain decision making

is fraught with difficulties,

largely because decision

making frequently is regarded

purely as a mental process

in which there is an illusion

of knowledge. Supply chain

managers rely too much in

guts and intuition, the authors

contend, and not enough on

structure and information.

They offer seven steps to

improving that problem in

your supply chain.



Decision Making

are wrong, some models are useul.What are the implications o this quoteespecially

in this context o supply chain decision making? Thereare many, and we will explore some o them in the dis-cussion that ollows.

Complexity Drives Decision-MakingReaders o SCMR likely will agree that most supplychains are complex, particularly as they extend past sup-pliers to suppliers suppliers, and past customers to cus-tomers customers. Most companies also have multiplesupply chains, which add additional complexity. In manyindustries, supply chain management perhaps is the sin-gle most important driver o the companys success and

the single most difcult area o decision making.Complexity drives uncertainty and risky environ-

ments or the decision maker. Uncertainty o demandand supply, uncertainty o the global business environ-ment and many other uncertainties are prevalent. Risko product ailures, o customers or suppliers going out

o business, o fnancial meltdownsand many otherrisks also are widespread. Decision makers need to dealwith all o these issues.

Importantly, please also recognize that relying onfnancial data alone can lead us to sub-optimal deci-sions. The message here is that we need a robust tech-nique or handling multi-criteria decision making, andor validating decisions. A robust technique is one thatis straightorward and that is difcult to misuse to getthe wrong answer. The guts and intuition approach todecision making, by contrast, can easily lead a person oran organization astray.

The theory and practice o decision making (ordecision theory) classifes decisions as ollows: deci-sion making under certainty (DMUC), decision makingunder uncertainty (DMUU), and decision making underrisk (DMUR). Decision theory contains well-knownapproaches to each.

When we know or certain the outcome associatedwith each decision alternative, its called decision mak-ing under certainty. Some, but not many, supply chaindecisions all into this categorymostly short-term deci-sions. O course, how certain is certain? No decision

outcome is really certain, as we say, down to the 47thdecimal place! Scheduling decisions, or example, some-times might be considered relatively certain. When youplace an order with Dell, they promise delivery o a com-puter in a certain number o days ater your order, and

its there right on schedule with very little variation romone customer to another. Dell has made their masteryo the supply chain a real competitive advantage in themarketplace. FedEx is another example o a companythat can promise delivery with some degree o certainty.

When a decision alternative can result in more thanone possible outcome, along with an uncertain probabil-ity o occurrence, then we call it decision making underuncertainty. However, when we know the probability o

each possible outcome, wecall it decision making underrisk. But, you argue, isnt ourknowledge o the probabilitiesreally just a guess and reallya continuum? In most cases,that is true.

Consider a simple example.Say, we have a supply chain decision about a possible newsupplier with two possible outcomes: successul choiceor unsuccessul choice. I we really do not know the trueprobabilities or either, we may choose to assign a 50/50likelihood. But this really is a perception-driven interpre-tation. Maybe its 60/40 or 40/60. Or, maybe its a nor-

mal distribution with a mean o 50 percent and a standarddeviation o 10 percent? We have just turned DMUU intoDUMR or, perhaps, into a continuum DMUR.

Case Study: Decisions Under RiskIn supply chain decision making, DMUR likely is themost prevalent condition. Lets look at a real case o alarge lubricants company. We have disguised the caseor use in our MBA and executive education teaching.Further, or this article we obviously have simplifedthe details o the case while nevertheless retaining theessence o the decision problem.

One o the companys specialty lubricants plantsis experiencing a substantial increase in business.Feedstock supply is not an issue. The plant is an essen-tial element o the companys supply chain as well as oits customers. The company is considering three cours-es o action:

A. Subcontract or additional capacityB. Construct new plantC. Do nothing (no change)Demand may be low, medium, or high with probabili-

ties estimated to be 10 percent, 50 percent, and 40 per-

Many surveys have shown that an

overwhelming majority of professionalsr o coo-ss d -f rro of

d d sbsq dcso k.



cent respectively.The company frst is interested in the fnancial

impact o the decision. It can estimate the net presentvalue o profts rom the three alternatives (A, B, and C)under the diering levels o demand. This is the Payo

Matrix shown in Exhibit 1. Several criteria can be usedto make the decision regarding which o the three alter-

natives to choose.The Maximin Criterion. Maximin is a criterion o

extreme pessimism. It attempts to maximize the mini-mum possible gain. This is done by determining thesmallest possible gain or each alternative i the worst

possible event occurs. The alternative is selected thathas the best outcome under the poorest conditionsthemaximum o the minimums. Under this criterion, thebest choice is C with a return o $20 million. The mini-mum or A is $10 million, and or B it is minus-$120

million.The Maximax Criterion. This is the criterion o

extreme optimism. Its objective is to selectthe decision alternative that will provide themaximum possible return, regardless o theassociated probabilities. Under this crite-rion, the best choice is B with a maximumreturn o $200 millionthe maximum othe maximums. The maximum or A is $90million, and or C it is $60 million.

