DataResource

ManagingThe Data Resource: A Contingency Perspective By: Dale L. Goodhue Carlson Schoolof Management Universityof Minnesota Minneapolis,MN55455 JudithA. Quillard Centerfor InformationSystems Research Sloan Schoolof Management Massachusetts Institute of Technology Cambridge,MA02139 JohnF. Rockart Centerfor InformationSystems Research Sloan Schoolof Management Massachusetts Institute of Technology Cambridge,MA02139

Abstract Today,corporationsare placing increasingemphasison the management of data. To learn more about effective approachesto "managingthe data resource,"casestudies of 31 data managementefforts in 20 diversefirms havebeenconducted. Themajorfinding is that there is no single, dominantapproach to improvingthe managementof data. Rather, firms have adopted multiple approaches that appearto be very diversein (1) business objective,(2) organizational scope,(3) planningmethod,and (4) "product," i.e., defiverableproduced.Thedominantbusinessobjective for successfulaction is improved managerialinformation; most data management efforts are "targeted"withouta formaldataplanning process;and the dominant productwas"information databases."In addition, several key organizational issues must be addressedwhen undertakingany data management effort. Keywords:Data administration, data management, information resource management, information systems management, strategic data planning ACM Categories:H.2.7, J.1, K.6.0, K.6.1

Introduction Althoughthe literature presentsboth conceptual justifications for managing data as a resource (Diebold, 1979; Edelman,1981; Horton, 1985) andapproaches/methods that describehowto do so (IBM,1981; Martin,1982,Ross,1981), it unfortunatelycontainslittle discussion of actualbusiness problemscausedby poorly managed data, of businesssuccessesmadepossible by wellmanaged data, or the data management actions that makea difference. Theresult maybe the impressionthat data management is a technologydriven conceptin searchof a concretebusiness need. Theproblemsof unmanaged data are, however, quite real andexist in a broadrangeof organizations. A majorbankseeking to shift its strategytowarda focuson customers finds that it cannot determinehowprofitable individual customers are, or evenwhatits total businessis with each customer,becauseits customercodesare not common acrossbranchesor lines of business.A manufacturing firm with nineplants cannotnegotiate favorablepurchasingagreements with its major suppliers,because it cannotpoolinconsistent datafromtheseplantsto find out howmuch it buys from eachsupplier. An insurancecompany discoversit cannotcheckgrouphealth insurance claimsagainstpreviousclaimsfor the same individual participant, because the structure of the data precludesit. A company attemptingto merge two divisionsfinds that incompatibilitiesin data definitions andsystems provideoneof the greatest obstaclesto attainingthis importantstrategic action. Thesefew examples mightbe dismissedas unfortunateabberationsif similar problems werenot foundin so many organizations today.In fact, it is likely that management’s demand for up-to-date informationfrom manydifferent sourcesat many different levels of aggregration will increaseeven further as the businessenvironmentbecomes morecompetitive (EDPAnalyzer, 1986). This makes it critical that businesses better appreciate the implicationsof poorlymanaged data, andunderstandthe variety of waysin whichtheycanimprove the management of data. The three-year study describedhere attempts to addressthis needby looking at somethirty data management effortsin 20firms. Themajorfindingfromthis studyis that thereis no single dominantapproachto improvingthe managementof data. Rather, firms have adopted multiple approaches, all of whichneedto beconsideredby anyorganizationstriving to leverage

MIS Quarterly~September1988 373

DataResource

dataas a valuablebusiness resource.In addition, this studyidentifiesa set of keyorganizational issuesthat shouldbe addressedwhenundertaking any data management effort. Therest of this section summarizes somecurrent approaches to data management in the literature and describes the research method.The next sectionpresentsthe majorfindings usinga frameworkthat representsthe variety andcontingency seenin the data management efforts studied.This is followedby a brief discussion of severalkeyorganizationalissuesthat managers shouldexplicitly address as they consider the options suggestedby the framework.The article concludes with a discussion of the majorlessons learned.

