1
Thisisthepostpeerreviewedfinaldraftversionofthefollowingarticle:Heeks,R.
“Informationsystemsanddevelopingcountries:failure,successandlo cal
improvisat ions”,TheInf ormationSociety,18(2),101112,2002,whichhasbeen
publishedinfinalform at:
http://www.t andfonline.com/doi/abs/10.1080/01972240290075039
InformationSystemsandDeveloping
Countries:Failure,SuccessandLocal
Improvisations
RichardHeeks
InstituteforDevelopmentPolicyandManagement
UniversityofManchester
ManchesterM139GH
England
Email:richard.heeks@man.ac.uk
Abstract
Thispaperpresentsevidencethat–alongsidethesuccesses–manyinformation
systems in developingcountriescanbecategorisedasfailingeithertotallyor partially.
Itthendevelopsamodelwhichseekstoexplainthehighratesoffailure.Themodel
drawsoncontingencytheoryinor dertoadvancethenotionofdesign—actualitygaps:
thematchormismatchbetweenISdesignsandlocaluseractuality.Thishelpsidentify
twohighriskarchetypesthataffectISindevelopingcountries:countrycontextgaps
and'hard—soft'gaps.T hemodelisalsoofvalueinexplainingtheconstr aintsthat
existtolocalISimprovisationsindevelopingcountries.Overall,t hepapershowshow
modelandt heoryhelpunderstandIScasesindevelopingcountriesbut,equally,how
thosecasesprovidevaluabledatatohelpdevelopISmodelsandtheories.
Introduction:De finingandMeasuringSuccessandFailure
Domostinformationsystems(IS)projectsindevelopingcountries(DCs)succeedor
fail?Anyattempttoanswerthisquestionmuststartbycategorisingsuccessand
failure.Thebasisforcategorisationwasqualitativereviewof alargenumberofDCIS
casestudies(in InformationTechnologyforDevelopmentjo urnalandinedited
volumessuchasRocheandBlaine1996;OdedraStraub1996;AvgerouandWalsham
2000a;IFIP WG9.42000) .
Anysuccess/failurecategorisationrunsintosomeimmediatedifficultiesthatthispaper,
whilerecognising,cannotcompletelyresolve.Thefirstdifficultyis thesubjectivity of
evaluation –viewedfromdifferentperspectives,oneperson'sfailuremay beanother's
success(Lyyt inenandHirschheim1987;Sauer 1993).Thecategorisatio ndoestr yto
addressthiswithinthelimitsimposedbythesubjectivityofthecasestudy writers
themselves.
2
Theseconddifficultyisthetimingo fevaluation –today'sISsuccessmay be
tomorrow'sISfailure,andviceversa.Giventherelianceonreportedcases,whichare
dominatedbycrosssectionalratherthanlongitudinalanalyses,thisissuecouldnotbe
adequatelyincorporated.However,thechangingproclivitytofailureduringIS
implementationisdiscussedbelow.
InanalysingcasestudiesofISinDCs,threedominantcategoriesofreportedoutcome
emerged,asdescribedbelow.Whilenottheoreticallyexhaustive–where,forexample,
wouldoneplacea'failurefollowedbysuccess'case–thisthreewaycategorisation
did coverallthecasesreviewed.
First,therewasthetotalfailureofaninitiativeneverimplementedorinwhichanew
systemisimplementedbutimmediatelyabandoned.Suchanoutcomecanbedefined
relativelyobjectively.Forexample,India'sIndiraGandhiConservatio nMo nitoring
Centrewasintendedtobeanationalinfo rmat ionproviderbasedonasetofcore
environmental informationsystems.Despitemorethanayearofplanning,analysisand
designwork,theseinformat ionsystemsneverbecameoperational,andthewhole
initiativecollapsedshortlyafterwards(Purietal2000).
Asecondpossibleoutcomeisthepartialfailureofaninitiativeinwhichmajorgoals
areunattainedorinwhichtherearesignificantundesirableoutcomes.Insome cases,
whereonlyasubsetofinitiallystatedobjectiveshasbeenachieved,thenotionof
partialfailuremay berelativelystraightforward.Forexample,theTax
ComputerisationProjectinThailand'sRevenueDepart mentsetoutsevenareasof
taxationthatweretobecomputerised.Attheend oftheproject,onlytwoareashad
beenpartlycomputerised,andfiveotherswerenotoperational(Kitiyadisai2000) .
Wherecasesareanalysedlongitudinally,anothertypeof partialfailurecanemerge–
onethatparticularly seemstoaffectdevelopingcountries.Thisisthe'sustainability
failure'ofaninitiat ivethatatfirstsucceedsbutist henabandonedafterayearorso.
Anexampleisthecreationofasetoftouch screenkiosksforremoterural
communitiesinSouthAfrica'sNorthWestProvince.Thesewereinitiallywellreceived
byt hecommunities.However,thekiosks'lackofupdatedorlocalcontentandlackof
interactivityledtodisuse,andtheywerer emovedlessthanoneyearlater(Benjamin
2001).