The Minimax Regret Criterion.

Regret is synonymous with the opportunity cost onot having made the best decision or a given outcome.It ollows that the decision maker would like to makea decision that minimizes regret. This is the RegretMatrix shown in Exhibit 2. Say, high demand occurs,

and the payo rom decision A would have been $90million, the payo rom decision B would have been

$200 million, and the payo rom decisionC would have been $60 million. The regretrom A would have been $110 million ($200million minus $90 million). This is the min-imax regret (the minimum o the maximumregrets) choice.

The Expected Value Criterion. Thiscriterion weights the outcomes by theirprobability o occurring. For example, usingthe payos and probabilities cited above,the expected value or decision A is $62 mil-lion ($90 million x 0.4 + $50 million x 0.5 +

$10 million x 0.1). This is shown in the Expected ValueMatrix in Exhibit 3. The maximum expected value is$80.5 million, decision B.

We oten ask questions such as the ollowing. Giventhe inormation as presented, what would you do with

the lubes plant decision and why? Especiallysince all three alternatives were chosenunder dierent decision criteria. And whatabout the model? In what ways is the model

an abstraction rom reality, and what dier-ence does it make? What additional inor-mation would you want i this were yourdecision?

There are many ways we can make thismodel more complex and thus more realis-tic. One example is sensitivitywe could

test the sensitivity o the decision to the probability esti-mates, to the payo estimates, and even to the decisionalternatives themselves. Another example o additionalcomplexity is with Alternative B: What size o a newplant should we constructsmall, medium, large, or

very large? Any number o additional complexities couldbe added to the decision.

EXHIBIT 1

Payoff Matrix

Probabilities of Demand

Decision Alternatives

A.Subcontract of Additional Capacity

B. Construct New Plant

C. Do Nothing (No Change)

Net Percent Value of Profits from the Decision

Demand Low Demand

50%

$50m

$25m

$40m

Medium Demand High Demand

10%

$10m

-$120m

Maximin $20m

40%

$90m

Maximax $200m

$60m

EXHIBIT 2

Regret Matrix

A.Subcontract of Additional Capacity



Decision Alternatives Net Present Value of Regret from Decisions

Demand Low Demand

$0m

$25m

$10m

Medium Demand High Demand

$10m

$140m

$20m

$0m

$140m

Minimax Regret$110m

EXHIBIT 3

Expected Value Matrix

A. Subcontract of Additional Capacity



$62m

Maximum Expected Value = $80.5m

$46m

Expected Value



Decision Making

So, what has this revealed? We have used this sim-plied (but realistic) example o a supply chain decisionwith three choices. By applying some very commonsensenancial decision rules, all three choices have beenshown to be reasonable under dierent decision criteria.

Whats a supply chain manager to do? To try to putsome light on this subject, lets look into some key sup-ply chain decisions that need to be made and a struc-tured approach or making them.

Key Supply Chain DecisionsIt is useul to consider ways to classiy supply chaindecisions so that we can apply decision-making process-es appropriately. One o the most eective classicationswas presented by David Simchi-Levi and his co-authors

in Designing and Managing the Supply Chain.3 The ol-lowing is adapted rom that work:

1. Strategic decisions have a long-lasting (one to tenyears typically) eect on the organization. These includetarget customers and their characteristics, product and

service selection and design, distribution network con-gurations such as numbers and locations o acilities,structure and processes o the supply chain, supplierrelationships, and so orth. Since these decisions areboth long lasting and typically o great consequence,more time, eort, and ormality usually are taken in thedecision process. The lubricants plant above is an exam-ple o a strategic decision.

2. Tactical decisions include decisions that typi-cally can be updated anywhere between quarterly and

yearly. These likely include purchasing and supply con-tracts, production decisions, inventory policies, sales and

operations planning, and so orth. These decisions areimportant but reversible in the intermediate time period.Thus, they would deserve some intermediate level o or-mality in the decision making process.

3. Operational decisions reer to day-to-day decisionssuch as scheduling, quotations, transportation, and soorth. These decisions typically last only a short periodo time and are o little consequence i wrong. Thus,people likely will spend relatively little time or resourcesin making them individually. However, management re-quently develops decision rules that apply to an entire

class (say, quotations) o these decisions. For example,quotation decision rules may be something like take theproduct standard cost and add a 40 percent markup orthe price.

Structured decision processes, the ocus o our discus-

sion, usually deal with strategic and tactical decisions.