Current approachesto data management As the business needfor information has increased,so hasthe technicalcapability to handle information.A rapidly growingamount of data is nowavailable in electronic form--mostof it designed and organized to meet the needs of specific applications,with little thoughtgivento compatibilityof dataacrossapplicationsor businessfunctions. Managing this data in a manner that best contributesto businessobjectiveshas becomea complex problem (EDP Analyzer, 1986). Various solutions that havebeendiscussedin the existing data resourcemanagement (DRM)literature canbe categorizedinto three types of approaches: Approaches with a technical focus. Theseinclude tools and techniques such as database management systems (Curtice, 1986; Date, 1981),data dictionaries (Appleton,1987;Ross, 1981), and data entity-relationship modeling (Chen,1976;1983). Approaches with a focusonorganizationalresponsibilities.Theseinclude the establishment of organizationalunits suchas database administration anddata administration(Gillenson,1985; GUIDE,1977; Kahnand Garceau,1985), and the formulationof administrativepolicies andprocedurescoveringareassuchas data ownership,access, and security (Appleton, 1984; Weldon, 1986). Approaches with a focus on top-down,business-relatedplanning.Theseinclude planning processesand methodssuch as Martin’s (1982) strategicdataplanning,Holland’s(1983)strategic systemsplanning, and IBM’s businesssystems

374 MIS Quarterly~September1988

planning (BSP)(IBM, 1981). All of these proaches link the acquisitionanduseof datawith businessobjectives. It is increasingly evident that neitherof thefirst two categories providesa completelyadequateapproach. Coulson(1982)acknowledges that many efforts to solve data management problems throughthe implementation of a data dictionary havefailed. Kahn(1983)presentsempiricalevidencesuggestingthat mostdata administration groupshavehadlittle or nosuccess in correcting key data management problems, and Tillman (1987)discusses severalreasons for the failure data administrationorganizationsto live up to management’s expectations,suchas insufficient management support. Thethird type of approach hasreceivedgreatattention, becausethe ultimate goal of data resourcemanagement is not to put tools in place or to createorganizationalunits but to provide data to supportthe needsof the business.Such planningapproaches, however,require significant resourcecommitments and, as will be discussedlater in the article, often are not easyto accomplish.

Research approach Giventhe limited generalknowledge of firms’ experienceswith existing approaches to data managementand the needto developnewwaysof thinking aboutthe area,exploratorycasestudies havebeenconductedacrossmultiple firms and industries. Oneto six daysof interviewswith IS managers and user managers werecarried out in eachof 20 large corporationsduring1985-86.The firms werefroma rangeof industries, including electronics, consumer goods,insurance,banking, computers,andenergy.Thelargest company wasin the top 10of the Fortune500;the smallest hadannualrevenues of $500million. Table1 providesbrief descriptionsof the firms (with disguisednames)and their data managementefforts. In each company,wefocusedon oneto four datamanagement projects, for a total of 31data management efforts studied.Almostall of theseefforts wereviewedas successfulby both information systemsand user management. In all, over230managers andotherprofessionals wereinterviewed;approximately 70 percentwere fromIS departments and30percentfromuser departments.Theintervieweesdiscussedthe datarelated policies, processes, controls, standards, andtools that their firm hadin placeor hadat-

DataResource

Table1. FirmsandDataManagement Efforts in the Study Disguised Names FinancialServices CoolidgeInternational

Description

Data Management Efforts

Financialservices,over $3billion in revenues

OperationalSDABs to support core customer financial services

DobbsInsurance

Among Top 10 insurance cos. in assets

Change in the basic businessdata collectedby onedivision

Taft Insurance

Over$3billion in revenues

Informationdatabase to support auto insurancebusiness

VanBuren Bank

Over$50billion in assets

Customer information database for top 1000customers in corporate and government bankinggroup

