(DSSORA)isaninteractivemathematicalprogrammingsystemforoptimalresourceallocationdevelopedtosupportdecisionsofinvestment



Similar documents
CHANCELLOR'S COMMUNICATION

b) since the remainder is 0 I need to factor the numerator. Synthetic division tells me this is true

Paper 3A: Cost Accounting Chapter 9 CA. Dharmendra Gupta

Xerox WorkCentre Xerox ColorQube

CFABAI132 Inform and facilitate organisational decision-making

Adform Brand Identity Guidelines

Executive Director Performance Evaluation and Development Plan Self-Evaluation Summary

Michigan Public School Accounting Manual presented by Glenda Rader Grand Ledge Public Schools September 23, 2015

Council of the European Union Brussels, 8 April 2016 (OR. en) Permanent Representatives Committee/Mixed Committee

Are you interested in working in ministry for the Catholic Church in our local community?


How to sign up to Ultimate Black Belt Guitar

Single Stage Light Business Case Template

PLAN YOUR BUSINESS. Instructions page. Business Plan Template

Program Description. [Program, e.g. B.S. in Nursing] To Be Offered by [Campus] at [Location]

OPERATIONAL REQUIREMENTS DOCUMENT

Defense Human Resources Activity Handbook on Teleworking for Supervisors and Employees

Target Strategies Resources Personnel timescale Impact through Monitoring Accurate groupings established to provide

Getting Started with Business Intelligence

Financial Services MOU PROCEDURES MATRIX

The Performance Management Process How to establish goals, objectives and KPI s

Ciphermail Gateway EJBCA integration guide

Using Microsoft s CA Server with SonicWALL Devices

10 3,5. 0,5 x 0,5 0,12 0,16 0,20 0,25 0,32 0,40 0,50 0,70 1,00

Figure2:Themixtureoffactoranalysisgenerativemodel. j;j z

Asset/Liability Management

G e r m a n T O & E TABLE OF CONTENTS. PANZER German TO&E

Project Management Plan (PMP)

Sample Resume BETTY CASEWORKER

Kansas State University Milling Specialist and Expert Credentials Program. Launching January 2014

Logistics Officer Association Guidance (G 3) National Scholarship Program

Lake at IPEN location

Product Composer System

Bogdan Vesovic Siemens Smart Grid Solutions, Minneapolis, USA

is identically equal to x 2 +3x +2

Best Practices in Performance Measurement Developing Performance Measures

Early Detection Research Network: Validation Infrastructure. Jo Ann Rinaudo, Ph.D. Division of Cancer Prevention


SETTING UP INTERNATIONAL CLINICAL TRIALS IN EUROPE: THE EORTC EXPERIENCE

Policy. VBA Enterprise Risk Management. Governance Unit

PROCUREMENT MANAGEMENT PLAN <PROJECT NAME>

Process Improvement Plan

Strategic Technology Plan

Design & Analysis of Ecological Data. Landscape of Statistical Methods...

Transportation Network Companies: Insurance Industry Advocacy Toolkit

Defense Logistics Agency INSTRUCTION. SUBJECT: Department of Defense (DoD) Financial Management (FM) Certification Program

R Power Supply (USA, CA, MX, JP, TW)

Zscaler Support Best Practices Guide. Version April 1, 2016

Dr. Tarek A. Tutunji Philadelphia University, Jordan. Engineering Skills, Philadelphia University

Contents. Financial Analysis Report

Connection cable for amplifier Type 4701AxA and torque sensor Type 4501A... (0143SD) version Q/R/H

Standard Practice for Evaluating the Cache Packaging Weight and Volume of Robots for Urban Search and Rescue 1

USERS MANUAL AREX DIGICOMM

Integration ALGEBRAIC FRACTIONS. Graham S McDonald and Silvia C Dalla

is identically equal to x 2 +3x +2

Table of Contents. Volume No. 2 - Classification & Coding Structure TOPIC NO Function No CARS TOPIC CHART OF ACCOUNTS.

Your gas and electricity bill actual readings

Quick Installation and Setup for WinPower Software

Computer Engineering (COE) PROGRAM LEARNING OUTCOMES (10/19/09) Students completing the Computer Engineering program should be able to demonstrate:

FOIL FACTORING. Factoring is merely undoing the FOIL method. Let s look at an example: Take the polynomial x²+4x+4.

