The Case Research Strategy in Studies of Information Systems - Wiki

CaseResearch Strategies

The Case Research Strategyin Studiesof Information Systems By: Izak Benbasat Professor & Chairmanof MIS Faculty of Commerceand Business Administration The University of British Columbia Vancouver, Canada V6T IY8 By: David K. Goldstein Assistant Professor of Business Administration Harvard Business School Harvard University Boston, Massachusetts 02163 By: Melissa Mead Assistant Professor of Business Administration Harvard Business School Harvard University Boston, Massachusetts 02163

Abstract This article definesanddiscussesone of these qualitative methods--thecaseresearchstrategy. Suggestions are providedfor researchers who wish to undertake research employing this approach.Criteria for the evaluation of case research are established and several characteristics useful for categorizing the studies are identified. A sampleof papers drawnfrom information systemsjournals is reviewed. The paper concludeswith examples of researchareasthat are particularly wellsuited to investigation usingthe caseresearch approach. ACMcategories: H.O., J.O. Keywords: Casestudy research, data collection, informationsystems,qualitative methods,research methods.

Introduction Therehas beena growinginterest in the use of qualitative techniquesin the administrative sciences. For example, a full issue of Administrative Science Quarterly (Volume24, 1979) has been devotedto qualitative methods. This interest has been sparked by a general dissatisfaction with the type of research information provided by quantitative techniques [51]. The dissatisfaction stems from several sources: the complexity of multivariate researchmethods,the distribution restrictions inherentin the useof these methods (e.g., multivariate normality), the large samplesizes these methodsdictate, and the difficulty of understandingand interpreting the results of studies in whichcomplexquantitative methodsare used. Similarly, in the informationsystems (IS) field, Franz and Robey[14] have suggestedthe use of idiographic rather than nomothetic research strategies. Idiographic research attempts to understand a phenomenon in its context. In suchresearch,the investigator intensely examines a single entity or a particular event. Nomotheticmethods,on the other hand, seek general laws and draw solely on proceduresused in the exact sciences[54]. This article discussesthe useof one qualitative technique,the caseresearchstrategy, in studies of information systems.It provides somesuggestions about how to conduct and evaluate case study research. A sampleof case-basedresearch from selected IS journals is categorizedaccordingto a set of characteristics developed in this paper. Thearticles in the sampleare then evaluated. Weare not advocatingan exclusive use of the case strategy. Manyauthors have commented that each research strategy has advantages and disadvantages;no strategy is more appropriate than all others for all research purposes.Benbasat[3] showedthat the goals of the researcher and the nature of the research topic influence the selection of a strategy. Caseresearchis particularly appropriate for certain types of problems:those in whichresearchand theory are at their early, formativestages[44], and"sticky, practicebasedproblemswherethe experiencesof the actors are importantand the context of action is critical" [4].

MIS Quarterly~September1987 369

CaseResearch Strategies

The information systemsarea is characterized by constanttechnological changeand innovation. IS researchers,therefore, often find themselvestrailing behind practitioners in proposing changesor in evaluating methods for developing new systems. Researchers usually learn by studying the innovations put in place by practitioners, rather than by providing the initial wisdomfor these novel ideas. For example, whencompaniesexperienced a growthin end-usercomputingin the late 1970sand early 1980s, academicswere not able to’offer a set of guidelinesdescribing howan organization could effectively manage the introduction of end-usercomputingtechnology. Researchers first descriptively studied how organizations were managingenduser computing. Thesestudies then formed the basis for the development of prescriptive management guidelines (for example, Rockart andFlannery[43]). Webelieve that the caseresearchstrategy is well-suited to capturing the knowledgeof practitioners and developingtheories from it. Christenson[9] points out that the trial-anderror processin which practitioners are engagedis necessaryfor knowledgeto accumulate. It is incumbent uponthe scientists to formalize this knowledgeand proceedto a testing stage. Before this formalization takes place, case studies could be employedto documentthe experiencesof practice. TheIS field has also seena shift fromtechnological to managerial and organizational questions, and consequentlymoreinterest in howcontext and innovations interact. For example,airline reservation systemswere very innovative technical achievements in the early 1960s. However,they becamea key competitive factor in the changingairline industry within the last few years. In order to understand this phenomenon,one must examine the structure of the industry, the role of deregulation, and the federal laws governing the industry. To summarize,there are three reasons why case study research is a viable information systems research strategy. First, the researcher can study information systemsin a natural setting, learn aboutthe state of the art, and generatetheories from practice. Second, the case methodallows the researcher to answer"how" and "why"questions, that is, to understand the nature and complexity of

370

MISQuarterly~September 1987

the processestaking place. Questionssuch as, "Howdoes a managereffectively introducenewinformation technologies?"are critical ones for researchersto pursue. Third, a case approachis an appropriate way to researchan area in which few previous studies havebeencarried out. With the rapid paceof changein the information systemsfield, many newtopics emergeeach year for which valuable insights can be gainedthroughthe use of case research.

Case Research: Definition There is no standard definition of a case study. For our purposes,wewill drawour definition from those presentedby Benbasat [3], Bonoma [5], Kaplan[23], Stone[46], and Yin [56]. A case study examinesa phenomenon in its natural setting, employingmultiple methodsof data collection to gather information from one or a few entities (people, groups, or organizations). Theboundariesof the phenomenon are not clearly evident at the outset of the research and no experimental control or manipulationis used. Table 1 contains a list of eleven characteristics of case studies summarizedfrom the papers mentioned above. To place case studies in perspective, it is useful to contrast this approachwith other methodscommonlyused by IS researchers. In laboratory experiments the researcher measuresdependentvariables while manipulating independentvariables in a controlled, environment.Similarly, field experimentsinvolve the manipulation and measurement of clearly definedvariables, but in a natural setting. Finally, in field studies researchersmeasure independentand dependentvariables in their natural context; however,no control or manipulationis involved. A fundamentaldifference betweencase studies and these alternative methods is that the case study researcher mayhave less a priori knowledge of whatthe variables of interest will be and 1 howthey will be measured. Wemustemphasize, however, that this is a matter of degree.Thereare instancesof casestudies where theinvestigators hada prior’notionof certain critical variables,suchasthetypeof industriesand thetypeof firms theywanted to examine. Forexample, Lawrence andLorsch[29] choseindustriesthat

CaseResearch Strategies

Table 1. KeyCharacteristics of CaseStudies 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Phenomenon is examinedin a natural setting. Data are collected by multiple means. Oneor few entities (person, group, or organization) are examined. Thecomplexityof the unit is studied intensively. Casestudies are moresuitable for the exploration, classification and hypothesisdevelopmentstages of the knowledgebuilding process; the investigator should have a receptive attitude towardsexploration. Noexperimentalcontrols or manipulationare involved. Theinvestigator maynot specify the set of independentand dependentvariables in advance. Theresults derived dependheavily on the integrative powersof the investigator. Changes in site selection and data collection methods couldtake place as the investigator develops newhypotheses. Caseresearchis useful in the study of "why"and "how" questionsbecausethese deal with operational links to be traced over time rather than with frequencyor incidence. The focus is on contemporaryevents.