Blaine Corporation

Personalcare products; Fortune500

Dataaccessservicesfor andusers

Crockett

Canadian subsidiaryof Fortune500U.S. computer company

OperationalSADBs for finance and administrativeapplicationsand managerial reporting

EaldnsCorporation

Multibilliondollar international 1. Informationdatabase for conindustrial andservice company solidatedfinancial datafrom subsidiaries 2. Informationdatabase for employeeinsuranceclaims

Foothill Computer

Fortune250

1. Informationcentersproviding data anddata consulting 2. Setof standard datadefinitions establishedby corporatetask force 3. Strategicdataplanningfor the orderflow function

GlobalProducts,Inc.

Fortune500manufacturing

Subjectarea databases being gradually implemented

LDI Electronics

Over$2billion in sales

1. Corporateinformation database for engineering specifications for product 2. Informationdatabase qualitytracking 3. Informationdatabase for managerial analysisacrossproduct divisions in the components group

Manufacturing

MIS Quarterly~September1988 375

DataResource

Table1. (continued) Disguised Names

Description

Data Management Efforts

MatracCorporation

Fortune100 manufacturing

Dataaccessservicesfor endusers

SpectrumElectronics

Fortune100

Integrated manufacturing database in onedivision

ProcessIndustries Derrick EnergyProducts LargestSubsidiaryof Fortune 100 companies

1, BSPin ProductionDivision as part of 2. Six operationalSADBs its asset management program

Sierra Energy

Amongtop 10 energy companies

Informationdatabase for consolidatedfinancial reporting

WaverlyChemicals

Fortune500diversified chemical company

1. Common manufacturing systems in largestdivision 2. Common accounting systems usedby multipledivisions 3. Strategic data modelingdone for IS planningin largestdivision 4. Setof standarddatadefinitions establishedby corporateIS group

WindsorProducts

Fortune100

Other Consumer Publications

Severalinformationdatabases built "Datacharting"effort

Over$1billion in revenues

1. Strategicdataplanningfor the corporation 2. OperationalSADBs to support the promotionfunction

Diverse Conglomerate

Fortune 500

Informationdatabase for useby senior management

National Technologies

Over$5 billion in revenues

Strategicdataplanningfor the corporation

Winslow

Aerospace division of a multibillion dollar conglomerate

Dataaccessto divisional data throughfourth generationtools

376 MIS Quarterly~September1988

Data Resource

tempted.Also discussedwere the factors that motivated eachorganization to take action and the results that wereachieved.In addition, the interviews gathered opinions regarding the most important problems and issues concerning the management and use of data. An initial assumption,supported by the present findings, wasthat similarities in data management issues faced by large corporations outweighdifferences amongindustry groups. In choosing companies to study, the goal wasto find a rangeof firms that were actively trying to improvetheir managementof data. The sample was drawn from firms knownto have forward-thinking IS groups. This meansthat the sampleis not random. However,becauseof the numberand variety of firms studied, the findings should be generalizable to other large organizations that have recognized data management as an important issue.

A ContingencyApproachto Data Management The picture emergingfrom these case studies is that effective data management efforts fit no single clear pattern. In analyzingthe cases,there is an interlinked set of choices that dependheavily on organizational considerations. Certainly, one must be careful in generalizing from a non-random sample of 20 companies, but, at least in these organizations, successful efforts appearto be diverse in terms of (1) businessobjective; (2)

Business Objective

Scope

¯ Coordination (6)

¯

Corporate (10)

¯ Flexibility(3)

¯

Division(8)

¯ Improved Managerial Information (16)

