Factors Affecting the Success Rate of Software Dev

FactorsAffectingtheSuccessRateofSoftwareDevelopmentProjectsRobertMiltonUnderwood,Jr.©2001RobertMiltonUnderwood,Jr.Page2©2000TableofContentsSectionpageAbstract3Introduction3Method6Participants7Materials7DesignandProcedure7Results8Discussion11Conclusions14References18Appendix20Exhibit121Exhibit222RobertMiltonUnderwood,Jr.Page3©2000FactorsAffectingtheSuccessRateofSoftwareDevelopmentProjectsAbstractResearch(StandishGroup,1994)hasrevealedthatthirty-onepercentofsoftwareprojectswereterminatedpriortocompletion.Onlyninepercentofsoftwareprojectsforlargecompaniesand16percentofsoftwareprojectsforsmallcompanieswerecompletedontimeandwithintheinitialbudget.Itisimportantformanagerstorealizethatimprovedcommunicationsbetweenallpartiesinvolvedandbetterplanningintheearlystagesofthesoftwaredevelopmentlifecyclewillimprovethesuccessrateofsoftwareprojects.IntroductionAccordingtoPfleeger(1998),mostsoftwareproductsarenotfreefromfaults.Withthousandsorevenmillionsoflinesofcode,itcanbeadauntingtasktodeliverabug-freeproducttoatargetmarket.Whileitmaybeunreasonabletorequirea100percenterror-freeproduct,thereisstillmuchroomforimprovement.Afewdistinctionsexistbetweensoftwareengineeringandtraditionalfieldsofengineering(Brookshear,2000).Traditionalfieldsofengineering(e.g.,mechanicalengineering)havealwaysbeenabletousepreviouslydevelopedcomponentsasbuildingblocks.Withsoftwareengineering,previouslydevelopedcomponentstendtohaveaninternaldesignthatisdependentonaspecificapplication.Toreuseacomponentforanewapplicationwouldrequirethatitbereengineered.RobertMiltonUnderwood,Jr.Page4©2000Aseconddistinctionbetweensoftwareengineeringandtraditionalfieldsofengineeringdealswiththeroleoftolerances.Thefunctioningofanairconditioningsysteminahomemaybeconsideredacceptableifitcoolsthehomewithinacertainacceptablerangeofcoolingdifferential.Software,incontrast,eitherworksproperly,oritdoesnotoperateproperly.Forinstance,aspreadsheetprogramthatoperateswithinaonepercentmarginorerrorisnotacceptable.Athirddistinctionbetweensoftwareengineeringandtraditionalfieldsofengineeringisidentifiedasthelackofquantitativesystemsformeasuringpropertiesofsoftware.ConsumerReportsMagazinereportsmonthlyonthequalityofvariousmechanicaldevicesbymeasuringthemeantimebetweenfailures,whichmeasureshowwelltheitemhandleswearandtear.Incontrast,softwaredoesnotwearout,soproceduresofmeasuringqualitydonotapply.Therearetwomajorapproachestosoftwareengineering:thesystemsapproach,andtheengineeringapproach.Thetwoapproachesarenotnecessarilymutuallyexclusive.Theengineeringapproachcanandshouldbuildfromthesystemsapproach.Thatis,assoonastheelementsofthesystemareidentified,categorized,andprioritized,thentheengineeringapproachcanbeusedtomaketheelementsofthatsystemadevelopmentalproject.Besidesthetwobasicapproachestosoftwareengineering,modelingshouldbeconsideredwhenattemptingtoimprovesoftwaredevelopmentprojects.Modelsarebasicallycategorizedeitherstaticordynamic.AccordingtoPfleeger(1998),astaticmodeldepictsaprocess,showingthattheinputsaretransformedtooutputs.Adynamicmodelcandepicthowproductsaretransformedatvariousstepsoftheprocess.RobertMiltonUnderwood,Jr.Page5©2000Therehavebeenanumberofmodelsthathavebeendesignedbyacademicsthatpurporttoconsiderallfactorsrelevanttoandnecessaryforthesuccessofaproject.Pfleeger(1998)presentsanintroductiontoseveralmodels,includingthewaterfallmodel,thewaterfallmodelwithprototyping,theVmodel,theprototypingmodel,theoperationalspecificationmodel,McCall’squalitymodel,thetransformationalmodel,thephaseddevelopmentmodel,theCapabilityMaturityModel,andthespiralmodel.Eachmodelpresentsauniquewayofrepresentationoffactorsthatmustbeconsideredinaproject.Itmightbewonderedwhetherornotatwo-dimensionalrepresentation(i.e.,awrittendiagram)caneverbesufficienttomodelthe“realworld,”especiallysinceweliveandoperateinathree-dimensionalworld.Actually,weliveinthethreedimensionsofphysicalspace,plusthefourthdimensionoftime(Rucker,1984).Wecontinuallyprogressforwardintime.Wecanalsogobackintime,figurativelyspeaking,bystudyinghistoricallyrelevantsituationsand/orcasestudies.Soitseems,therefore,thatamodelshouldbeusedonlyasaguideline.Perhapsathreedimensionalmodelwouldbemoreaccuratethanatwo-dimensionalmodelinthatactionsandproceduresrelatedtoadesignprojectcanoccurnotonlyatdifferenttimes,butatdifferentplaces.Computermodelingcanapproximatethree-dimensionalprojections,butaneffectivemanagermustconsiderhowthatmodelwillevolveordevelopovertime.Projectmanagerswhorelysolelyonmodelsdepictedtwodimensionallyorlinearlywillneglectsomeoftheconsiderationsnecessarytosuccessfullycompleteasoftwareprojectontimeandwithintheinitialbudget.RobertMiltonUnderwood,Jr.Page6©2000Perhapsanintegrationofmodelingapproacheswouldbemoreeffective.Ingeometry,forinstance,thestudyofpolyhedron1modelshasrevealeddualmodels,inwhichtwodifferentthree-dimensionalshapescomplimenteachotherbybeingsymmetricallycompatible(Wenninger,1983).Aneffectivemanagerofsoftwaredevelopmentprojectsmayneedtointegratetwoormoremodelingapproachesthatmayseemdissimilar,butareactuallycomplimentary.Theimpo