In our surveyof the literature, weidentified three categoriesof qualitative researchthat appearto be consideredas case studies. Our focus in this article, however,includes only oneof these. For clarification wewill define the two excluded categories-application descriptions and action research. Thenwe will describekey aspectsof the executionof case research. Applicationdescriptions--Written by practitioners, application descriptions detail the author’s experiencesimplementinga particular application, such as a databasemanagement system. The outcomesof these published projects are almost alwayssuccessful andthe authorconcludesby providinga list of "dos" and "don’ts" for the implementationof similar systems.Theauthor does not conduct a researchstudy; instead, the objective is to successfully implementa specific systemfor a given assignment.Application descriptions are not included in our definition of case research. Action research--We have also excluded action research. Theseare studies in which the author, usually a researcher, is a participant in the implementation of a system,but simultaneouslywantsto evaluatea certain intervention technique. An examplemight be the use of the sociotechnical approachfor differedin their ratesof technological change. They alsostudiedeffectiveandineffectiveorganizations withinthesame industryas wellas comparing effective organizations in differentindustries.

systemdevelopment. Thearticles in both this and the application descriptions categories are written by individuals whohave an insider’s view of the systemin question. However, action researcharticles are authoredby those whoseoriginal intent is to conduct research while effecting change[47]. The action researcheris not an independentobserver, but becomesa participant, and the processof changebecomes the subject of research. Thus, the researcher has two objectives: to take action to solve a problemand to contribute to a set of system development concepts[8]. Thestrength of these studies is the indepth and first handunderstandingthe researcher obtains. Conversely,a weakness is the potential lack of objectivity stemmingfrom the researcher’s stake in effecting a successful outcome for the client organization. Further, generalizationsto other situations wherethe intervention technique is applied by people less knowledgeablethan the researcher may be difficult. Examples of action researchin IS are foundin Gibson[16], Ginzberg[17], Mumford [35] and Morton[34]. Casestudy research-In case studies the clear objective is the conduct of research. Theseare efforts where research questions are specified prior to the study by researchers whoare observers/investigators rather than participants. Wewill discuss considerations that are important in selecting the caseresearch approach and we will detail the mechanicsof executing case research.

MIS Quarterly~September1987 371

CaseResearch Strategies

Conducting Case Research Research themes and objectivesDeciding on case research Given a specific researchquestion, one must ask whether the case methodis a useful approach. To judge the appropriatenessof the casestrategy, wecan ask the following questions (drawnfrom Table 1): 1. 2. 3. 4.

Canthe phenomenon of interest be studied outside its natural setting? Must the study focus on contemporary events? Is control or manipulationof subjects or events necessary? Does the phenomenon of interest enjoy an established theoretical base?

Casemethodologyis clearly useful whena natural setting or a focus on contemporary events is needed. Similarly, research phenomena not supportedby a strong theoretical base maybe fruitfully pursuedthrough case research.A rich natural setting canbe fertile groundfor generating theories. Conversely, whensubjects or events mustbe controlled or manipulatedin the course of a researchproject, the caseapproachis not suitable. Quite often, however,the decision to use a caseapproachis not clear-cut. Both Yin [56] and Bonoma [5] discuss the usefulnessof the caseapproachin various phasesof research. Table2 gives their terminologyfor the traditional exploration, hypothesis generation, and testing phasesof knowledgeaccrual. Research that is not strictly exploratory or descriptive maybe enhancedby using quail-

tative methods.For example,Yin [56] states that case studies could be used to explain phenomena. Heconsiders Allison’s [1] study of the Cubanmissile crisis an exampleof suchstudies. Allison proposeddifferent theories to account for the same course of events, identified the one that provided the best explanation, and suggested how this theory could be useful to understandother situations. Bonomasuggests that the case strategy could play a role in both hypothesis generation and hypothesistesting. Bauer(as reportedin Towl[50]) refers to the useof the critical case(crucial experiment) to test a wellfoundedtheory. Whenthe case approachis deemed appropriate, researchersmaybe uncertain about how to proceed. The remainder of this section offers practical aid to researchersfor understanding and implementing case research. Unit of Analysis Prior to searchingfor sites, the researcher shoulddeterminethe unit of analysis mostappropriate for the project. Will the studyfocus on individuals, groups (e.g., a task force, profit center, IS group)or an entire organization? Alternatively, the unit of analysismaybe a specific project or decision. In makingthis determination, the researchershould closely examinethe research questions to be pursued. Theseoften indicate an appropriate unit of analysis. Finally, the researchershould consider what generalizations are hopedfor at the project’s completion. Does the researcher hope to generalize to other organizations, individuals, or decisions, for instance?

Table 2. Terminologiesfor Stagesof Case Research Programs Traditional Phases of Knowledge Accrual

Yin’s [56] Framework

Bonoma’s[5] Framework

Number of Cases

Exploration

Description

Drift

Single or multiple case(s)

Hypothesis generation

Exploration

Design

Multiple cases

¯ Confirmation

Explanation

Prediction

Multiple cases

¯ Disconfirmation

Explanation

Disconfirmation

Singlecritical case.

Hypothesistesting

372

MISQuarterly~September 1987

CaseResearch Strategies

Whenthe research is highly exploratory, a single case maybe useful as a pilot study. Thegoal will be to determinethe appropriate unit and familiarize the researcherwith the phenomenon in its context. Single-Casevs. Multiple-Case Designs Central to caseresearchdesign is the decision to includeoneor severalcasesin the project. Mostresearchefforts require multiple cases,but single casesare useful in specific instances. Yin suggestssingle-case studies are appropriateif: 1) 2) 3)

It is a reve/atory case,i.e., it is a situation previously inaccessible to scientific investigation. It representsa critica/case for testing a 2 well-formulatedtheory. It is an extremeor uniquecase.

As Table 2 shows,single-case study projects are mostuseful at the outset of theorygeneration andlate in theory testing. A single case used for exploration maybe followed by a multiple-case study. This correspondsto Bonoma’sdrift stage in which researcherslearn first handthe relevant jargon and context in which the phenomenon will be studied. A single casemayalso be used to test the boundaries of well-formedtheory. Multiple-case designs are desirable whenthe intent of the researchis description, theory building, or theory testing. Thesethree correspondto Bonoma’sdesign, prediction, and disconfirmation stages, respectively. Multiple-case designsallow for cross-caseanalysis and the extension of theory. Of course, multiple cases yield moregeneral research results. Site Selection Thefactors that dictate a single-case design also determinesite selection. Whenmultiple casesare to be included in a study, however, choices must be made.It is quite useful to considera multiple-caseproject as analogous to the replication that is possiblewith multiple traditional experiments[19]. Adopting this point of view, Yin proposestwo criteria for selecting potential sites. First, sites where According to Yin[56, p. 42]"Toconfirm,challenge or extend a theory,theremay exist a singlecase,meetingall theconditions for testingthetheory."

similar results are predicted maybe used as "literal" replications. Second,sites maybe chosenfor "theoretical" replication. Thatis, chosensuch that contradictory results are predicted. With careful site selection, the researchercan extendand revise the initial propositionsof the study. Site selection shouldbe carefully thoughtout rather than opportunistic. Researchersmay begin site selection by consideringthe nature of their topic. Research on organization-level phenomenawould require site selection basedon the characteristics of firms. These mayinclude the industry, company size, organizationalstructure, profit/not-for-profit status, public or private ownership,geographic coverage,degreeof vertical or horizontal integration, and so on. Researchers interested in specific technologies,IS methodologies or organizational structures should consider these characteristics whenselecting sites. Oncethe limiting factors are determined, specific sites maybe identified and approached. Regular scans of business newspapersandperiodicals often turn up potential sites. Library researchusing indexesof industries, businessliterature, and marketing or financial researchdata maybe helpful. Finally, talking with friends, colleagues,or acquaintances is a goodwayto identify potential researchlocations. Approachingthe potential site is a crucial point in orchestratinga caseresearchproject. Hereagain, the topic of studyis key to determining whomto contact. The researchermust eventually contact the individual with enough authority to approvethe project. Colleagues maybe able to help with introductions. If not, preparecarefully beforeplacing a cold call or writing to the organization. The researcher should clearly describe the project and who will be involved--researchers,assistants, or companyemployees. The contact should be told the amountof time, effort andexpense required of the organization. Twokey points to be addressedin order to gain cooperationare confidentiality and benefits to the organization. The researcher mustprovide assurancethat the organization will not be harmed by its participation. Theorganization and its employeesmust knowthat the researcher will not betray their con-