Is there a Doctor in the House? A Business Analyst in an Agile Business Dot Tudor TCC Agile Business Conference 2015

Data Visualization and Feature Selection: New Algorithms for Nongaussian Data

Quick Installation and Setup for WinPower Software

Institute of Certified Bookkeepers. Business Plan Template

Substitutes for Mobile Voice Roaming Services ( MVRS )

How to Craft a World-Class Work Breakdown Structure

PA 590 Application Upper Iowa University MPA Program

Utility Maximization

CERTIFIED TRANSLATION

System Behavior Analysis by Machine Learning

Accounting Notes. Types (classifications) of Assets:

BEHAVIOR BASED CREDIT CARD FRAUD DETECTION USING SUPPORT VECTOR MACHINES

Hybrid Cloud Architecture: How to Streamline Hybrid Cloud Migration

WHY CLOUD COMPUTING MAKES SENSE FOR NONPROFITS

ROYAL MALAYSIAN CUSTOMS

GAMING MANAGEMENT SEMINAR SERIES. Prices of Casino Games. Jason Zhicheng Gao 高 志 成

2011 Latin American Network Security Markets. N July 2011

ESnet SSL CA service Certificate Policy And Certification Practice Statement Version 1.0

Transcription:

Jean-MichelThizy1,DanielE.Lane1,SavvasPissarides2&SurendraRawat1;3? 1FacultyofAdministration,UniversityofOttawa,OttawaONK1N6N5,Canada, InteractiveMultipleCriteriaOptimizationfor CapitalBudgetinginaCanadian 3StentorResearchCentreInc.,160ElginSt.,Ottawa,Ontario,K1G3J4Canada. 2BellCanada,160ElginSt.,Ottawa,Ontario,K1G3J4Canada, TelecommunicationsCompany (DSSORA)isaninteractivemathematicalprogrammingsystemforoptimalresourceallocationdevelopedtosupportdecisionsofinvestment incapitalintensivetelecommunicationsprojects.thesystemstrivesto Abstract.DecisionSupportSystemforOptimalResourceAllocation maximizecorporategoalswhilerespectingnancialconstraintssuchas researchconductedwhileattheuniversityofottawa, cess(ahp)isusedforquanticationofqualitativemanagerialjudgment pendenciesorsynergiesamongprojects.theanalyticalhierarchypro- inregardtotherelativevalueofprojectsthroughatwostageprocess.an theavailabilityofcapitalfunds,institutionalrequirementsandvariedde- managementandtheimplementationofprojectportfolioswithsignicantlyconsistentresults.theexibility,userfriendlinessandquicktimecientsofwhicharedeterminedbytheahp.then,amodiedsimplex ofdssoraandcometorelyonitseciency-seekingcapabilities.dss ORAhasbeentestedbyseveralgroupsofmanagersresponsibleforthe userscanbuildgradualcondenceintheconstraintcheckingmechanism algorithmproposessomeratesoffundingsubstitution.userschoosethe amountofsubstitutionorcanoverridethesubstitutionsproposed.thus, initialresourceallocationisfoundbyalinearprogram,theobjectiveco- responseofdssoramakeitaneectivenegotiationtoolinagroupsetting. 1TheCapitalBudgetingProcess Whilecapitalbudgetingconstitutesaclassicalresourceallocationproblem,its exercisebycorporateplannersisregulatedbyinstitutionalprocedurestoensure Keywords.Interactivemultiplecriteriadecisionmaking,optimization, etc.forexample,determiningthetotalamountofcapitaltobeallocated,what accountability,respectorganizationalstructures,safeguardminorityinterests, capitalbudgeting,telecommunications,mathematicalprogramming activitiesshouldbeconsideredforfunding,whichcriteriaarecompatiblewiththe?researchsupportedbybellcanada,nsercoperatinggrantsogp0042197and OGP0043693,andAUCCGoingGlobalProgram

muchbeforethedecisionmodelleadingtoacapitalallocationisdened. directionofthermareallstepsthatrequirecarefuljudgmentanddeliberation measureofthecostsandbenetsofeachprojectandtheiruncertainty.systematicmethodsforcapitalbudgetinghavebeenactivelyinvestigatedbypubligram(mop)fortheannualcapitalbudgetingexerciseofacanadiantelecommunicationscompany.ithasbeenusedbyafunctionalgroupofthecompanywhich utilitiessuchastelecommunicationscompanies(salo,1989;hoadley,1993).the decisionsupportsystemdescribedherecentersaroundamultipleobjectivepro- isresponsibleforthemanagementofaportfolioofprograms4.theseprograms Theadequatedesignofananalyticalmodeltomeettheserequirementsis challenging:capitalbudgetingmodelsareoftenhinderedbyapooreconomic constructofmathematicaloptimization. resourcesamongallfunctionalgroupsoftherm.eachprogramcoordinatesa allyusethenarrowertermproject,reservingthetermprogramtothefamiliar catetheseresourcestoindividualprojectsparticipatingintheprogram.hence, inagreementwithmostoftheliteratureoncapitalbudgeting,wewillgener- arepresentedtothecompany'scentralcapitalbudgetingcommitteeallocating coherentsetofprojectsmeetingacommonsetofobjectives.giventheresources awardedtoaparticularprogram,themethodproposedcaninturnhelpallo- oforganizationalandoperationalconstraints,suchasprojectinterdependencies.thecommitteemakesdecisionsthatmaybecontestedbysomefunctional groups'managers.legitimateconcernspromptthecommitteetoreconsiderthe cordingtotherequestsofthefunctionalgroups.thereforetheprojectsarepri- oritized,usingexplicitcriteriathatreectthecompany'smission.currently, judgmentbasedupontheknowledge,inuenceandexperienceofeachmember inthebudgetingcommittee.thedecision-makersmustalsoconsideranumber theevaluationofeachprojectwithrespecttoeachcriterioninvolveshuman LimitedresourcespreventtheCommitteefromfundingalltheprojectsactributionsoftheprojectstothesecriteriaandthenatureofprojectdependencies. Theappealmechanismprovidesevidenceofthecomplexityandtheuncertainty existingintheactualbudgetingdecisionprocess.muchofthedicultycomes fromqualitativeandsubjectiveviewsofthecompany'smissionandtheroleof theprojectstoaccomplishthismission. allocationbymodifyingtherulesusedtodenethecompany'scriteria,theconnativecapitalprojectstoachievethem.itcanbeusedeitherbythecommittee quicklyintheeventthatresourcesrequestedarenotfullyallocated. orbydierentunitmanagerstopreparetheirfundingrequestortorespond evaluationofthemissionandgoalsofthecompanyandthesuitabilityofalter- 4Itispreciselysuchprogramplanningthatleadtotherstuseoflinearprogramming Ourdecisionsupportsystemaimsatformingamoreexplicit,quantitative (Dantzig,1963),originallycalledProgramminginalinearstructure.