¯

Function (13)

organizational scope; (3) planning method; and (4) "product," i.e., majordeliverable produced. As a starting point for visualizing the contingency approachto data management, Figure 1 presents a simpleframework that reflects the variety of options foundin the casestudies. Thefour mainelements of the framework represent the key components of the data management efforts studied. Theseelementsare: ¯ The identification of a businessobjective. In the successful companiesin the sample, data management actions were almost always justified not by conceptualor technical arguments, but by one of four compelling business needs: operational coordination, organizational flexibility; improvedmanagerialinformation; or IS effectiveness. The scope of the data managementproject. Thefirms ~tudiedexplicitly defined andlimited the organizational scopeof their efforts. Some focusedon a functional area (such as finance), others on a division, while somewere corporate-wide. The data planning method. Top-down, indepth stategic data modelingwasnot the only data planning process.In fact, there appearto be a numberof obstacles to accomplishing a large-scale strategic data planning effort. The planning processesutilized varied widely in termsof their formality, their detail, andtheir emphasison data models. The range of options varied from strategic data planning to more limited planning approachesto no planning whatsoever.

Data Planning Process

Product

¯ Strategic Data Modeling (5)

¯ SADBs(7) (operational data)

¯ "80/20" Methods(4)

¯ Common Systems (2) ¯ Information Databases (11

¯ Targeted(15) ¯ Data Access Services (4)

¯ IS Effectiveness (6) ¯ None(7)

¯ Architectural Foundations (7) Note: The numbersin parenthesesshowhow manydata management efforts fell

into each category.

Figure 1. Frameworkof Data Management Choices MISQua~erly/Sep~mber

1988 377

Data Resource

¯ The "product" of the data management effort. Muchof the existing data resource managementliterature centers on the implementation of subject area databases(Martin, 1982). In the case studies, however,were seen five distinct "products," which were the end results of the data management project team’s work. These products are: subject area databasesfor operational systems; common systems; information databases;data accessservices; and architectures or standardsfor future systems. Thenext four sections of the article discuss each of the elementsin the framework.To facilitate the discussion, the elementsare discussedin reverse order from that shown above, with the most tangible element, the "products," first, followed by planning processes, scope, and business objectives.

Five data management "products" Any successful data management effort results in a product or deliverable such as a newsystem, service, or policy. The "product" most common in the existing data resource management literature is a set of subject area databasesusedby multiple operational systems. Wealso found, however, four other products. Like subject area databases, two of the other product types--commonsystems and information databases--require a systems development effort. The final two product types, whichusually do not involve newsystemsor databases, were labelled data access services and amhitectural foundations. Subject Area Databasesfor Operational Systems Subject area databases (SADBs)contain data that is organizedaroundimportant businessentities or subject areas, such as customer and product, rather than aroundindividual applications, such as order processing or manufacturing scheduling. Manydifferent operational applications mayshare data (i.e., both access and update data) from a single set of SADBs. In sevenof the 31 cases, the product can best be described as SADBs for operational data. Consumer Publishing,Inc.1 doesmanydirect mailings using its large customer baseandother purchasedlists in order to promoteits booksand magazines. Tohelp supportandcontrol this important activity, Consumer Publishing hasbuilt a set of 1All company nameshavebeendisguised.

378 MIS Quarterly/September

1988

SADBs (including product,vendor,postal regulations, andpromotionplan) that are usedby three major applications: (1) an MRP-likesystemfor planningandexecutionof mailings;(2) a purchasing system for mailing-related materials;and(3) inventorysystem.Theimplementation occurredin severalphases between 1981-84, andwasthe first time that Consumer Publishing’s IS department took a data-focused approach to systemsdevelopment.TheSADBs andapplications havehelpedto simplify andimprovemailing-relatedoperations, andhavegreatly reduced inventorycosts. CommonSystems A secondtype of data management product is the set of operational data files or databases that are developed for commonsystems. Common systemsare applications developedby a single, most often a central, organizationto be usedby multiple organizational units. Physically, there can be one or multiple copies of the system. The concept of commonsystems is not new. They have often been developed not for data managementpurposes but rather to ensure common proceduresor to lower IS costs. Common systemscannot be developed, however,without surfacing and resolving data definitional issues, since old systems (and old definitions) will be discontinued. WaverlyChemicals is a major,diversified chemical company, andits largest division operatesabouta dozenplants. Sinceabout1980,this division has emphasized the development of common systems for manufacturing applicationssuchas production scheduling and spare parts inventory. The availability of well-defined,standardized datahas enabledthe division to reduceinventorycostsand to greatly improvecoordinationamong the plants. While only two of the companiesstudied developed new common systems, several of the other companies had a significant existing base of commonsystems. Thesefirms often were able to use the standard data in their commonsystems to leverage other data management efforts. Information Databases A third new systems product is an information database, whichcan be defined as a subject area databaseintended for use by staff analysts and line management.Information databases are "secondary" databases, which periodically draw their contents from operational databases and (sometimes) external sources, and often store data in aggregatedforms. Significantly, information databasescan provide data without requiring