MIS Quarterly~September1987 373

CaseResearch Strategies

fidence. Onthe other hand, the researcher should seek assurancethat reasonable candor will be provided and that essential data will be madeavailable. Thebenefits to an organization participating in a researchproject are varied. Theymayinclude learning moreabout the organization, getting feedback and newinsights from the researcher, and developing a relationship with the researcher.In addition, there is the opportunity to contribute to knowledgeand business research. The organization mayor maynot wish to be identified whenthe researchis published.If it does,thereis the additional benefit of recognitionandpublicity. Data Collection Methods Multiple data collection methods are typically employedin case research studies. Ideally, evidence from two or moresources will converge to support the research findings. Yin identifies several Sourcesof evidencethat workwell in caseresearch[56, p. 78]. 1. 2. 3. 4.

5.

Documentation--Written material ranging from memoranda to newspaperclippings to formal reports. Archival records--Organization charts; service, personnelor financial records. Interviews--These maybe open-en.ded or focused (See Bouchard[6]; Cookand Campbell[10]). Direct observation--Absorbing and noting details, actions, or subtleties of the field environment(see Webb,et al., [53] on unobtrusive measures). Physical artifacts--Devices, outputs, tools.

Thegoal is to obtain a rich set of data surroundingthe specific researchissue, as well as capturing the contextual complexity. Specific data to be collected will dependon the researchquestionsand the unit of analysis. Prior to site visits, the researchershould outline, in detail, the datato begathered.This mayinclude a list of materials to be collected (documentation,archival records and physical artifacts) as well as questionsfor interviews and plans for direct observation. This formalization helps coordination whenmultiple investigators worktogether. It also provides someseparation of data collection from data analysis. The goals of this planning should be to ensuregoodcoverageof the re-

374

MISQuarter/y/September 1987

search questions and excellent use of time spent on-site. This planning stage helps to structure projects that are inherently flexible. It gives the researchera guide from whichto work. As the project unfolds, the plan will be revised accordingto the researcher’s judgement,unexpected observations, or limitations and opportunities. Finally, the researchershould be meticulous in record-keeping.Preciousdata maybe lost whenentrusted to memoryor not organized as soon after collection as possible. This is particularly important in multiple case designs where, as time passes, the details of various sites tend to run together. The researcher’sgoal shouldbe to collect data in such a waythat anotherresearchercould pick it up and immediatelyunderstandit and work withit.

Dataanalysis andexpositionTheanalysis of case data dependsheavily on the integrative powersof the researcher. Using multiple methodsof data collection, however, offers the opportunityfor triangulation and lends greater support to the researcher’s conclusions. Workingwith a researchpartner mayalso provide invaluable assistance. Two researcherscan capture greater richness of data and rely moreconfidently on the accuracy of the data. The key elements of data analysis are also critical to the written results of caseresearch. As muchas possible, the contextual and data richness of the study should be presented, and a clear chain of evidenceshould be established. The researcher’s reasoning in establishing causeand effect or drawingout hypotheses should be clearly stated and defended. The research should move from objectives and questions, to assumptionsand design choices, to specific data uncovered, and finally, to results and conclusions.Readers shouldbe able to follow this path readily.

A Critique of Case-Based Research To gain an understandingof the nature and quality of case research on IS, wesurveyed

CaseResearch Strategies

the following journals and conferencepro-’ 3 for the period January1981to Deceedings cember1985: Communicationsfor the ACM, The Proceedingsof the International Conference on Information Systems,Information and Management, Quarterly, and Systems,Objectives, Solutions. After excludingthe two categoriesof application descriptionsandaction researcharticles, we identified only five case papers in the Communicationsof the ACM,an average of one per year. Therewere few case papers in Information and Management; even if we include the action research and application description articles in this journal, caseresearchmakesup only 10%of all articles. The Proceedingsof the International Conference on Information Systemsincluded seven case papers, again about one per year. Finally, about10%of the articles in the MISQuarterly were case studies. Compared to these journals, Systems,Objectives, Solutions had a substantial proportion of caseresearcharticles (about25%of the articles published).This journal publishedcase research articles in each issue and encouragedresearchersto submit case studies. After four years as an independentpublication, the journal mergedwith Information and 4 Management in 1985. Weapproached the critique of the articles in our samplein two parts. First, welooked at four casestudies in detail andevaluatedtheir strengths and weaknesses.Then, we rated the whole sampleof case studies based on our guidelines for conductingcase studies.

Fourcaseresearchstudies Wechosethe following four casestudies becauseeachtook a different approachto investigation andthey illustrate both the strengths and weaknessesin case study research. 3 In ourestimation,thesewerethe IS journalsmost likely to publishcase-type articles.Ourintentionwas to provideexamples of casestudies,notto doanexhaustive search. 4 Hamilton andIves[18], based ona sample of IS articles from15journalspublishedbetween 1970-79, observedthat 14%of these werecasestudies. Based ona similar sample between 1977-83,Vogel ¯ andWetherbe [52] reportthat about17% of the publishedIS articleswerecasest,udies.

Markus: IS Implementation-Markus [32] examinedthe use of a "production planning and profit analysis system"in two manufacturing plants within the samedivision of a company. The syste m was readily accepted in oneplant, but wasat first stronglyrejected in the other (eventually it wasput into use). Her case study attemptedto find the reasonsfor the contradictory outcomesof implementation. Markusstated that an explanationbased on user participation could only partially account for these findings. Sheobservedthat the plant that acceptedthe systemexhibited a higher degreeof user participation than the onethat rejected the systeminitially. But how does one accountfor the later reversal? Markusproposeda "distribution of power" modelto explain the different reactions to the system. That is, the lack of consonancebetweenthe distribution of powerimplied by the systemand the distribution of powerwithin the organization causedthe failure of one plant to adoptthe system.Basedon this same model, Markusalso explained why one systemwasrejected at first, but waseventually, acceptedwhenthe organizational conditions that led to the initial rejection werechanged. Markus’research had to be carried out in a natural setting sinceit tracedthe evolutionof the system’simplementationfrom rejection to acceptanceand the reasons for the switch. Markuswasinterested in answeringa "why" question: Whywas the system used in one plant and not in another?The fact that both units were within a single company increased the internal validity of the casestudy. A large numberof possible causes for the acceptance in one plant and not in the other could be eliminated since the sites sharedthe same organizational setting and companyhistory. Thus, Markustook advantageof a unique opportunity to study an implementationissue. Since the opportunity was unique, Markus addedto our knowledgeabout the implementation of systems,eventhoughsomeprevious researchon the sametopic existed. The history of the implementationprocess, which spanneda period of six years, wasdescribed in detail. Oneflaw in the study though,wasthe total lack of detail aboutthe data collection methodology. The reader could only infer that interviews took place, since direct quotations from the people in-