Yetotherpartialfailuresaremoredifficulttoidentifybecauseidentificatio ngrapples
withtheissueofsubjectivity.Thisrequiresevaluationtoask:"Whosego alsare
unattained?"and"Forwhomaretheoutcomesundesirable?". Answerswillonly
appearwhereevaluat ionmet hods recognisefailure'ssubjectivity,andrecogniseand
interactwithmultiplestakeholdergroups.Suchrecognitionis,unfortunately,rarein
evaluationsofdevelopingcountry(andother)ISprojects.
TherewassuchrecognitioninanalysingtheAccountsandPersonnelComputerisation
Project ofGhana's VoltaRiverAuthority.Mostmanagerialstaffinthefinance
depart mentwerepleasedwit hthechangesbroughtbythenewsystem.However,the
implementation"bredafeelingofresentment,bitternessandalienation"amongsome
lowerlevelst aff,andledtoresistanceandnonuse,particularly amongolderworkers
(Tett ey2000:72).
3
Finally,onemayseethesuccessofaninitiativeinwhichmoststakeholdergroups
attaintheir majorgoalsanddonotexperiencesignificantundesirableoutcomes.This,
again,requirestherelativelysophisticatedapproachtoevaluationthatisabsentin
manycases.Inoneindepthevaluation,aSouthAfricantyremanufacturingfirm
introducedarelativelysimpleworkflowtrackingsystemusingbarcodesonthetyres.
Analysisfrommultiplest akeholderperspectivesshowedthatallthreekeygroups–
managers,supervisor s andworkers–perceivedthesystemtohavebrought benefitsto
theirwork(Calitz2000).
TheExtentofSuccessandFailure
WhatproportionofDCISprojectsfallintoeachofthet hreeoutcomecategories?
Nooneknowsforcertain.Thequestionishardeno ughtoanswerintheindustrialised
countries.There,atleast,acertainlevelofsurveys,evaluationsandanalysisispresent
(KoracBoisvertandKouzmin1995;James1997;Sauer1999; TheEconomist2000).
Onthebasisoftherangeoffiguresprovidedinthesesurveys,onemayestimatethat
somethinglikeonefifthtoonequarterofindustrialisedcount ryISprojectsfallintothe
totalfailurecategory;somethinglikeonethirdtot hreefifthsfallintot hepartialfailure
category;andthatonlyamino rityfallintothesuccesscategory.
This,atleast,canbeusedasathresholdindicatortoanswerthequestionabove.
Thereisnoevidence,noristhereanytheoreticalrationale,tosupporttheideathat
failureratesindevelopingcountriessho uldbeanylowerthanthoseinindustrialised
countries.Conversely, thereisevidenceandthereareplentyofpracticalr easons–
suchaslackoftechnicalandhumaninfrastructur e–tosupporttheideathatfailure
ratesinDCsmightbehigher,perhapsconsiderablyhigher,thanthisthreshold.
WhatistheevidencerelatingtoISsuccessandfailureindevelopingcountries?
Evidencetoaddresstheearlierquestion,andmovebeyondthethreshold estimations
offeredabove,isverylimited.Inadditiontopoorr ecognitionofsubjectivityand
timingof evaluat ion,theconstraintsonevidenceareseveral:
· Lackofliteratureingeneral:untilveryrecently,theentireliteratureonISand
developingcountrieswouldstruggletofillasinglebookshelf.Theattentionof
writers–fromresearcherstoconsult antstojournalists–hasbeenfocused
elsewhere.
· Lackofevaluation:thosewhohavethewilltoevaluate–suchasacademics–often
lacktheresourcesandcapacity.Thosewhohavetheresources–suchasaiddonor
agencies oftenlackthewillto evaluate.
· Focusoncasestudies:theliterat ureonISinDCshasgrown,butitisaliterature
dominatedbycasestudiesofindividualISprojects.Takenalone,theseprovideno
basisforestimationofoverallfailure/successrates.
Despitetheselimitations,therearesomeglimpsesofevidence.Anoverviewofthe
literatureconcludes,"successfulexamplesofcomputerisationcanbefound…but
frustratingstoriesofsystemswhichfailedtofulfiltheirinitialpromisearemore
frequent"(AvgerouandWalsham2000b:1). Thisshowsupinoverallevidence:IT
capitalshowsnosignificantcorrelationwithproductivityindevelopingcountries
whileinindustrialisedcountriesthereisapositivecorrelation(KraemerandDedrick
2001).Likewise,I TinvestmentshowsnosignificantreturnsinDCsbut80%gross
returnsinOECDcountries.
4
Afewmorespecificmultiplecasestudieshavebeenconducted,withexamples
summarisedbelow:
· HealthinformationsystemsinSouthAfrica:BraaandHedberg(200*)report
widespreadpartialfailureofhighcostsystemswithlittleuseofdata.
· IS in theThaipublicsector:Kitiyadisai(2000)reports"failurecasesseemtobethe
norminThailandatallgovernmentallevels".
· DonorfundedITproject s in China:BaarkandHeeks(1999)reportedthatallwere
foundtobepartialfailures.
· WorldBankfundedIT projectsinAfrica:MoussaandSchware(1992)report
almostallaspartial –oftensustainability –failures.