Budgeting Wenowfocusonamathematicalformulationofamultipleobjectiveprogram 2AMultipleObjectiveProgrammingSystemforCapital forcapitalbudgeting. 2.1ReviewofPreviousModels gramtosolveacapitalbudgetingproblem: Linearprogrammingwasappliedtocapitalbudgetingfromitsinception;there existsavastnumberofanalysesofcapitalbudgetingbylinearprogramming.we 1984;Crum,1981;Dean,1951;Wilkes,1983). shouldconsultmonographssuchas(bierman,1988;bromwich,1979;clark, brieyreviewlandmarkanalysesformop;forextensivebibliographies,readers Charnes,CooperandMiller(1959)werethersttoformulatealinearpro- maxnxj=1cjxj subjectto aijistheamountofresourceirequiredbyprojectj, where: cjisthenetpresentvalueofprojectj, 0xj1forj=1;...;n nxj=1aijxjbifori=1;...;m xjisthefractionofprojectjaccepted, biisthetotalamountofresourceiavailable, nisthenumberofprojectscompetingintheallocation, misthenumberofresourcesconsideredintheallocation. threebasicsetsofgoals:maximizationofnetpresentvalue,cashowbudgeting conventionalgoalprogrammingtechniques. proposinganinteractivemultiplegoalprogrammingmethodbasedonamutual preferencefunctionortrade-osamongcompetingobjectives,assuperiorto andsuccessiveinterplaybetweenadecision-makerandananalyst.spronkviewed hismethod,whichdoesnotrequireexplicitrepresentationofthedecision-maker's Deckro,Spahr,andHerbert(1985)presentedanon-lineargoalprogramwith Spronk(1981)extendedtheuseofgoalprogrammingincapitalbudgetingby andcontrolofrisk.

designedtosupportcollectivecapitalbudgetingdecisions. DecisionSupportSystemforOptimalResourceAllocation(DSSORA)isaninteractivemultipleobjectiveprogrammingsystemforecientresourceallocation 2.2TheMultipleObjectiveProgramofDSSORA fundingofeveryprojectislimitedbyupperandlowerbounds.ourprecise orsynergiesamongprojects:forexample,someprojectscannotbeimplemented unlessatleastacertainportionofanotherprojectisimplemented.finally,the fortheplanningperiod.constraintscanalsobeusedtorepresentdependencies inequalities,themostimportantofwhichisthecapitalfundsconstraintthat Otherconstraintsincludeoverallshorttermnancialimpactoftheportfolio suchasthemaximumlevelofacceptabledepreciationexpenseforeachproject representsthetotalavailabilityofcapitalthatmaybeallocatedtoallprojects. Theallocationofresourcestotheprojectsismadesubjecttoanumberof formulationis: EcientfMissionsatisfaction(x1;...;xn)g AvailabilityofCapital: subjectto: nxj=1xjc (1) Employeesforimplementation:nXj=1ejxjE DepreciationLimit:nXj=1djxjD (3) (4) (2) tionofthecorporatemissionviaamulti-valuedfunctionfin(1)thatcanbe wherexjmeasurestheleveloffundingofprojectj.dssoraassessessatisfac- Softwareandotherexpenses:nXj=1sjxjS Dependencies&Synergies:nXj=1aijxjbifori=1;...;r(6) Bounds:ljxjujforj=1;...;n(7) (5) ciationallowed.inasimilarway,ejrepresentsthenumberofemployeesneeded representsthedepreciationconstraint,wherethecoecientsdjarethenumber ofdollarsdepreciatedperdollarallocatedanddisthetotalamountofdepre- protection,savingsandstrategicimportance,themselvesimplicitlyknownfunctionsoftheallocation.inequality(2)representsthecapitalconstraint,inwhich Cdenotesthetotalbudgettobeallocatedamongthenprojects.Inequality(3) improvedbyincreasingcorporatecriteriasuchasrevenuegeneration,revenue