MISQuarterly~September1987 375

CaseResearch Strategies

volved were usedto support the author’s arguments.Asidefromthis, it is not clear what, if any, additional data sourceswereused for triangulationandvalidation. Thisis of particular concernbecausethe history of the project covered six years and the memoriesof the participants mayhave beeninconsistent. Dutton: Adoption of a Fiscal Impact ModeI--Dutton’s [12] study of the City of Tulsa’s adoptionof a fiscal impactmodelandits eventual rejection was outstanding among casestudies welookedat for the rich level of detail it provided about the implementation process. According to Dutton, the sequence of events wasreconstructed basedon newspaper accounts, governmentreports, documents, memos,over 20 lengthy unstructured interviews with participants having different perspectives on implementation, and telephonecalls that precededand followed site visits. Asa final checkfor accuracy,a draft of the paper was sent to several respondents and important participants whohad not previously beeninterviewed. The purposeof the study wasto examinethe limitations of both the technical andorganizational perspectives of the implementation process. Dutton stated that his study contributed to our understandingof the process of innovationrejection, a topic weknowlittle about. Following a 16-pagedescription of the events, the organizational and political environment, and the actors associatedwith the implementation process, Dutton explained howtechnical, organizational, and interpersonal factors influenced the implementation process. However, he concluded that the greatest determinant wasthe political environment; he describedin detail howthe organizational and technical factors were contingent on the political ones. Dutton described events in such detail that readerscould maketheir own,different interpretations if they wished.It is an exploratory paperthat reveals an important factor in the implementationprocess. Onecriticism of the study is that Dutton did not clearly define his original researchobjectives. Further, wedo not knowwhyhe chose this particular site. Did he knowaboutthe outcome(failure of system) before he analyzed retrospectively the eventsthat took place?If this is the case, then his original objective

376

MISQuarterly~September 1987

might have beento explore whya failure had occurred. Alternatively, washis intention to study the implementationof computer-based systemsin city governments and describe the factors influencing the outcomes? In the latter case, the investigator wasin the discovery stage; in the formercasehe wasattemptingto find the factors that causeda particular outcome.Since these concernsmight affect the way the researcher approacheddata collection and thoughtaboutthe issues in the case, webelieve that a clear statementof the initial objectives of the researcher should have beenprovided in the study description. Pyburn: Strategic IS Planning-Pyburn [39] investigated IS strategic planning processes that wereunderwayin several companies,examinedthe businessand technical context in which these plans were developed, and generated preliminary conclusionsaboutthe success or failure of the planning activity. He wantedto understandwhya particular methodologyworkedwell in somesettings but not in others. Theimpetusfor the study wasthe growingimportanceof suchplans in the successof the overall IS effort, coupledwith the lack of effectiveness of existing IS planning methodologies. The method used was a comparative case study which, as described by Pyburn,"relies on the fact that outcomes in the different sites were the results of identified differences in those factors measuredfor its conclusions" [39, pp. 4-5]. Therefore,he choseeight sites that were as similar as possible basedon a number of characteristics,i.e., a literal replication. Someof these characteristics, such as "the companyhad formal business planning for five years" and "top managers were willing to committime to assist the research," madeobvious sense. However, the reasons for other characteristicsof the site selection, such as "the companieswere dominatedby a founding family" and "had corporate headquarters in relatively small cities," werenot immediatelyevident. It appearedthat Pyburn was more concerned with choosing highly similar sites, rather than choosinga model group that wasrepresentative of companies involved in strategic IS planning. In eachsite, Pyburnconductedindepth interviewswith the senior IS executiveand the top management team(four to six individuals). Heused a series of questionsto gather data

CaseResearch Strategies

concerningthe nature of the business, the factors critical to its success,the company’s IS planningpractices, andthe extent to which IS wasaddressingthe critical needsof the firm. In orderto increasethe reliability of the data collection processand to reduceintervieweebias, prior to the first interview, headministered a questionnaire that contained items similar to the questions subsequently posedin the interviews. The questionnaire data were analyzed using nonparametric statistical methods.For the interview data, a casedescription waswritten for eachorganization. For a given topic, a side-by-side presentation of all comments madeby each interviewee for each topic was provided. (However, the case descriptions and commentswere not included in the publishedpaper and are presumablypart of Pyburn’sdoctoral thesis on whichthe article wasbased.) Basedon the interview data andpartially supported by questionnaireresults, Pyburnidentified three IS planning styles: personalinformal, personal-formal,andwritten-formal. Hethen determinedthe degreeof successof IS strategic planning in eachcompany.Since none of these planning approachesappeared to be uniformly successful in all the companies in which they were implemented,Pyburn pursueda contingencyanalysis. A number of factors, suchas the perceivedstatus of the IS managerand the complexity of the IS environment,wereidentified as important influenceson the successor failure of the planning process.Finally, all of this information wastied togetherandthe critical factors in the successof a particular planning style were identified. For example,a personal-informal style dependedon both an informal general managementstyle and high IS manager status. Pyburnset forth clear researchobjectivesprior to data collection. Thisled to a focusedapproachto the interviews and site selection, andthe useof triangulation in data collection. Hecarefully explainedeachstep of a logical processthat eventually culminatedin a preliminary typology of strategic MISplanning. Theuse of multiple sites allowed him to attempt a contingencyanalysis. This is a good exampleof an exploratory case study that proposed a classification schemefor MIS planning that could be further refined and tested in other studies.

A drawbackof this workwasthat the type of detailed descriptions found in the Dutton and the Markuspapers were not provided in this paper. Of course, the journal mayhave imposedlength limitations that made it impossible to presentthe data collected from eachof the eight sites. Booksor monographs might be better vehicles to publish case study research. Nevertheless,the reader has to rely on the author’s interpretation and cannotinterpret the data independently. Olson: Centralization of the SystemDevelopmentFunction--Olson [37] examined the issue of whether the systems developmentfunction shouldbe centralized or located in the user organization. Basedon an analysis of the literature, Olsoncontendedthat there is no best wayto organize the system developmentfunction; the decision depends on factors outside the IS function. Shetherefore conductedan in-depth study of two organizationsin orderto identify the factors that influence the organization of the systemdevelopmentfunction and the quality of the developmentprocess. She conductedextensive interviews with IS managers and key participants in the systemdesign process, and administeredquestionnairesto users as well as those whowerepart of the designeffort. Two computer-based systems (one an accounts receivable system, the other a benefits system) were selected from two large multidivisional companies.In one case, the systems development function wascentralized; in the other it wasdecentralized to businessdivisions. Olson’sresults showedthat in the decentralized developmentgroup users had a higher level of informationsatisfaction, usersparticipated less in the design of the system, and analysts wereless satisfied with their jobs than in the centralized group. Theseresults were baseduponvery small samplesof users and analysts in eachorganization, a fact acknowledgedby Olson. Shesuggestedseveral conclusions about the effectiveness of user participation in the development process. Olson chose her sample to bring out the differences associated with alternative organizations of the systemdevelopmentfunction. Theviewsof the various participants in the development process were sought and data were collected by multiple means.Furthermore, recently completed systems were

MIS Quarter/y/September 1987 377

CaseResearch Strategies

chosento minimize the problemsassociated with retrospective data collection. A strength of the paperwasthe author’s attemptto provide explanations for the outcomesbasedon what was observed from the cases. The IS departmentsof the two companiesand their activities weredescribedin reasonabledetail, but little wasoffered to the reader aboutthe companies themselves.Finally, since the outcomeswere based only on a sample of two companies, this study could be considered only a basic exploratory casestudy.

study methodology is well-suited to identifying key eventsand actors and to linking them in a causal chain. Examplesof case study topics are: the impactof organizationalstrategyon the IS organization’sstructure[45], the impact of IS on organizational change[41, 42], the impact of technology on personnel [28], the influence of technologyon organizational communications [15], the factors affecting the successof end-userdevelopedapplications [40], and the role of users in DSS development [31 ].

