Schedule risk and contingency using @RISK and probabilistic ...

  • Published on
    15-Jan-2017

  • View
    216

  • Download
    3

Embed Size (px)

Transcript

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysisAugust2010

    IanWallace

    PalisadeCorporation

    798CascadillaStreet

    Ithaca,NY14850USA

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    2

    CONTENTS Page

    1Introduction 3

    1.1Background 3

    1.2Whatisprojectriskanalysis 5

    1.3Whatisprobabilisticanalysis&MonteCarloSimulation 6

    1.4Probabilitydistributions 7

    1.5Scope/approach 8

    1.6Overviewof@RISKforProject 10

    2Quantifyingscheduleriskandcontingency 11

    2.1Overview 11

    2.2Step1enterprobabilitydistributions 13

    2.3Step2defineoutputsandrunthesimulation 20

    2.4Step3interpretresults 21

    3Benefits 24

    4Conclusion 25

    AppendixIUsefulreferences

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    3

    1 Introduction

    1.1 BackgroundThepercentagefailureofprojectsandthenumberofprojectoverrunsiswelldocumented.TheStandishGroup,forexample,intheir2009CHAOSreport,whichtracksprojectfailureratesacrossa

    broadrangeofcompaniesandindustries,reportedamarked decrease in project success rates. Inparticular:

    Only32%ofallprojectsweredeliveredontime,onbudget,andwiththerequiredfeaturesandfunctions,

    44%werechallenged,i.e.theywerelate,overbudget,and/orwithlessthantherequiredfeaturesandfunctions,

    24%failedduetobeingcancelledordeliveredandneverused.

    Previousreportshadsuggestedthat56%ofprojectshavecostoverrunsand84%ofprojectshavetimeoverruns,implyingthat,onaverage,only16%ofprojectsareontimeandonly44%achievetheirbudget.Allinall,apoorrecordinanyonestermsandthesituationseemstobegettingworse.

    Therecessionmayaccountforsomeoftheprojectcancellationsbutexperiencedprojectmanagerswillprobablyrecognisemanyofthecommonreasonsforprojectfailure,suchas:

    Projectgoalsarepoorlydefined,andoutcomesarenotidentifiedinspecificandmeasurableterms;

    Projectplanslacksufficientdetail,leadingtoinsufficienttimeallocationandinadequatefinancialsupportand/orotherresources;

    Keystakeholdersdonotbuyinandagreetoprovideadequatesupport; Ariskanalysisisnotperformed; Theprojectscopeexpandsuncontrollablyscopecreep(lackofchangecontrol); Poorcommunication.

    TheUKOfficeofGovernmentCommerce(OGC)andtheNationalauditOffice(NAO)provideasimilarlistforcommonreasonsofprojectfailureinGovernment,including:

    Lackofclearlinksbetweentheprojectandtheorganisation'skeystrategicpriorities,includingagreedmeasuresofsuccess.

    LackofclearseniormanagementandMinisterialownershipandleadership.

    Lackofeffectiveengagementwithstakeholders.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    4

    Lackofskillsandprovenapproachtoprojectmanagementandriskmanagement.

    Toolittleattentiontobreakingdevelopmentandimplementationintomanageablesteps.

    Evaluationofproposalsdrivenbyinitialpriceratherthanlongtermvalueformoney(especiallysecuringdeliveryofbusinessbenefits).

    Lackofunderstandingof,andcontactwiththesupplyindustryatseniorlevelsintheorganisation.

    Lackofeffectiveprojectteamintegrationbetweenclients,thesupplierteamandthesupplychain.

    Asmanyoftheseproblemsaredifficulttoaddressintheshortterm,itisclearthatprojectmanagersneedsomehelpinobtainingsufficientscheduleandcostcontingencytoavoidoverruns.

    ThepurposeofthisseriesofwhitepapersistosuggestwaysofusingprobabilisticanalysisandMonteCarlosimulationsothatmanagerscanvisualiseandquantifytheuncertaintyintheirprojectsandmakethoughtprovokingpredictionsofthelikelihoodofbeingontimeandonbudget.Withthisnewinformation,itwillbepossibletomakemoreinformeddecisionsabouttargetdates,pricing,budgetingandriskmanagement,aswellasmanagecustomerandstakeholderexpectationsmoreeffectively.