perdollarallocatedandeisthetotalnumberofemployees;sjdenotesthe softwareandotherexpensesperdollarallocatedtoprojectjandsisthetotal budgetforsoftwareandotherexpenses.constraints(6)representdependencies andsynergiesbetweenprojects.finally,in(7),ujandljrepresenttheupper andlowerlimitsofallocationtoprojectj.foreaseofpresentation,themodel issummarizedas: wherex=(x1;...;xp)t,aisafullrowrankmpmatrix,bisanm-vectorand Fisamulti-valuedfunction.TheoperatorEseekstheecientregionspecied bytheprogram,althoughtheinteractiveproceduredescribedinparagraph2.4 EfF(x)g Ax=b s.t. x0 (10) (8) isdesignedformoreexibleexploration. (9) Step1:DecisionmakersmustchooseanumberofcriteriaimportanttotheCompany'smissionagree,acompromiseonagivencomparisoncanbeobtainedeitherbydiscussion,voteorbytakingthegeometricmeanofeverymember'scomparison. anoverallscaleofimportance.whenindividualsintheevaluationgroupdis- ThismeasurementresemblesthefamiliarMOPmethodofpointestimate valueofeachofitscoecients,calledpriority. criterion.theanalytichierarchyprocess(saaty,1980)isusedtoassessthe ofthemulti-valuedcriterionfunction:f=(fh(x))h=1;...;h,wherefhisalinear Aninitialallocationwillrstbeobtainedbycalculatinglinearapproximations 2.3TheAnalyticHierarchyProcess Step2:Therelativeimportanceofeachpairofcriteriaismeasuredinordertoobtain Step3:AssessthevalueofeachofthecoecientsofFh,i.e.therelativeimpactof (thesimilaritywillberenedinstep4oftheprocedure.) that:phh=1h=1;thesecanbeusedtoreducethecriterionfunctiontothe linearobjective:maxhxh=1hfh(x1;x2;...;xn) weightedsumsthatdenesnonnegativemultipliershforh=1;...;hsuch Step4:Theoverallpriorityofeachprojectisobtainedastheinnerproductofthe andp4. Considerforexampletheallocationofresourcestofourprojects,P1,P2,P3 vectorsobtainedinstep2andstep3(forthisproject). eachpairofprojectsoneachcriterionh,byaseriesofpairwisecomparisons analogoustothoseofstep2.

Step1:Supposethatthecapitalbudgetingdecisiongroupchosethecriteria:revenue B,CandD).ThecorrespondinghierarchyisdisplayedinFig.1. generation,revenueprotection,savingsandstrategicimportance(labeleda, Mission Revenue Revenue Operational Strategic generation protection savings Importance (A) (B) (C) (D) Step2:Forthesakeofsimplicity,weassumethateverycomparisonwasmadeunanimouslybythedecision-makinggroup.Thefollowingisthecomparisonmatrixforthecriteria(forinstance,thesecondcoecient:5signiesthatthe Fig.1.AHPhierarchyforselectionoftelecommunicationsprojects criterionbisvaluedasmoreimportantthana): Project A1535 Project Project Project P1 B1=511=31 0B@ABCD P2 P3 P4 D1=511=311CAyielding=0B@ C1=3313 A0:558 B0:096 C0:249 D0:0961CA

Step3:Nextarethematricesofcomparisonsoftheprojectsundereachcriterion: P11175 A:P1P2P3P4 P21175 P41=51=5311CA P31=71=711=3 0B@ B:P1P2P3P4 P21=3111=9 P11351=5 C:P1P2P3P4 0B@ P41=31=71=311CA P25157 P311=513 P111=513 0B@ projectsyieldonecolumnofthefollowingmatrix: FollowingthemethodofStep2,foreachcriterion,comparisonsbetween P31=5111=9 P459911CA P11571 D:P1P2P3P4 P21=5131=3 P413711CA P31=71=311=7 0B@ Step4:BymultiplyingthismatrixbythevectorofmultipliersfoundinStep2, P40:1020:0620:6770:3871CA P30:0500:1530:0580:052 P20:4240:6320:0660:121 P10:4240:1530:1990:440 0B@ABCD onegetstheoverallpriorities0b@ constraints(2)-(7)toproduceaninitialallocationofresources: Ṫheprioritiesarethenusedasobjectivecoecientsofalinearprogramwith P10:34 max0:34x1+0:33x2+0:06x3+0:27x4 P20:33 P30:06 P40:271CA overallallocationof$180million,andadependencyconstraint(6): Givenindividualbounds(7)displayedinTable1,thecapitalallocationobtained bylinearprogrammingiscontainedinitslastcolumn. TheconstraintsoftheexampleincludeonlyCapitalAvailability(2)withan 0:5x1+x380:

Table1.Allocationconstraints(therightmostthreecolumnsdisplay$million) ProjectPrioritiesDependencyLowerUpperInitial P2 P10.34 0.33 0.5 LimitLimitAllocation 2050 2535 50 2.4TheInteractiveAllocation P3 P4 0.06 0.27 Dependencyrighthandside:80 Capitalavailable:$180million 10 3565 4555 30 55 Theallocationdelineatedpreviouslymayneedfurtherrenementorsensitivity 45 constraints. leavesagreatleewaytodecision-makerswhileenforcingthebudgetorother analysis.totheseends,thedecisionsystemoersaninteractiveprocedurethat proposed.thereforeuserscanproposetheirownsolutions,buildgradualcondenceintheconstraintcheckingmechanismofthesimplexmethod,andcometo algorithmproposessomeratesoffundingsubstitution.then,usersdecideinteractivelywhatamountofsubstitutionispreferable,andcanoverridetherates notresorttoaninteractiveoptimizationinthespaceofcriterionmultipliers analyticalformofeachcriterioncouldbeelusive.thus,thedecisionsystemdoes asclassicalmopmethodspropose,butallowsuserstoassesstheirpreferences directlyinthespaceofbudgetallocations.ateachstep,amodiedsimplex Infact,itwasfoundthatnotonlythevaluesofthepriorities,buteventhe relyoftheeciency-seekingcapabilitiespatternedafter(georion,1972;zionts, 1976). partitionedasx=(xn;xb),wherexnisthesubvectorofnonbasicvariablesand xbisthesubvectorofbasicvariables.correspondingly,aispartitionedasa= [NjB],whereNconsistsofthenonbasiccolumnsofAandBconsistsofthebasic columns.hence: MultiplyingbothsidesbyB?1yields: Considertheprecedingformulation(8){(10).Letthecurrentsolutionxbe whichisequivalentto:xb=b?1b?b?1nxn: B?1NxN+IxB=B?1b; NxN+BxB=b: ofdecreaseinthebasicvariablexbicausedbyaunitincrementinthenon-basic DenethematrixY=B?1N.Eachofitscomponentsyijdescribestheamount

Step0:Settheiterationcountertat0.Selectaninitialsolution(e.g.fromtheinitial allocation,eachoftheprecedingquantitiesreceivesasuperscriptt. vectors.atthet-thiterationofthefollowingalgorithmforinteractiveresource variablexnj.hence,itscolumnsy1;y2;...;yp?mcanbeusedtodenetrade-o Step2:Eitherselectaproperamountoftrade-oxtNjforsomej2[1;...;p?m], Step1:ForeverynonbasicvariablextNj,j2[1;...;p?m],atrade-ovectorytjis calculatedandpresentedtousers. using:min linearprogram):xob=(bo)?1(bo?noxon): i:ytij>0xtbi andlet: ytijxtnjmax?xtn;max x'b=xtb?ytjxtnj; x'nk=xtnkfork6=j; x'nj=xtnj+xtnj; i:ytij<0xtbi ytik; Step3:If,forsomeindexq2[1;...;m],x'Bq=0,thenlet orterminatewithsolutionxt. xt+1=(e1;...;ek?1;ep?m+q;ek+1;...;ep?m+q?1;ek;ep?m+q+1;...;en)x'; Forexample,attheoptimumofthelinearprogramdisplayedinTable1,part Sett=t+1andgotoStep1. elseletxt+1=x'. and(bt+1)?1=(e1;...;ek?1;k;ek+1;...;en)(bt)?1; k=(?y1k=yqk;?y2k=yqk;...;1=yqk;...;?ym?1=yqk;?ym=yqk) oftherowsofthematrix?ycorrespondingtotheoriginalvariablesx1;x2;x3; andx4isshownbelow:0b@ characterizedbythedirectionoftherstcolumn,inanamountxtn1=5.the allocationtoprojectp1decreasesby5,theallocationstoprojectsp2andp3 Toillustratetheuseoftheprecedingmatrix,supposeuserschooseatrade-o P40:00:00:01:01CA: P30:51:00:00:0 P20:5?1:?1:?1: P1?1:0:00:00:0 (11)