Data Resource

major rewrites of current systems. Instead, "bridges" are built from the existing operational systems to provide the appropriate data to the new database.

deliversthe datain the user’schoiceof formats. Thegroupmaintainscopiesof mostof the datait providesandarranges for periodicupdates.

Information databases were the most prevalent productin the sample,occurring in 11 (35 percent) of the 31 cases.

Dataaccessservices are expect to be morehelpful in companies whereexisting data is of reasonable quality. Wheredata is of poor quality, the processof delivering data in its current form to managersmayspur action toward increasing data standardization and control mechanisms.Some firms are following upinitial efforts to providedata accesswith the developmentof a data directory designedspecifically to assist endusersby listing andcross-referencingthe available data.

At WindsorProducts,a consumer goodsmanufacturer, corporatemanagement’s demands for informationled to the developmentof several new databases.Standardcodesanddefinitions were definedfor thesedatabases; however, existing applications andthe operationaldatabases on which the transactionsystems depend wereleft in place. Automated bridgesfromthe existing systems populate the new"information-only" databases for customer,product, andshipment.Theseinformation databases are nowwidelyused. LDIElectronicsis buildinganinformationdatabase to containdataonkeycharacteristicsof the parts andmaterialsthe company usesin its electrical components, includingspecifications,availability, reliability, andcost.Mostof the datais already collected, in variousforms,by existing procurement, engineering, and manufacturingsystems.Thus, bridgesandtranslationroutinesare beingbuilt to the newdatabase.A prototypeis in use by engineersin nineproductgroups.

Data AccessServices The first three "products" discussed emphasize developingnewdatabasesor files with pertinent, accurate, and consistent data, Four firms in the study, however, focused mainly on improving managerial accessto existing data, without attemptingto upgradethe quality or structure of the data. Data access services are usually provided by a small cadreof personnel,often part of an information center, whosegoal is to better understand what data is available in current systemsand to put in place mechanismsto deliver this data. These mechanisms include locating appropriate data, extracting data from production files, or training users in fourth generation languages. Such efforts are widely applauded by managers whonowhave help in "getting their hands" on existing data. A multi-billion dollar manufacturing firm, Matrac Corporation, hasputin placea "dataservice"for its corporateendusers. Thedata service organization is a smallgroupwithinthe corporate IS organizationthat locatesdata,arranges for extracts,and

Architectural Foundationsfor the Future In mostof the firms studied, managers focusedon a limited set of data servinga portion of the corporation. However,there clearly is a dangerin approaching data management function by function, businessunit by businessunit, or subject area by subject area. A companymayfind itself facing problems in the future if it desiresto integratedata acrossthese boundaries.To avoid these future incompatibility problems, someorganizations have focused on developing "architectural foundations," whichare policies andstandardsthat force systemsdevelopment efforts to conformto a wellstructured, overall data plan. Onetype of architectural foundation is a corporate-wide strategic data modeldesignedto serve as an underlying blueprint for all future systems development. Martin’s (1982) Strategic Data Planning approachproducessuch a modelas one of its products. IBM’s BSPmethodology (IBM, 1981.) and Holland’s (1983) methodology are others that producea data architecture. Proponents of these approaches argue that a strategic data model provides an architectural foundation that will lead to consistencyof data, more easily integrated systems, and improved productivity in systems development and maintenance. Five of the firms studied developeda strategic data modelprimarily as an architectural foundation for future systemsdevelopment. WaverlyChemical’slargest division begana strategicdatamodeling effort for the entiredivision after the successful implementationof common manufacturingsystems.The modelwasused to helpcreatea strategicIS planthat identified key business areasthat hadlittle systems support.Due to a downturn in the division’s primarybusiness,