The Structured ApproachMany o us are accustomed to making decisions on thebasis o quantitative data. The rate o return on a proj-ect, or the expected demand or a product, are easilyexplained in terms o numbers. In the above example,we developed a payo matrix and applied nancial cri-teria to the decision. But this is just one criterion. In thereal world, there are many actors at play; these are what

we reer to as criteria.Inherent in many supply

chain decisions are actorssuch as company strategy,competition, customers, sup-pliers, bureaucracy, languagebarriers, governmental issues,and so orthall o which are

qualitative in nature. Thus the subjective human being(the decision maker) adds inevitable biases when he orshe includes certain criteria, and excludes others.

To most people, a good decision is one that producesthe desired outcome. Unortunately this is a very limited

denition. David Skinner makes a good point about out-comes and the relative desirability o possible outcomes.1

Outcomes are what can happen. As an example, i youwere going to have heart surgery, the outcomes could be:

complete recovery, no side effects, partial recovery, some side effects, or death.Clearly, we all would have our preerences about the

outcome! So part o having a robust decision making pro-cess is a strong linkage o the decision(s)say, not onlywhether to have heart surgery, but whom to choose asthe surgeon, where to have the surgery, and the surgeons

decisions beore, during, and ater surgeryto the pos-sible outcomes. We all know that dierent surgeons anddierent hospitals have dierent outcome probabilities.

We want to weigh these in our decision process.The quality o the decision will depend on a num-

ber o things: the decision alternatives, decision criteria,available data and inormation, context and domain othe decision, analysis techniques used, the expertise othe decision maker, and so orth. However, the outcomeo the decision will, additionally, depend on appropriatetiming, adequate resources, commitment to execution,

The guts and intuition approach todecision making is easily confused d cs d rso or orzo sr.



and changing circumstances, among other actors.As an alternative to gut-eel and pure common-sense

techniques, we oer a seven-step process to making solid,structured decisions. We will discuss the steps in terms othe previous example. This approach is invaluable when

we deal with qualitative inormation and criteria.

1. Frame and describe the situation about which adecision is to be made.

We paraphrase a amous anonymous quote: A decisionwell ramed is hal made. The rst step is to dene thedecision in light o the companys strategy, along withthe decisions boundaries, interaces, and infuences.Strategic thinking drives decision making by:

Setting strategic direction for the business. Establishing the objectives to deliver results. Creating measures to track progress toward objectives. Specifying behaviors that are required to imple-

ment the strategy.Each o these is important as we rame our supply

chain decisions.

2. Defne the objective(s) o the decision and thecriteria that defne the objectives.Objectives must be stated in unambiguous terms.Additionally, they should be time-bound, and progressmust be measurable. In the earlier lubes plant example,we may have an objective stated as: Determine the best

option to meet the expected demand growth or the nextve years. Wherever possible it is essential to have astatement (such as this one) that would allow us to mea-sure progress in terms o achieving this objective. Thus,i the expected progress is not being achieved, we cantake corrective measures in a timely manner.

Criteria explicitly dene the objectives o the deci-sion. As we structured the problem previously, the pay-o matrix is just one criterionnancial. Most supplychain decisions (and most others or that matter) aremultiple-criteria decisionsmeaning that there are mul-tiple objectives which the decision needs to achieve.

The nancial payo matrix indicated no clear preerredoption, but we would have needed to consider other cri-teria anyway. These could be, or example, the ollowingour additional criteria.

Strategic ft. How well do the candidate decisionalternatives t with the companys strategic direction?Does the company preer outsourcing or not outsourc-ing? Does the company preer to own excess capacity orto be tight on capacity?

Management capability. Does the company havethe necessary management capability to implement the

options? I outsourcing is unamiliar to the managementteam, this may be a dicult alternative to implement.

Risk. Do the options present unacceptable risk tothe companydened in whatever way the companydeems useul? For example, part o a robust decision

process likely should be the possible environmental risk. Customers needs. How well do the alternatives t

the needs o the customers?

3. Extract obligatory criteria.With criteria, more is not merrier. As we add more cri-teria, the signicance o the existing criteria is diluted.More than about 10 criterion require some sort o priori-tization to extract the obligatory criteria simply becausewe may not be able to deal eectively with that manychoices. The remaining criteria may be desirable or nice-to-have but not necessary. I there is a tie between thetop two candidates, then we may revert to these desir-able criteria to make a nal decision.

The Analytic Hierarchy Process4 (AHP) is a struc-tured pair-wise comparison technique or dealing withcomplex decisions. Rather than prescribing a correctdecision, the AHP helps the decision makers nd theone that best suits their needs and their understandingo the problem. Several rms supply computer sotwareto assist in using the process. AHP is beyond the scopeo this article, but we wanted to introduce the idea andsuggest that our readers go beyond the present discus-

sion in this important area.