The characteristics of the four case studies discussed above, as well as a sample of othersidentified in our survey,are outlined in Table3. It is clear that the casestudies have both similarities and differences. For example, the Pyburnand Dutton studies were both exploratoryin nature, but differed in termsof sample selection and data gathering methods. Onecommon feature of all four studies, whichis not evident in the larger sampleoutlined in Table 3, wasthe detailed explanations that accompanied the findings. An investigator conductinga field study or experiment mainly relies on theory or a priori reasoningto deductively arrive at the outcomes.In these four studies, the investigators collected data, distilled the evidence,and inductively developedcausal links to explain particular outcomes.While on the positive side the explanations were groundedin observedfacts, our concernis that these facts werefiltered throughthe subjective lensesof the investigators.

ResearchObjectives In the published case research we surveyed, the objective of the study wasseldomclearly specified. Amongthe exceptions were Pyburn [39] and Hirschheim [20], who stated that their objectives were to describe and explore a phenomenon that was not wellunderstood.

An overall evaluation of case studies In this section, wewill describe the nature and general quality of case research in IS. We’ll evaluatethe casestudies in our survey basedon our guidelines for conducting case research. Research Themes The predominant themein the case studies wasimplementation, that is, the possible causesof the successor failure of an information or decision supportsystem(e.g., [11]). Since the process of implementation takes place over time, is a complexprocessinvolving multiple actors, and is influenced by events that happen unexpectedly, a case

378

MISQuarterly~September 1987

Wewould characterize most of the case studies as exploratory in nature. They described the context in which an intervention occurredand the intervention itself. For example, Kraemer[27] and Dutton [12] described the implementation of model-basedsystemsin the public sector. Almostall of these exploratory studies concludedwith a list of suggestionsto improvethe successof future implementation efforts. It wasdifficult to determine if the researcherswere at the same time attempting to generate hypotheses. An exceptionis the study by Ives and Olson[21] which basically describedthe nature of an IS manager’sjob. Somecase studies pursued an explanatory strategy by first describing the events that took place and then presenting multiple competing theories to explain the courseof events (e.g., Franzand Robey[14], Kling and lacono [26], Markus[33]). Eachof these studies used a sample of one case. For example, Markus [33] evaluated three theories of resistance: peopledetermined, system-determined, and interaction-determined, and concluded that the interaction theory did a better job of explaining the causesof resistanceto a system. Someother studies also followed an explanatory strategy by testing hypothesesderived from a single theory (or a priori reasoning). White[55] examined the influence of the cognitive style compositionof a project teamon

Table 3. An Illustrative Categorizationof a Sampleof CaseStudies Level of Description About Units

Theme

Research Thrust Sample Selection

Data Collection Units of Analysis Method

OIson [37]

Centralization/ decentralization of system development

Exploration

2 IS groups, differing on the degreeof centralization of system development

Group

Interview and questionnaire

High

Markus [32]

Implementation

Explanation

2 plants within samecompany differing on impleme_ntation success

Organizational subunit

Unspecified, inferred to be interviewed

High

_Pyburn[39]

IS strategic planning

Exploration

8 companies chosen based on similar characteristics

Organizations

Interview and questionnaire

Low

Dutton [12]

Implementation of models in the public sector

Exploration and explanation

1 U.S. city

Organization

Multiple sources, high degreeof triangulation

Very High

Hirschheim[20]

Participative systemsdesign

Exploration

20 individuals from 8 organizations with experiencein participative design

Individuals and organizations

Interviews

Low

White [55]

Cognitive styles of MIS project teams

Explanation

2 MIS project teamsin the samecompanydiffering in the cognitive style compositionof their members

Groups

Interviews

Low

Ives and Olson [21]

Nature of IS manager’sjob

Exploration

6 IS managers chosen based on several common criteria

Individuals

Observations of each managerfor 3-4 days

High, but at aggregate level

Keen, Bronsema, and Zuboff [24]

Implementation of common systemsin an international banking setting

Exploration and explanation

Banksin 9 countries (part of an international bank) selected on a numberof criteria

Organizational subunits

Interviews

High

Fulk and Dutton [15]

Effect of videoconferencing on organizational communication

Exploration

2 organizations, newuser, one localized experienced user

Organization

Interview

High

Franz and Robey [14]

User-led system design

Exploration and explanation

1 IS development team

Group

Longitudinal and multiple sources

High

CaseResearch Strategies

its performance,and Schonberger [45] tested the hypothesis that organization strategies will influencethe structure of the IS organization. Although Schonbergerclearly stated that he first observedtwo organizations with different IS departmentstructures and then got interested in testing the influence of strategy on structure basedon Chandler’s[7] theory, White appearedto be moreambiguous.It is not clear if shefirst observed twoproject teamswith different cognitive styles and decidedto test a hypothesis,or if sheintended to test a hypothesisand then searchedfor an ideal sampleof project teamswith different compositions.However,wedo not wish to single out her workfor criticism since this was typical of the casestudies in our sample. In only oneinstance (Keen,et al., [24]) were the authors’ stated objectives both the exploration and testing of the explanatorypowerof theories. Twoorganizational changeparadigms (the Lewin-Schein and Kolb-Frohman models)were used for theory testing. These provideda priori explanationsof majorfactors that influence the successor failure of implementationefforts. Theorytesting wasdone both by goodness of fit and by counterexamples,i.e., by the prediction and disconfirmation strategies suggested by Bonoma [5]. In summary, we found that the IS case researchers we surveyed did not provide clear descriptionsof wheretheir topics fit in the knowledgebuilding process. They thus did not justify their researchpurpose(drift, description, exploration and explanation) and did not allow readersto judge their workon a moreinformed basis.

Unit of AnalysisandSite Selection Theunit of analysis for a casestudy, and consequentlyfor the selection of a particular site to study, wasnot providedin manyof the published works. This is a problem consistent with the lack of clear researchobjectives discussedabove,and probably an outcomeof it. Several of the casesexamineda single company or a subunit. Since these typically did not represent critical, unique or revelatory cases, they were presumably chosen based on availability and evidently the researchers’ goals were to conduct exploratory case studies.