MIS Quarterly~September 1988 379

Data Resource

limited action hasbeentakenbasedon the plan. Thedivision, however, hasshown its datamodelto otherWaverly divisions;at least oneotherdivision hasdecidedthat the modelfairly closely fits its business andhasusedit, with some modifications, asa basisfor its IS plans. A second, morelimited approachto data architecture is the standardizationof data definitions and codes. Most of the new system products developedby the firms in this study require that line management and IS technical personnel agree on the precisedefinitions of a specific set of data elementsas a prerequisite to building the system.In two firms, however,a set of corporate-wide data definitions wasdevelopedsolely as an architectural foundationto facilitate meetingfuture information requirements. In 1984, Foothill Computer formeda task force, chairedby corporateIS, to identify anddefinekey dataelements beingusedin multiple areasof the business.Therewereno immediate plansto implementthe agreed-upon definitions. Rather,it was assumed that future systemsdevelopment work wouldconform to thesedefinitions. In addition,it wasestablishedthat anygroupsupplyingdata to anothergroupwithin the corporationwouldbe requiredto deliver that datain conformance with the definitions, if asked.After coming to agreement on definitions for over 200data elements,the task forcehasmorerecentlyrefocused its efforts to concentrateonly on definingthoseelements for which a specific businessimpactcan be identified and pursued. As these examplesillustrate, either a widescope strategic data modelor a set of standarddata definitions can be a product in its own right. When data modelsor standardsare enforced, an organization should gain a major asset of interpretable, shareabledata. But usefulnessof these architectural foundations can always be questioned, unless the data model is used to guide future systemsdevelopment, or unless the definitions eventually becomeincorporated in either operational or managerialdatabases.

A rangeof data planningprocesses This section focuses on the planning processes organizations use to identify the target for data management action, and to choosethe action (or "product") to pursue. To manypeople, planning for data resource managementis synonymous with a large-scale strategic data planning and modeling effort. There are, however, other less comprehensiveplanning approachesthat can be

380 MIS Quarterly/September 1988

extremelyeffective. This section categorizes the planning processes from the case studies into four types: strategic data planning; "80/20" approaches; "targeting;" and no explicit planning. Theseapproachesrepresent a continuumof planning processesthat ranges from global, well-defined, rigorous methodsthrough morelocal, often less formal approaches,, to no data planningat all. Strategic Data Planning Strategic data planning is the category that encompasses rigorous top-down planning approaches focused~on understanding and modeling data in the context of business functions, Theresulting plan definesan architecture of major subject area databasesand prioritizes their implementation. 2 The diagram in Figure 2, adapted from Martin’s (1982) Strategic DataPlanning Methodologies, is representative of these approaches. The left side of the diagram showsa top-down planning approach, leading to the identification of logical subject area databases. In general, only selected portions of the plan are chosen for bottom-up design and implementation. The underlying assumption of top-down data planning methodologies--thatit is impossible to plan effectively if one does not knowwhat the businessis, whatit does, andwhatdata it uses-is difficult to contest. However, of the 31 data management efforts studied, only five useda strategic data planning approach. In general, these five firms undertooka strategic data planningeffort to produce both a strategic data model (which we call an architectural foundation) and a plan of action. Noneof these firms sawthe kind of successenvisioned in the literature a master data architecture that identifies strategic opportunities and guides all newdevelopment.The outcomesof the five efforts varied, but in nonedid the plansdirectly lead to the implementationof newsystemsor sub~Thewordplanhastwodefinitions: (1) a scheme of action or procedure; and(2) a representation drawnon plane, e.g., a map,a model.Althoughthe termsdata planninganddata modeling haveoften beenusedsynonomously, it is helpful to distinguish between them. Dataplanningis usedhere to m~&an aneffort to develop a "schemeof action." Datamodelingrefers to the preparationof a "representation"with predetermined scopeandlevel of detail. Therepresentation maybe a useful aid in develoPing a scheme of action, or when donein great detail, maybethe basisfor a system design. Manystrategic data planningmethodologies aspire to produce both a model anda plan.