eachincreaseby2:5.theallocationtoprojectp4staysthesame.therefore afterdecreasingthefundingofprojectp3by5,theallocationisgivenby(12). 0B@45 standpoint(itenablesdecisionmakerstoreducetheresourcesallocated)and 57:5 32:5 fromatheoreticalone:userscanavoidthepervasiveassumptionthattheirutilityfunctionmustbepseudoconcave,underwhichlargerallocationsarealways preferredtosmallerones(zionts,1976). ofpurelyreducingresourcesforaprojectmaybeusefulbothfromapractical decreasetheallocationtoproject2.unlikelyatrstbrush,suchadecision Thethirdcolumnofthematrix(11).indicatesthatitisfeasibletosimply 451CA requirements,allowabledepreciationexpense,softwareandotherexpenses,manpowerneedsandrelationshipsbetweenprojects.amongthemodulessupporting directlytheinteractiveresourceallocation: {amoduleforprojectvalueassessmentimplementstheahpmethodologyto availabilitiesandprojectrequirements,e.g.,bystoringinformationaboutcapital 1992).Financialinformationmodulesaredesignedtomonitorcapitalresource 3ComputerImplementation mentedasasmalllibraryofobjectswrittenformicrosoftwindows(pissarides, nancialinformationmanagementandresourceallocation.eachmoduleisimple- DSSORAprovidesadecisionsupportenvironmentthatcomprisesmodulesfor- Fig.2representsthemutualrelationshipsbetweenthemodulesofDSSORA {amoduleofbudgetreallocationallowsuserstoexplorealternativefunding {anoptimizingmoduleusesthemtodetermineaninitialallocationbylinear programming,and calculateprioritiesforeachproject, whicharedescribedinmoredetailnext. 3.1ProjectValueAssessment byproposingsomesubstitutions,oneofwhichtobeselectedbydecision ThismoduleimplementstheAHPmethodologyonahierarchyformedbyatop makers. projects,asdescribedinsect.2.3andfig.1. levelcomprisingcriteriaforthecorporatemissionandabottomlevelforthe

Project requirements Project evaluation Resource availability Initial resource allocation limits.usersmustthencompareeachcriterionwitheachoftheotherones, ria,ithasbeenfoundthatvekeptthenumberofcomparisonswithintolerable corporatemission.althoughthereisnotheoreticallimitonthenumberofcrite- Usersarerstaskedtoinputatmostvenamesofcriteriacharacterizingthe Comparisonsofmissioncriteria Fig.2.DSSORAmodulediagram choosingoneofthefollowingcharacterizations: Interactive {Equal resource {ModeratelyMoreImportant allocation

TheinterfaceusedforthecomparisonsisshowninFig.3.Usersarerequiredto llintheboxeswithappropriatesymbols.whenallthecomparisonsaremade, therightmostcolumnoffig.3. thesystempresentsuserswithavectorofprioritiesforthecriteriadepictedin {StronglyMoreImportant {AbsolutelyMoreImportant {VeryStronglyMoreImportant Forthesecondlevelofthehierarchy,usersmustcompareeachpairofprojects,as Comparisonsofprojects followingthesameprocedureasforcriterioncomparisons.attheendofthe comparisonsforagivencriterion,thesystempresentsuserswiththeproject showninfig.4.theprojectsarecomparedpairwiseaccordingtoeachcriterion, Fig.3.Projectvalueassessment:missioncriteria;projects. prioritiesaccordingtothecurrentcriteriontogetherwithaninconsistencyratio (Saaty,1980).Ifthisratioistoohigh,userscangobackandcomparethepairsof projectsagain.whenalltheprojectsarecomparedaccordingtoallcriteria,the systempresentsuserswiththeoverallprojectandcriterionprioritiestogether withtheoverallinconsistencyratio.again,iftheratioistoohigh,userscan revisethecomparisonsinordertoobtainabetterratio.

algorithmwaseasedbyimplementingitasanobjectforanintegratedmanipulationofthesimplextableau,performingthefollowingfunctions: determinesaninitialallocationofresourcestoprojects.designingasimplex 3.2InitialResourceAllocation Theprioritiesareusedasobjectivefunctioncoecientsofalinearprogramthat Fig.4.Projectvalueassessment:projects. 3.3InteractiveResourceAllocation Themoduleforinteractiveallocationdisplaysthecurrentallocationinaneasy {keeptrackofalltheelementsinthetableau, inordertoobtainamoredesirableallocation,ordirectlytoswapsomeresources format.itallowsuserstochangesomeoftheinputsfromthepreviousmodules {keeptrackofthesizeofthetableau, betweenprojectswithoutviolatingtheconstraints.foreachproject,theinteractiveanalysisdisplaysinformationonobjectivecoecientsoftheoriginallinear program,itsmodiedvaluesandcorrespondingallocations,asshowninfig.5 {currentobjectivecoecient, {performpivotoperations. anditemizedbelow: {originalobjectivecoecientdeterminedbyahpanalysis,