380

MISQuarterly~September 1987

Markus’[32] study previously describedwasa single-case sample containing two embedded units of analysis (two plants within the samedivision). It wasrare becauseof the unique outcomesof the implementationprocess(oneplant hada success,the other a failure with the samecomputerizedsystem) and the unusualinsights it offered. Basedon a large numberof criteria, Keen,et al. [24] selectedninesites (countries)affiliated with an international bank. Thecriteria included IS development strategy, pace, focus, perceived ease of implementation, geographical dispersion, the size of the country, and the complexity of the operation. Twoof the sites, which differed the most in terms of criteria, wereselected for indepth analysis. Even though this study appeared to be a single-case, embedded unit analysis, it could be considereda multiple-case design, due to the decentralizednature of the sites. A study by Ein-Dor and Segev[13] measured the relationship betweenthe successof an IS and the perceived importance and investmentin that system. They examined10 subsystemssupportedby a logistics IS in a large organization in Israel. This wasan embedded case study design in which the researchers conductedquantitative analysis of a large numberof subunits. However,a problem of embeddeddesigns that Yin [56] mentioned and this study appearedto have, is that it makesthe subunitsthe sole focus of the study and ignores the context--the characteristics of the organizationas a whole.Therewasless focus on the context and morefocus on a few aspects of the subunits. Thus, this study seemedto fall somewherebetween a case and a field study, but had the strengths of neither. The sites for someof the multiple-case research studies were chosen based on the . study’s researchthemeand objective. For example, to measurethe effects of centralization, Olson[37] chosetwo sites--one in which the system developmentfunction was centralized andone in whichit wasdecentralized. To examinethe influence of strategy on structure, Schonberger[45] observedtwo organizations that haddifferent IS department structures. The sites were chosenpartly because they differed in termsof the outcomevariable of interest, andpartly on an opportunistic basis. Similarly, in their paperonfactors affect-

CaseResearch Strategies

ing user involvement in DSSdevelopment, Mannand Watson[31] presented three cases that demonstratedsubstantially different degrees of user involvement. However,the reader is not told whether .the three cases were chosenfrom a larger samplebecauseof their uniquecharacteristics. In other instances, the researcher sought a homogeneoussample with as manysimilar characteristics as possible [39]. However,as we mentionedin our discussion of Pyburn’s study [39], it wasnot obviouswhythese particular characteristics were chosen. In any case, the investigator did not knowin advance what independent variables would prove to be important. The idea wasto minimizethe extraneous variablesso that if significant factors wereindeedfound, there would be a high degreeof confidencethat only those factors causedthe observeddifferences. In contrast to Pyburn’sapproach,Fulk and Dutton [15], in their study of the impact of videoconferencingon organizational communications, chosetwo companies with different characteristics. Onewasa large aerospace companywith offices on the east and west coasts that had recently started using videoconferencing. The secondwas a large westernpublic utility with severalyearsof routine utilization. Noneof the multiple-casestudies clearly stated the site selection objectives, i.e., whether the investigatorpursued a literal or a theoretical replication. Nevertheless, in most instances,the rationale for choosinga particular site combinationwasstated. Sometimesthe reasons for site selection were not easy to infer. For example,Robey (1983) studied eight companieslocated different countries. Eventhoughthe samples appearedto be from differing organizations, the researcherdid not explicitly state whether a literal or theoreticalreplicationwasthe goal. In one study [28] the three caseswerechosen for a variety of disparate reasons.Onecompany waschosenbecauseone of the authors wasinvolved in the implementationof a system there. In the secondcompany,2 out of 17 branches were selected becausethe implementation of a system went smoothly in one and was"less successful" in the other. The third company wasincluded in the samplebecauseit hada recent, majorshift from a manual system to a computerizedsystem and it

wasself-contained. It thus appearedthat the first casewaschosenopportunistically, the second due to the observed outcomes, and the third basedon a possiblecausalfactor. In this particular study, it wasdifficult to infer whetherthe authors’ goal wasa literal or theoretical replication, or whetherthe caseswere chosenfor exploratoryor illustrative reasons. Authorsin our sampledid not indicate if their case studies were part of systematic/programmatic research plans. Most seemedto be stand-alone, one-shot studies. Only one casestudywasa triangulation in that it wasa follow-up to a survey[15]. Anothercasestudy waspart of a large-scaleeffort to studyvarious aspectsof IS and their organizationalimpact, but the linkage to the larger study was not describedin the paper[41]. Data Collection In about half of the case studies, the data were collected by multiple means;the other half relied solely on interviews. In two case studies, though, the data collection method wasnot specified at all. Webelieve that a clear description of data sourcesand the way they contribute to the findings of the research is an importantaspectof the reliability andvalidity of the findings. Yin’s [56] suggestions aboutdescribing the case study protocol and having a case study database can serve as important guidelines. Almostall of the studies used interviews for data collection. Theinterview questionswere rarely specified and, whenthey were,it wasin a very general form. Hirschheim[20], a notable exception, included the questions used in his semistructuredinterview in an appendix. In somestudies, researchersinterviewed individuals whohad different perspectiveson a given process, e.g., managers,users, and designers. Sometimesthe researchers mentioned that they used documentsand observations, but they did not provideany moredetail aboutthem. A few studies of implementationin the public sector collected data from a large numberof sources[25, 27]. Thesestudies weresimilar to the Dutton[12] studyin the richnessof their descriptions and data sources. The only study that useda longitudinal methodology is also an exemplaryeffort of data collection [14]. To study user involvementin IS design,

MISQuarterly~September1987 381

CaseResearch Strategies

the investigators collected data over a twoyear period. In order to achievetriangulation, data were gathered through questionnaires, critical incident files, unstructured interviews, documentsand memoranda,observations at meetings and tape recordings. In this way, they attemptedto get both an objective view of eventsandthe subjective interpretations of participants. Keen,et al. [24] usedan outside expert with no knowledgeof the systembeing implemented to interview participants with whom the researchteam(authors) had already talked and whowerein conflict. This wasdoneto reduce any bias arising from the expectationsof the researchers.If other studies usedsuchtechniquesto increasethe objectivity of data collection, no mention wasmadeof them. Case data in the research surveyed was mostly qualitative. A few studies included quantitative observations,mostly in the form of questionnairedata [12, 22, 37]. In general, however,the degreeof detail about data collection methodswasnot very revealing, a substantial problem with most of the case studies observed.

Concluding Comments Our intent in this article wasto clarify the nature of the case researchmethod,explain whyit might be utilized in IS studies, survey its uses in IS research within the last five years, and offer somesuggestions for improvement. The case research strategy has mostly been used for exploration and hypothesis generation. This is a legitimate wayof addingto the bodyof knowledge in the IS field. Exploration is, however,not the only reasonfor applying the case method.As discussedearlier, various authors have suggestedthe use of cases for providing explanation and for testing hypotheses. Noresearchstrategy is better than all others. Unlike someof the contributors to Mumford, et al. [36], wedo not advocateexclusivereliance on case/action research methods. The selection of a research strategy dependson the current knowledgeof a topic and the nature of the topic, amongother factors. The casestrategy is particularly well-suited to IS researchbecausethe technologyis relatively

382

MISQuarterly~September 1987

newand interest has shifted to organizational rather than technical issues. For example, case studies have beenhelpful in identifying the causalchainthat led to the success or failure of an information systemby revealing in chronological fashion the various actors and events that influenced the final outcome. Severalcurrent topics within IS researchare amenableto the case study approach. The use of an expert systemfor management support isone such area. Since expert systems are just beginning to be introduced into organizations, a case study of companiesthat are rather far alongin the useof sucha system wouldprovide valuable insights. Sviokla [48] has examinedhowan expert systemthat assists financial plannersaffects their jobs andthe quality of the plansthey produce.This revelatory use of the case methodcan provide hypothesesabout the impact of expert systemstechnology on organizations. These hypothesescan then be tested using another researchmethod,like a field experiment. The relationship betweeninformation technology and corporatestrategy is anotherarea that could be explored further using a structured programof multiple casestudies. From the case studies conductedto date, there is evidence that somecompaniesuse information technologymoreeffectively as a strategic weaponthan others [38]. A systematic study of several companieswithin one industry could provide important insights into why somecompaniesuse information technology moresuccessfully than others. One of the more difficult decisions that researchers must makeis to determine when further case studies are neededin an emerging researcharea. For example,in the study of information technology and corporate strategy, BakosandTreacy[2, p. 107] believe that: As this area of researchmatures,there is an increasing need to movebeyond frameworks and toward explanatory models of the underlying phenomena. This type of research will allow us to build a cumulativetradition and to make normative statements to guide managerialactions. In the IS area, there is somemerit to both the view that morecase research is neededand