Likewise,reportsfromindividualdevelopingcountr ies(e.g.WorldBank1993;
Oyomno1996)findISfailuretobethedominanttheme.
Insummary, theevidencebaseisnotstrong–anditurgentlyneedsstrengthening–but
itallpointsinonedirection:towardshighrat esofISfailureindevelopingcountries.If
thisisso,weshouldseektounderstandwhy.Thatistheintentionofthispaper–to
developandthenapplyamodelthathelpsexplainwhysomanyinformationsystemsin
developingcountriesfail.
Beforemovingontothis,though,onefurtherquestionshouldbeaddressed.Isthe
prevalenceoffailureaproblem?Forexample,failurecanhavebenefits,especiallyin
relationtolearning.Unfortunately,whilelearningfromISfailuredoesoccur,itis
generallyfortuitousrathert hanplanned(MaciasChapula2000).Therearefewsigns
ofthepresenceoflearningsystemsinDCorganisations,andsomesignsoftheir
absence(Shukla1997).
Inaverydirectsensefailureisalsoaproblembecauseoftheopportunity costsof
resour ceinvestmentinfailur e,asopposedtosuccess.Suchopportunitycostsare
likelytobeparticularlyhighinDCsbecauseof themorelimitedavailabilityof
resour cessuchascapitalandskilledlabour.
Finally,thecostsofalltypesoffailureidentifiedabove–uncompleted/abandoned
projects;projectsthatfailtomeetobjectivesorwhichfailtosatisfykeystakeholders;
andpr ojectswhichcannotbesustained–arehighbecauseonlysuccessful projectswill
ensureglobaleconomicconvergence(Kenny2001).Thefailureskeepdeveloping
countriesonthewrongsideofthedigitaldivide,turningI Tintoatechnologyofglobal
inequality. Forallthesereasons,ISfailureisthereforeaveryrealandverypractical
problemfordevelopingcountriesthatneedstobeaddressed.
UnderstandingDCISSuccessandFailure
Wehave anestimationthatasignificantnumbero fISprojectsindevelopingcountries
failinsomeway.Whyshouldthisbe?Twobodiesofliteraturerelatet othisquestion:
thegeneralliteratureonISfailur e(e.g.LyytinenandHirschheim 1987;Hortonand
Lewis1991);andthespecificliteratureonISfailureindevelopingcountries(e.g.
MattaandBoutros1989;Boon1992;BeeharryandSchneider1996). Bothhavebeen
usefulinhelpingbuildtheoverallbodyofknowledge. However,t herehavebeen
criticismsthatsuchwritingshavebeenpooratexplainingcauses;pooratidentifying
5
responsestofailure;andthatworkis toonormativeandprescriptive,failingtotake
accountofthemanydifferingcontextsinwhichinfo rmat ionsystemsareimplemented
(Sauer1993;PoulymenakouandHolmes1996;Montealegre1999).
Thispaperthereforeseekstobuildanewmodeltounderstandthesuccessandfailure
ofinformatio nsystems;amodelthatwillbothexplaincausesandidentifyresponses.
Inordertoavoidthepit fallsofthenormativemodels,thestartingpointmustbe
contingency,whichhasagreatvalueindealingwiththecomplexityofISand
organisations (RobeyandBoudreau1999).Contingencyseesnosingleblueprint for
successandfailureinorganisationalchange.Instead,itrecognisesthatthereare
situationspecificfactorsforeachinfo rmat ionsystemwhichwilldeterminesuccessand
failureand,hence,strat egiesfor success.
Inherentwithinmuchoftheorganisationalliteratureoncontingencyistheideaof fit or
congruence:ofmismatchandmatchbetweenfactors (LorschandMorse1974;But ler
1991).Failureisseentoderivefromlackoffitbetweenfactors;successfrom
congruencebetweenfactors.ISimplementationmodelsdrawingonthisthemehave
involvedfitbetween'tool'/'technology'and'task'(GoodhueandThompson1995;
AgarwalandSinha1996),orbetweenmultiplefactors(ScottMorton1991;Southon
etal1997).
AcommonbaseforthesemodelsisLeavitt's(1965)diamond,illustratedinFigure1.
Inallcases,theconceptisoneof'dimensionalfit':theneedforoneormoredifferent
dimensionso forganisationand/or environmentt obebroughtintocongruenceatthe
sametime.
Figure1:DimensionalCont ingency
Althoughvaluable,theseideashavetwokeyshortcomingsinhelpingunderstandand
explainISfailure:
· Thereisarelativelypoor conceptionoforganisationalchange. Earliermodelsdo
notadequatelyaccountforthefactthat,thegreaterthedegreeofchange,the
great ertheriskoffailure(MoussaandSchware1992;Clemonsetal1995;Sauer
1999;Kitiyadisai2000).Dimensionalfit–requiringdifferentdimensionstobe
congruentatthesametime–istoostaticamodeltohandle theprocessofchange.
Insteadamoretemporalconceptionoffitisneeded;onethatconceivesthesame
dimensionsbeingcongruentatdifferenttimes.Thisrequir esthenotionoffittobe
Processes
People
Structure
Technology