Forsensitivityanalysisoftheresources,themodulealsodisplaysthefollowing {ascrollbartochangetheobjectivecoecient, {originalallocationdeterminedbylinearprogramming, {currentallocation, {ascrollbartochangetheallocation. Fig.5.Userinterfaceforinteractiveresourceallocationbysensitivityanalysis informationrelativetoeachconstraint: Themoduleoersfouroptionstochangetheallocationofresources: {currentrighthandsidevalue {originalrighthandsidevalue {ascrollbartoallowthechangeoftherighthandsidevalue, {originalvalueofthelefthandsidedeterminedbylinearprogramming, {currentvalueofthelefthandside. \Optimize"onthescreen,whichacceptsthenewdataasinputtothesimplex algorithmcontainedinthemoduledescribedpreviouslyanddisplaysthenew {changeofobjectivefunctioncoecients, {computer-assistedchangeoffundingallocation,and {unassistedchangeoffundingallocation. {changeofamountsofresources(therighthandsideoftheconstraints), Inthersttwooptions,theallocationismodiedbyselectingthemenuitem

allocationsbothnumericallyandonscrollbars.thereforeuserscanperform sensitivityanalysisonboththeobjectivecoecientsandtheconstraintright handsidevaluesbydraggingthescrollbarcursorsineitherdirection.infig.5, foranewobjectivefunction asalgebraicfunctions.inthelastoption,i.e.unassistedchange,theycansimply bylinearprogramming:inparticular,theyneednotexpressthemissioncriteria theallocationbecomes: Thelasttwooptions,illustratedinFig.6,allowuserstoby-passre-optimization max0:31x1+0:31x2+0:06x3+0:32x4; changetheallocationofresourcestotheprojectsbychangingthepositionof thecursorsonthescrollbars,usingthemouse.eachcursoronthescrollbars canmoveuntiloneoftheconstraintsisviolated.forexample,userscanachieve amoredesirablesolutionbydecreasingtheallocationtoprojectp2from30to x1=50;x2=25;x3=55;x4=50: Ofcourse,thenewsolutionmaynotbeoptimalwithrespecttotheobjective 26andincreasingtheallocationtoProjectP4from45to47.Inthisunassisted change,anyarbitraryallocationwillbepossiblewithintheconstraintsspecied. coecientsdisplayedbytheinterface. Fig.6.Userinterfaceforinteractiveresourceallocation:computer-assistedtrade-os

coecients(11). example,choosingthersttrade-oyieldsthenewallocationgivenby(12). generatedbythesystem,showninfig.6,thattransposesthematrixofreduced changetheallocationbypressingabuttononthescreen.thechangesarere- ofvalues,bothofwhichrepresentingtheallocationlevelstoeachproject.for ectedbythepositionofcursorsonthescrollbarsandthecorrespondingdisplay Eachtrade-oindicatesarateoffundingsubstitution.Userscandecideto Inanassistedchange,userscanselectoneallocationtrade-ofromalist problemathandbyexploringtheregionspeciedbytheconstraints.itdoes updatesthelistoftrade-os. oftherighthandsidevaluesbecomeszero,thesystemperformsapivotand inparticularneednotbepseudoconcave,aqualicationthatwouldrestrictthe notrequireanyanalyticexpressionoftheobjectiveorutilityfunctionwhich searchtoextremepointsonly. Thealgorithmisdesignedtohelpusersgetabetterunderstandingofthe Usersareallowedtoswitchtoadierenttrade-oatanytime.Whenone withdssoraconsistedofvemanagers.first,afunctionalgroupofsta 4FieldTrial AeldtrialwasconductedtotestwhetherDSSORAcouldsupportgroupdecisionsasrequiredbytheCompany'sbudgetingprocess.Twogroupsthatworked managersresponsibleforthefundingofaportfolioofveprogramslabeledp1, P2,P3,P4andP5(eachconsistingofnumerousprojects)hadtodeveloptheir fundingallocationrequestfortheprograms.thegroupheldamanagerialviewof eachprojectanditsroleintheportfolio,ratherthanatechnicalunderstanding ofitsfunctionsandvalueinthecorporatetelecommunicationsnetwork.consequently,itsmembersdieredfairlysubstantiallyintheirinitialevaluationofthsionaimedatdevelopingacommonunderstandingoftheroleoftheprojects Thiswasconrmedinourcase,asarstevaluativesessionsparkedadiscusparisonshaveintuitiveappeal,pinpointinginternalinconsistenciesofjudgment. responsibleforimplementationoftheseprogramsbeforetherequestcouldbe forwardedtothecontrolcommitteeresponsibleforfundingallocation. intable2.yet,consensuswasrequiredofthisrequestbythelinemanagers importanceofeachprogramtowardeachcorporatemissioncriterion,asshown adequacyofthemethod,theoutcomeofthesecondsessionofpairwiseevaluationswasthattheprioritiesofeachofthemanagersfortheprogramswereclose insatisfyingcorporatemissioncriteria,andofcriticalunderlyingassumptions underwhichallprojectshadtocompeteforfundingallocation.attestingtothe enoughtoreachaunanimousconsensusaboutthevalidityofthevaluesbased Saaty(1990)notedthatthehierarchicalrepresentationandthepairwisecom- ontheaverageinputofthegroup,displayedinthelastcolumnoftable2.the roleeachsessionofdssoratowardthisprogressiveconciliationisrepresented infig.7.