CaseResearch Strategies

the view that quantitative techniquesshould be employed.Case studies can provide the organizational context for the study of the relationship betweenstrategy and information technology.This is importantevenin this stage of the research. A formal model can provide significant insights into the more quantitative aspectsof the issue, suchas the economicimpact of investing in information technology. Recently there have been calls for a more detailed understanding of phenomena in the IS area. Thereis an interesting parallel betweenour call for higher quality caseresearch and the argument presented by Todd and Benbasat[49] for increased use of protocol analysis in the study of decision supportsystems. The use of protocol analysis in IS studies has increased within the last few years becausemanyresearchersbelieve that they haveto openthe "black box." Theywant to examineprocessesin order to better understandthe effects of information technology on the peoplewhoworkwith it, andtheir influenceon the technology.Theintent is to carry out a detailed, indepth examinationof a small numberof individuals. In a similar way, case research obtains detailed data about one or a few units. The key difference between the case researchmethodand protocol analysis is that, whileboth methods requirelittle a priori specification of dependent variables and their measurementallow the investigator to use a high degreeof discretion in structuring and interpreting the data, the casemethodalso offers little control over the antecedentcondition, e.g., independentvariables. In this article weidentified a number of problems that were common to most of the case research studies in the sample. Someof these might be alleviated by asking the authors to provide moreinformation about their researchobjectives and researchplans. However,it appearsto us that, in manyinstances, the investigators had not considered someof the methodologicalissues. In general, the objectives of the researcherswerenot clearly specified. Thereasonsfor selecting single-case versus multiple-case designs werenot explainedand the choiceof particular sites wasnot tied to the designapproach. In manycasesthe data collection methodwas ambiguousand details were not provided.

Theuseof triangulation to increasereliability wasrare. Yin states that the reader of the casestudy shouldbe able to follow the derivation of any evidencefrom initial researchquestions to the conclusionsof the study. This chainof evidencewill improvethe reliability of the data. Thepoint is that a casestudy should be more than an exercisein storytelling or an opinion piece; it should adhereto certain rules of procedure,as describedin Yin [56] and discussed here. Although these proceduresare not as detailed as one wouldfind in a field study, surveyor experiment,they are critical to allowing readersto assessthe reliability andvalidity of the study’sfindings. In this way, IS researcherscan better contribute to the knowledgebuilding process and IS case researchwill comeinto its own.

Acknowledgement Professor Benbasatwasa Marvin BowerFellow at the Harvard Business School while workingon this paper. Hethanks the Fellowship program for its support during the 1985-86 academicyear.

References 1. Allison, G.T. Essenceof Decisionmaking: Explainingthe Cuban Missile Crisis, Little, Brown, Boston, Massachusetts1971. 2. Bakos, J.Y. and Treacy, M.E. "Information Technologyand CorporateStrategy: A ResearchPerspective," MISQuarterly, Volume10, Number2, June 1986, pp. 107-119. 3. Benbasat, I. "An Analysis of Research Methodologies,"in The Information Systems Research Challenge, F. Warren McFarlan(ed.), HarvardBusinessSchool Press, Boston, Massachusetts,1984, pp. 47-85. 4. Bonoma,T.V. "A CaseStudy in CaseResearch: Marketing Implementation," Working Paper 9-585-142, Harvard University GraduateSchool of Business Administration, Boston, Massachusetts, May 1983. 5. Bonoma,T.V."Case Researchin Marketing: Opportunities, Problems,and a Process," Journal of Marketing Research, Volume 22, Number2, May 1985, pp.

MIS Quarterly~September1987 383

CaseResearch Strategies

199-208. Bouchard,T.J., Jr."Field ResearchMethods," in Industrial andOrganizationalPsychology, M.D. Dunnette (ed.), Rand McNally,Chicago,Illinois, 1976. 7. Chandler, A.D., Jr. Strategy and Structure: Chaptersin the History of the American Industrial Enterprise, MIT Press, Cambridge, Massachusetts, 1962. Checkland, P. Systems Thinking, Systems Practice, John Wiley and Sons, New York, NewYork, 1981. Christenson,C. "Proposalsfor a Program of Empirical Researchinto the Properties of Triangles," Decision Sciences,Volume 7, Number3, October 1976, pp. 631-648. 10. Cook, T.D. and Campbell, D.T. Quasiexperimentation:Designand Analysis Issues for Field Settings, RandMcNally, Chicago,Illinois, 1979. 11. Dickson, G.W.and Janson,M. "The Failure of a DSSfor EnergyConservation: A Technical Perspective," Systems,Objectives, Solutions, Volume4, Number2, April 1984, pp. 69-79. 12. Dutton, W.H. "The Rejection of an Innovation: The Political Environmentof a Computer-BasedModel," Systems, Objectives, Solutions, Volume1, Number4, November1981, pp. 179-201. 13. Ein-Dor, P. and Segev,E. "PerceivedImportance, Investment. and Successof MIS, or the MIS Zoo," Systems, Objectives, Solutions, Volume4, Number2, April 1984, pp. 61-67. 14. Franz, C.R. and Robey,D. "An Investigation of User-LedSystemDesign: Rational and Political Perspectives," Communications of the ACM,Volume27, Number12, December1984, pp. 1202-1217. 15. Fulk, J. and Dutton, W. "Videoconferencing as an OrganizationalInformation System: Assessing. the Role of Electronic Meetings," Systems, Objectives, Solutions, Volume4, Number2, April 1984, pp. 105-118. 16. Gibson, C.F. "A Methodology for ImplementationResearch,"in Implementing Operations Research/ManagementScince, R.L. Schultz and D.P. Slevin (eds.), AmericanElsevier, NewYork, NewYork, 1975. 17. Ginzberg, M. "A Prescriptive Modelfor System Implementation," Systems, Objectives, Solutions, Volume1, Number 1, January 1981, pp. 33-46. 6.