    Otherplannedpapersintheprojectseriesinclude:

    Costriskandcontingencyusing@RISKandprobabilisticanalysis,

    Projectprogressevaluationusing@RISKandprobabilisticanalysis.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    5

    1.2 WhatisprojectriskanalysisBestpracticeprojectmanagement,asadvocatedbyleadingmethodologiesprovidedbytheProjectManagementInstitute(PMI),theOGC(PRINCE2)andtheAssociationofprojectmanagers(APM)involvescarryingoutathoroughanalysisofpotentialrisksasearlyandasregularlyaspossible,inordertodeterminethelikelihoodofspecificriskeventsoccurringand,iftheydo,themagnitudeoftheirimpactontheproject.Usuallytheseeventshavenegativeconsequences,suchasdelaysorfinanciallosses,buttheycouldalsohavepositiveoutcomes.

    Perhapsmoreimportantly,theprocessofregularandongoingriskanalysisprovidesprojectmanagerswiththeopportunitytoorganisethedevelopmentofcontingencyplansandfunding,aswellasmanagetheexpectationsofimportantstakeholdersthroughformalandinformalreportingmechanisms.Belowisatypicaloverviewofprojectriskmanagementprocess:

    Theassessmentstageoftheprocesscanbeperformedbothqualitativelyandquantitatively,withthelatterattemptingtoassignnumericvaluestotheprobabilityandimpactofeachrisk,eitherbyusingempiricaldataorbyquantifyingqualitativeassessmentssuchashigh,mediumorlowlikelihoodofoccurrence,usingapercentage,anda3pointestimates(bestcase,mostlikelyimpact,worsecase)forpotentialimpact.Atypicalquantitativeprojectriskanalysis,usuallydevelopedinaspreadsheetandoftenreferredtoasariskregister,isshowninsection1.6.

    Althoughveryusefulforprioritisationandcontingencyplanning,thereareseverallimitationstothisdeterministicriskregisterapproach,including:

    Itconsidersonlythreediscretescenarios,ignoringhundredsorthousandsofothercombinationsofoutcomes.

    ProjectProject

    Identify

    Mon

    itor

    Mitigate

    Assess

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    6

    Itgivesequalweighttoeach3pointestimate.Thatis,noattemptismadetoassessthelikelihoodofeachoutcome.

    Interdependencebetweenrisks,aswellastheirultimateimpactontheprojectplanstargets,isignored,oversimplifyingthemodelandreducingitsusefulness.

    1.3 Whatisprobabilisticanalysis&MonteCarloSimulation InMonteCarlosimulation,riskanduncertaintyisrepresentedbyprobabilitydistributionswhichrecognisethateachvalueinarangeofpotentialoutcomeshasitsownprobabilityofoccurring.Probabilitydistributionsarethereforeamuchmorerealisticwayofdescribinguncertaintyinriskanalysis.Themostcommonlyusedprobabilitydistributionsaredescribedinsection1.4.

    DuringaMonteCarlosimulation(namedafterthefamouscasinos),valuesaresampledfromtheprobabilitydistributionshundredsorthousandsoftimesandthespreadsheetorprojectplanrecalculatedeachtime.Theserecalculationsallowustographthedistributionofhundredsorthousandsofpotentialscenarios.Inthisway,MonteCarlosimulationprovidesamuchmoreusefulviewofwhatmayhappenfordecisionmaking.

    Insummary,MonteCarlosimulationprovidesanumberofadvantagesover3pointdeterministicanalysis:

    Probabilisticanalysis,showingnotonlywhatcouldhappen,buthowlikelyeachoutcomeis.

    Graphicalanalysis.DuetothedataaMonteCarlosimulationgenerates,itiseasytocreategraphsofdifferentoutcomesandtheirchancesofoccurrence.Thisisimportantforcommunicatingfindingstostakeholders.

    SensitivityAnalysis.Workingwithjustthreescenarios,deterministicanalysismakesitdifficulttoseewhichrisksimpacttheprojectoutcomesmost(e.g.completiondatesandcostbudgets).MonteCarlosimulationallowsyoutoseewhichriskshavethebiggesteffectonbottomlineresultsveryusefulforallocatinglimitedmitigationresources.

    Correlationofrisks.Itisimportanttorepresentreallifeinterdependencesothatwhenaparticularriskoccurs,theprobabilityorimpactofothersgoesupordownaccordingly.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    7

    1.4 ProbabilityDistributions

    Inprojectmanagement,themostcommonlyusedprobabilitydistributionsareshownanddescribed

    below:

    Triangular PERTvsTriangular PERTvsUniform Discrete

    TriangularTheuserdefinestheminimum,mostlikely,andmaximumimpactofariskoruncertainty,similartothedeterministicapproach(inthiscase300,340and460daysrespectively).However,thevaluesaroundthemostlikelyhaveahigherrelativeprobabilitycomparedtothoseintheareaaroundtheminimumandmaximumvalues.Infact,usinghisdistribution,90%oftheprobability(i.e.areaofthetriangle)isbetween317.9daysand429days.