Fig.7.UsingDSSORAforconsensusbuilding

Project12345inputs P10.0790.0450.0370.0610.0660.063 P20.3220.1700.4660.3510.2510.311 P30.1350.0670.1110.0780.1580.111 Table2.Projectimportanceassessmentresults P40.2390.4670.1690.1250.0860.206 P50.2260.2500.2160.3860.4740.306 Basedoninputs frommanager Basedon average aninteractiveresourceallocationwasnecessary. 5Conclusion managers.giventheconsensus,inneitherexperimentdidthemanagersfeelthat theimplementationoftheprojectswithinthecorporatetelecommunications network.consensuswasreachedinonesession,andtheresultingprojectevaluationsweresignicantlyclosetotheresultsobtainedbytherstgroupofsta ThesecondgroupusingDSSORAcomprisedlinemanagersresponsiblefor ApplyingDSSORAtoresourceallocationwithinafunctionalunitlimitsthe projectmanagers.beyondtheinteractivemopmethodology,theeldtrial provedthatthesystemwasaneectivenegotiationtoolinagroupsetting. numberofprojectsunderconsideration,withseveralensuingbenets: DSSORAhasbeentestedwithveryconsistentresultsbyseveralgroupsof CapitalbudgetingintheCompanyspansmanyunitsandassociatedprograms. ToadaptDSSORAtointer-unitbudgeting,currentresearchfocusesoneective decompositionandaggregationtechniques(liang,1994).acentralissueisthe {increasedlikelihoodofconsensusbuilding, {limitedinterferenceofexternalconstraintswiththeconsensualevaluations, {relativelyfewpairwisecomparisons, {familiarityofthemanagerswiththeprojects, formationofsubportfoliosofprogramsbalancingtworequirements: {possibleacceptanceofresultsasabaseforimplementation. projects. portfoliosofprograms,followedbyintra-programallocationtotheconstituting Inthissetting,resourceallocationcouldproceedalongseveraltiers:allocationto {easethecomparisonsofprojects, {circumscribeprojectsthatshareimportanttechnologicaloroperationaldependencies.

References Bierman,H.Jr.:ImplementingCapitalBudgetingTechniques.BallingerPublishing Bromwich,M.:TheEconomicsofCapitalBudgeting.PitmanPublishingLimited,1979 Charnes,A.,Cooper,W.,Miller,M.H.:AnApplicationofLinearProgrammingto Clark,J.J.,Hinderland,T.J.,Pritchard,R.E.:CapitalBudgeting.Prentice-HallInc., Dantzig,G.B.:LinearProgrammingandExtensions.PrincetonUniversityPress, Crum,L.R.,Derkinderen,F.G.J.:CapitalBudgetingUnderConditionsofUncertainty.MartinusNijhoPublishing,1981 Company,Cambridge,Massachusetts,1988 FinancialBudgetingandtheCostofFunds.TheJournalofBusiness(1959)20{46 EnglewoodClis,NewJersey,1984 Hoadley,B.,Katz,P.,Sadrian,A.:ImprovingtheUtilityoftheBellcorebusiness. Georion,A.M.,Dyer,J.S.,FeinbergA.:AnInteractiveApproachforMultipleOp- Dean,J.:CapitalBudgeting.ColumbiaUniversityPress,NewYork,1951 Deckro,R.F.,Spahr,R.W.,Herbert,J.E.:PreferenceTrade-osinCapitalBudgeting Interfaces231(1993)27{43 timizationwithapplicationstotheoperationofanacademicdepartment.man- agementscience19(1972)357{368 Decisions.IIETransactions17(1965)332{337 Princeton,NewYork,1963 Pissarides,S:InteractiveMultipleCriteriaOptimizationforCapitalBudgeting.Master'sthesis,UniversityofOttawa,1992 Liang,Y.:CapitalBudgetingDecisionMaking:Database,AggregationandDisaggregationMethodsforaLargeScaleProblem.Master'sthesis,UniversityofOttawa, 1994 Wilkes,F.M.:CapitalBudgetingTechniques.Wiley,1983 Salo,A.,Hamalainen,R.P.:AModellingandDecisionsAidforSupportingTelecom- Saaty,T.L.:TheAnalyticHierarchyProcess.McGraw-HillInc.,1980. Spronk,J.:InteractiveMultipleGoalProgrammingasanAidforCapitalBudgeting tionsofuncertainty.martinusnijhopublishings(1981)188{212 municationsinvestments.proceedingsoftheiieeinternationalconferenceonsys- tems,manandcybernetics1(1989)115{118 andfinancialplanningwithmultiplegoals,in:capitalbudgetingundercondi- Zionts,S.,Wallenius,J.:AnInteractiveProgrammingMethodforSolvingtheMultiple CriteriaProblem.Management Science22(1976)652{663 ThisarticlewasprocessedusingtheLTEXmacropackagewithLLNCSstyle