384

MISQuarterly/September1987

i8. Hamilton, S. and Ives, B. "MIS Research Strategies," Information and Management, Volume 5, Number 6, December 1982, pp. 339-347. 19. Hersen, M. and Barlow, D.H., SingleCaseExperimental Designs: Strategies for Studying Behavior, Pergamon Press, NewYork, NewYork, 1976. 20. Hirschheim, R.A. "User ExperienceWith an Assessmentof Participative Systems Design," MIS Quarterly, Volume9, Number 4, December1985, pp. 295-304. 21. Ives, B. and Olson, M.H. "Manageror Technician? The Nature of the Information SystemsManager’sJob," MISQuarterly, Volume5, Number4, December 1981, pp. 49-63. 22. Jick, T.D. "Mixing Qualitative and Quantitative Methods: Triangulationin Action," Administrative SciencesQuarterly, Volume 24, Number4, December1979, pp. 602-611. 23. Kaplan, R.S. "The Role of Empirical Research in ManagementAccounting," Working Paper 9-785-001, Division of Research, Harvard Business School, Boston, Massachusetts,1985. 24. Keen, P.W., Bronsema,G., and Zuboff, S. "Implementing CommonSystems: OneOrganization’s Experience," Systems, Objectives, Solutions, Volume2, Number3, August 1982, pp. 125-142. 25. King, J. "Successful Implementationof Large-Scale DSS,"Systems,Objectives, Solutions, Volume3, Number4, November 1983, pp. 183-205. 26. Kling, R. and lacono, S. "TheControl of Information SystemsDevelopmentsAfter Implementation," Communications of the ACM,Volume27, Number12, December 1984, pp. 1218-1226. 27. Kraemer,K.L. "The Politics of ModelImplementation," Systems,Objectives, Solutions, Volume1, Number4, November 1981, pp. 161-178. 28. Land, F., Detjejarvwat, N. and Smith, C. "Factors Affecting Social Controls: The Reasonsand Values--Part I1," Systems, Objectives, Solutions, Volume3, Number 4, November1983, pp. 207-226. 29. Lawrence,P.L. and Lorsch, J.W. Organization and Environment, Harvard Business School Press, Boston, Massachusetts, 1986. 30. MacKenzie, K.D. and Howe, R. "Paradigm Developmentin the Social Sci-

CaseResearch Strategies

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

41.

42.

ences: A ProposedResearchStrategy," Academyof ManagementReview, Volume3, Number1, 1978, pp. 7-23. Mann,R.I. and Watson,H.J. "A Contingency Model for User Involvement in DSSDevelopment,"MIS Quarterly, Volume 8, Number 1, March 1984, pp. 27-38. Markus, M.L. "ImplementationPolitics: Top ManagementSupport and User Involvement," Systems,Objectives, Solutions, Volume1, Number4, November 1981, pp. 203-215. Markus, M.L. "Power, Politics, and MIS Implementation," Communications of the ACM,Volume26, Number6, June 1983, pp. 430-444. Morton, M.S. Management Decision Systems: Computer-Based Support for Decision Making, Harvard Business School, Division of Research, Boston, Massachusetts, 1971. Mumford,E. "Participative SystemsDesign: Structure and Method," Systems, Objectives, Solutions, Volume1, Number 1, January1981, pp. 5-19. Mumford,E., Hirschheim,R., Fitzgerald, G. and Wood-Harper,T. ResearchMethods in Information Systems,North Holland, NewYork, NewYork, 1985. Olson, M. "User Involvement and Decentralization of the Development Function: A Comparison of Two Case Studies," Systems,Objectives, Solutions, Volume 1, Number2, April 1981, pp. 59-69. Parsons, G.L. "Information Technology: A New Competitive Weapon," Sloan ManagementReview, Volume 25, Number 1, Fall 1983,pp. 3-14. Pyburn, P.J. "Linking the MISPlan with Corporate Strategy: An Exploratory Study," MISQuarterly, Volume7, Number 2, June 1983, pp. 1-14. Rivard, S. and Huff, S.L. "User DevelopedApplications: Evaluation of Success From the DPDepartment Perspeqtive," MIS Quarterly, Volume 8, Number1, March1984, pp. 39-50. Robey, D. "ComputerInformation Systems and Organization Structure," Communications of the ACM, Volume 24, Number10, October 1981, pp. 679-686. Robey,D. "Information Systemsand Organizational Change: A Comparative CaseStudy," Systems,Objectives, Solu-

43.

44.

45.

46. 47.

48.

49.

50. 51.

52.

53.

54.

tions, Volume3, Number 3, August1983, pp. 143-154. Rockart, J.F. and Flannery, L.S. "The Managementof End-User Computing," Communications of the ACM,Volume26, Number10, October 1983, pp. 776-784. Roethlisberger, F.J. The Elusive Phenomena,Harvard Business School, Division of Research, Boston, Massachusetts 1977. Schonberger,R.J. "Strategy and Structure: A Tale of TwoInformation Systems Departments,"Systems,Objectives, Solutions, Volume1, Number 2, April 1981, pp. 71-77. Stone, E. ResearchMethodsin Organizational Behavior, Scott, Foresmanand Company, Glenview,Illinois, 1978. Susman,G.L. and Evered, R.D. "An Assessmentof the Scientific Merits of Action Research,"Administrative Sciences Quarterly, Volume23, Number4, December 1978, pp. 582-603. Sviokla, J.S. "Planpower, XCON,and Mudman: An In-Depth Analysis into Three Commercial Expert Systems in Use," unpublisheddoctoral dissertation, Harvard Business School, Boston, Massachusetts, 1986. Todd, P. and Benbasat,I. "ProcessTracing Methodsin Decision Support Systems Research: Exploring the Black Box," WorkingPaperNo. 1140,Faculty of Commerce and Business Administration, TheUniversity of British Columbia,Vancouver, British Columbia,Canada,January 1986. Towl, A.R., To Study Administrations by Cases, Harvard Business School, Boston, 1969. VanMaanen,J. "Introduction," in Varieties of Quafitative Research,J. VanMaanen (ed.), Sage Publications, Beverly Hills, California 1982,pp. 11-29. Vogel, D.R. and Wetherbe,J.C. "MIS Research:A Profile of LeadingJournalsand Universities," Database, Volume16, Number1, Fall 1984, pp. 3-14. Webb,E., Campbell, D., Schwartz, R. and Sechrest, L. Unobtrusive Measures: Nonreactive Research in the Social Sciences, RandMcNally, Chicago, Illinois, 1966. Weick, K.E. "Theoretical Assumptions and ResearchMethodologySelection" in

MIS Quarterly~September1987 385

Information Research Planning

The Information SystemsResearchChallenge, F. WarrenMcFarlan(ed.), Harvard Business School Press, Boston, Massachusetts, 1984, pp. 111-132. 55. White, K.B. "MIS Project Teams:An Investigation of Cognitive Style Implications," MISQuarterly, Volume8, Number 2, June 1984, pp. 95-101. 56. Yin, R.K. CaseStudy Research, Design and Methods,SagePublications, Beverly Hills, California, 1984.

About The Authors Izak Benbasatis Professor and Chair of the MISDivision at the University of British Columbia. Hereceivedhis Ph.D. in MISfrom the University of Minnesotain 1974. During the 1985-86 academic year he was a Marvin Bower Fellow at the Harvard Business School. Heis currently studying the topic of knowledgeacquisition for expert systemsdevelopment, supported by a grant from the Natural Sciencesand Engineering Council of Canada. David Goldstein is an Assistant Professorof BusinessAdministration at HarvardUniversity. He received his Ph.D. in Management

386

MISQuarterly~September 1987

Science from Sloan School of Management at M.I.T. Goldstein arrived at the HarvardBusinessSchoolin 1984and has taught control in the MBAprogram and MIS in the MBAand Executive Education programs.His research interests are on the impact of information technologyon the effectiveness of managers, the management of the systems development function, andthe useof qualitative methods in MISresearch. MelissaMeadjoined the information systems faculty at the Harvard Business School in 1985. Her teaching at Harvard includes courses in both control and management information systems,in addition to knowledgebased systemsand doctoral seminars on research methodsand topics. Professor Mead completed a Ph.D. in ManagementScience and Information Systemsat TheUniversity of Texasat Austinin 1985.Prior to her arrival at Harvard, Professor Meadworkedas a consultant for Analysis, Research& Computation Inc. Her researchinterests include the incorporation of information technologyas an elementof business strategy, particularly the use of information technology by general managers during times of transition to foster and sustain changesin strategic direction.