    PERT(ProjectEvaluationandReviewTechnique)Again,theuserdefinestheminimum,mostlikely,andmaximumimpactofariskoruncertainty,justlikethetriangulardistribution.However,thevaluesaroundthemostlikelyareevenmorelikelytooccurasthereisabuiltinweightingof4:1towardsthemostlikelyvalue.Thisreflectstherealityofalotofprojectuncertainties,asmanyofthemarepartiallywithinthecontroloftheprojectteamandthereforelesslikelytoachievetheirextremevalues,asshownbytheredlineinthePERTvsTriangulardiagram.Here,thePERThasathinnertailbecauseitisinherentlymoreconfidentthantheTriangulardistribution.

    UniformAllvalueshaveanequalchanceofoccurring,andtheusersimplydefinestheminimumandmaximum.Thisdistributionistheleastconfidentavailableandshouldbereservedforriskswhereusershavenoideaofthepotentialimpact.ThePERTvsUniformdiagramshowsquitedramaticallyhowmuchthicker(andthereforeuncertain)theUniformtailiscomparedtoaPERT.Thisisbecausenooneiswillingtoestimateamostlikelyvaluealwaysasignoflackofconfidence.

    Triang(300, 340, 460)

    Val

    ues

    x 10

    ^-2

    0.0

    0.2

    0.4

    0.6

    0.8

    1.0

    1.2

    1.4

    280

    300

    320

    340

    360

    380

    400

    420

    440

    460

    480

    5.0%90.0%317.9 429.0

    Triang(300, 340, 460) vsPert(300, 340, 460)

    Val

    ues

    x 10

    ^-2

    0.0

    0.2

    0.4

    0.6

    0.8

    1.0

    1.2

    1.4

    280

    300

    320

    340

    360

    380

    400

    420

    440

    460

    480

    5.0%90.0%317.9 429.0

    Uniform(300, 460) vs Pert(300,340, 460)

    Val

    ues

    x 10

    ^-2

    0.0

    0.2

    0.4

    0.6

    0.8

    1.0

    1.2

    1.4

    280

    300

    320

    340

    360

    380

    400

    420

    440

    460

    480

    90.0%308.0 452.0

    Discrete({x}, {p})

    0.0

    0.1

    0.2

    0.3

    0.4

    0.5

    0.6

    0 10 20 30 40 50 60 70

    90.0%10.00 60.00

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    8

    DiscreteHeretheuserdefinesspecificvaluesthatmayoccurandthelikelihoodofeachone.Inthediagramwehaveused20%chanceof10days,50%chanceof20days,20%chanceof50daysand10%chanceof60days.TheMonteCarlosimulationwillthereforeselectthesevaluesinthesameproportionduringasimulation.

    1.5 Scope/approachForthepurposesofthispaper,wewilluseasimpleprojectplantodevelopanewengineeringproduct,togetherwithalistof13potentialrisks,typicallyfoundinadeterministicprojectriskregisterasshownbelow:

    Inreallife,alistsuchasthiswouldbedevelopedviaworkshopmeetingsand/orothersourcessuchasprojectmanagementmethodologieswhichoftenprovidelistsoftypicalrisks.Asimilarlisthasalsobeenusedforcostrisk,discussedinseparatewhitepaper(Costriskandcontingencyusing@RISKandprobabilisticanalysis).Inbothcases,eachriskeventisbrokendownintotwocomponentstheprobabilityofoccurrenceandthemin/mostlikely/maximpactifeachriskoccurs.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    9

    UsingPalisades@RISKforProject,thiswhitepaperdemonstrateshowtocalculateschedule/timeriskandcontingencies,usingthreesteps:

    Enteringprobabilityandimpactdistributionsforeachrisk,fromtheriskregisterabove.

    DefiningtheoutputsrequiredandrunningaMonteCarlosimulation.

    Interpretingtheresults.

    Themodelusedassumesthattheprojectisstillinitsveryearlydays,withtheprojectplanonlydrawnupatahighlevelforthepurposesofabid/tenderorbusinesscase.Often,intheearlystagesofplanningandduringbids,theriskmodelcomesbeforeanyonehasgotaroundtoplanninginearnest.Thisisthetime,however,whencrucialriskmanagementdecisionshavetobemadeand,indeed,whenriskanalysiscanbeatitsmostcommerciallyadvantageous.

    TheWhitePaperentitledProjectprogressevaluationusing@RISKandprobabilisticanalysisdiscusshowprobabilitycanbeusedtomonitortheoutturnduringtheprojectlifecycleandseeifconfidenceisgrowingordeterioratinginthelightofactualprogress.