4. Creatively identiy decision options that meet allobligatory criteria.

We have already identied the decision alternatives:sub-contract, build a new plant, or do nothing. Althoughwe noted that each o these could be expanded intonumerous sub-options such as build a small plant, or amedium-size one, or a large one, or on and on. We thuscould build a hierarchy o decision options. Such a hier-archy can be analyzed using AHP. As to the creativeapproach, Errol Wirasinghe, in his book, says it well:

The role o creativity is that o generating andidentiying options with which to solve a problem. Creativity consists largely o rearranging what we knowin order to nd out what we do not know. A pile orocks ceases to be a rock pile when someone contem-plates it as a cathedral.2

5. Gather inormation on decision alternatives, anddevelop the judgment table.This is where we note the pros and cons pertaining tothe criteria, or each o our candidates. The quality o



Decision Making

our nal decision will depend on how much eort wededicate to the task. Do not skip this step; withoutadequate inormation about the candidates, we cannotmake a reliable decision.

6. Assign weights to the obligatory criteria.All criteria do not have the same signicance. We mayrun an AHP evaluation to assign criteria weights. Thereis a danger in using common sense because you maybecome a victim o hidden traps in decision makingas outlined in a 2006 Harvard Business Review articleon the subject that we highly recommend.5 The article

mentions several types o traps, but postulates thatwhat makes all o them so dangerous is their invisibility.Because they are hardwired into our thinking process,and thus our common sense, we ail to recognize themeven as we all right into them.

7. Rank alternatives.Ranking alternatives typically begins by discarding those

that appear obviously inerior. In the lubes example, weintentionally did not include any inerior alternatives.Then, we run a pair-wise analysis on the decision alter-natives using our judgment table. The decision makersystematically evaluates the various alternatives by com-paring them to one another two at a time. That is, wecompare subcontracting with building, subcontractingwith doing nothing, and building with doing nothing. Inmaking the comparisons, the decision makers can useconcrete data, or they can use their judgments about theoptions relative meaning and importance.

To validate the decision, we described the deci-

sion situation, decided which criteria are relevant;andassigned weights to these criteria to refect their signi-cance. At this point, we need to ask, Did our humaneelings overly bias the decision in avor o some alter-native? These may be some o the decision traps men-tioned above.

Another very relevant validating question is: Will thewinning candidate be valid i we were to remove anyparticular criterion? We then remove one criterion at a

time and see its impact on the decision. Clearly, i weremove any single criterion, then we must re-evaluatethe weightings o the remaining ones.

Next Steps

The reader should be aware o one caveatwe cannotpossibly cover the subject o supply chain decision mak-ing in one short article such as this. Countless articlesand books have been written on the subject o decisionmaking alone, without adding the diculties o supplychains. We thus have chosen to take a very narrow sliceand not try to deal with such a complex subject in any-

thing near a complete manner.Our objective here has been to encour-

age the readers to consider how to structuredecisions so as to improve their decision-making process. We have examined thedynamics o the decision processthehow and why o decision makingandhave explained some elements o the struc-

turing process. We also have shown how one approach(the nancial one) easily can provide conficting resultsand why most decision situations are multi-dimensional.

The example o the lubes plant shows a commonstructure o the nancial role in the decision processwe can say (tongue rmly placed in cheek), you tell methe answer you want, and Ill tell you what criterion touse! We could expand this little statement to say that

we also can manipulate the payo matrix to give you theanswer you want. But, thats not air is it?So, our message is to be careul about how you dene

the decision under consideration. We eel making use othe seven steps described here is the best way o helpingto ensure a robust decision-making process.jjj

End Notes

1 Skinner, David C., Introduction to Decision Analysis: APractitioners Guide to Improving Decision Quality, SecondEdition, Probabilistic Publishing, 1999.

2 Wirasinghe, Errol, The Art of Making Decisions: Expanding

Common Sense & Experience, Shanmar Publishing, 2003.3 Simchi-Levi, David, Philip Kaminsky, and Edith Simchi-

Levi, Designing and Managing the Supply Chain: Concepts,Strategies, and Case Studies. McGraw-Hill Irwin, 2008.

4 See Wikipedia article on Analytic Hierarchy Process oran introductory explanation and other reerences. (AccessedFebruary 2010).

5 Hammond, John S., Ralph L. Keeney, and Howard Raia,The Hidden Traps in Decision Making, Harvard BusinessReview, January 2006.

We need to take a more analyticallook or s c dcsos, o j s bzz ord.

10. Putting the Structure in Decision Making

Documents

Transcript of 10. Putting the Structure in Decision Making