    NotethatalthoughitispossibletomodelhighlevelplansinExcel,ratherthanMSProject,thelattermakestheprocessmucheasierbecauseitwillautomaticallyworkoutthescheduledependingonhowtasksandactivitiesarelinkedtogether.UsingExcelinvolveshavingtowritecomplexformulafromscratchinordertomodeltherelationshipsbetweentasks.Inaddition,MSProjectallowsyoutoaddedandcostprojectresourcesatthesametime,andlater,controlactualworkagainsttheplan.

    Ineithercase,attemptingatahighlevelplan,evenifthereisverylittledetailavailable,helpstofocusandidentifyrisks(andopportunities)moreeffectively.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    10

    1.6 Overviewof@RISKforProject

    Triangularprobabilitydensityfunction

    Palisades@RISKforProjectisanaddinforMicrosoftProject,thewellknownplanningtool.Whenthesoftwareisloaded,anewtoolbarisprovidedinordertobeabletoaddprobabilitydistributionstothesinglepointestimatesinyourplanandrunaMonteCarlosimulation.TheMonteCarloalgorithmtakessampleswithineachdistributionbasedonthedensityofprobabilityimpliedbytheshapeandarea.Forexample,theshapeshownaboveimpliesthat90%ofthesamplesshouldbetakenbetween55.9and64.1days.Eachtimeasampleistaken,itisenteredintotheplansothatMSProjectcanrecalculatethefinishdate.Consequently,ifwerun1000iterationswewillobtain1000potentialfinishdates,alongwithadistributionofwheremostofthemlie.Thenarrowerthedistributionoffinishdates,themoreconfidenttheoutlook;thewiderthedistribution,themoreuncertaintheoutlook.Asweshallsee,thisinformationisextremelyusefulforquotingacceptablefinishdatesandsettinginternaltargetsandcontingencies.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    11

    2 Quantifyingscheduleriskandcontingency

    2.1 Overviewofriskmodel

    Baselinefinishdate10th Feb2012

    Estimatedmostlikelydurations

    Baselinestartdate1st Jan2011

    Activity/taskfinishtostartdependency

    Theplanwehaveusedtodemonstrateprobabilityanalysisisanearly,highleveldecisionmakingplan,consistingofthekeyactivities,linkedtogetherusingsimplefinishtostartdependencies.Therearealsothreemilestonedeliverableswhichareshownatthetopoftheplan:

    Agreeddesign/solution,

    Designtestingcomplete,

    Customeracceptance.

    Inreallife,thesemilestoneswouldmostlikelybelinkedtosomesortofcontractualpayment.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    12

    Startingon1stJanuary2011,thehighlevelfinishdateispredictedtobe10thFebruary2012usingsinglepointestimates.Oftenthisisknownasthebaselineplan.

    Inordertoaddriskanduncertainty,theprocessadvocatedhereistoleavetheoriginalestimatesuntouchedandtomodeltheriskregisterdescribedinsection1.5inaseparatesectionoftheplan,asshownbelow.Thishelpsavoiddoublecountingandforcespeopletobemorespecificaboutuncertainties,ratherthandescribingthemasvagueconcernsorpossiblypadding.

    Eachriskinitiallysetupasamilestoneswithzeroduration

    Eachriskislinkedtoanactivityormilestone

    Branchtyperisks

    UsingstandardMicrosoftProjectfunctionality,risks1to13havebeenenteredastaskswithzeroduration,likemilestones,andusingthesamedescriptionastheriskregister.TherisksmilestoneshavethenbeenlinkedtotheactivitiesorprojectmilestonesaffectedusingstandardMicrosoftprojectfinishstartdependencylinks.

    The@RISKmodellingforRisks13and813isdescribedinsection2.2.1.

    Forrisks4and5(bothqualitycontrolrisks),adifferenttypemodellinghasbeenusedcalledprobabilisticbranching,asdescribedinsection2.2.2.Thistechniqueintroducesadditionalactivitiesintotheplanifthequalitycontrolriskoccursandthedesignfailstomeettechnicalrequirements.

  • Scheduleriskandcontingencyusing@RISKandprobabilisticanalysis

    13

    2.2 Step1enterprobabilitydistributionsThreetechniqueshavebeenusedformodellingtheriskregister:

    RISK/IFstatementscontainingtheprobabilitydistributionsforeachrisksoccurrenceandimpact; Probabilisticbranchesformodellinguncertaintyaroundqualitycontrol(gettin...

Recommended

View more >