Data Resource

Enterprise Modeling

!

Application Programs

Data Model

Entities

Logical Subject Area Databases

A strategic data planning processbegins with the development of an enterprise or business model(Box 1, above). Theenterprise modeldepicts the functional areas of the firm, and the processesthat are necessary to run the business.Thenext step is to identify corporatedata entities andto link themto processes or activities (Box 2). Datarequirementsare thus mapped onto the enterprise model,leading to the identification subject areas for which databasesneedto be implemented(Box 3). In general only selected portions of the enterprise modelandsubject area databasesare chosenfor bottomup design. Building the logical data modelis the first step. Thedata model(Box5), results from a synthesis of detailed management and end-user data views (Box 4) with the results of the previous top-downentity analysis (Box 2). Databasedesign and subsequentdesign of application programs(Boxes6 and 7) proceed from the logical data model. 3 Figure 2. Strategic Data Planning

ject area databases. In one case, although the clearest benefit wasa better understandingof the data by those involved in the planning process, the effort wasviewedas worthwhile.Twoof the efforts weregenerally perceivedas failures. In 1982,the centralIS planningstaff at National Technologies, Inc., a large high technology firm with 20 divisions, begana comprehensive data plann!ngeffort in response to the complexityand variability in NationalTechnologie’s business envi3Adapted from James Martin, Strategic Data-Planning Methodolgies,1982, p. 109. Usedby permissionof Prentice-Hall,Inc., Englewood Cliffs, NJ.

ronment. Thegoalwasto link strategicIS planning with strategicbusiness p~anning andto link a logical modelof the businessdata with the developmentof physical databasesand systems. The planningeffort tookabouta yearandinvolvedeight peoplefromIS planningandthe user community. Althoughthe strategic data modelwascompleted andturned over to systemsdevelopment teams, the modelwasnot adhered to. Thepressures of the operatingenvironment took hold, anddeadlines, not the globaldata model,became the significant driving force. Also,conflicts among the usersabout definitionsandusesof dataarose,despitethe generally agreedupondatamodel.In addition,limitations in data-orienteddesigntools andrelational databasetechnologywerenoted.

MIS Quarterly~September 1988 381

Data Resource

In the other two of the five cases, while strategic data planningdid not lead directly to an IS plan of action, it did providean architectural basefor subsequent data management efforts. DerekEnergyProductscompleted a businesssystemplanningeffort for its productiondivision in 1982.Shortlyafter the planwascomplete, the divisionparticipatedin a majorreorganization, andthe plan wasshelvedas no longer appropriate. In 1984,Derekformulateda newlong-rangeIS strategy.Thestrategyidentified a "target future state" for the division’s informationsystems.Keyuser managers participatedwith IS personnel in the development of animplementation plan basedon the "targetfuture state," critical business objectives, andthe currentslate of system requests.Aspart of the planningeffort, the businessprocess/data modelfromthe BSPwasrevivedandmodifiedby a small groupof IS professionals. TheBSPmodel wasuseful in categorizing50 currentproject requestsandin makingit apparentthat a third of theserequestsdepended on the samesix subject area databases.Thesesix databases are nowbeing developed. In most of the data management efforts studied, the planning and implementationprocess did not

proceed as suggestedby strategic data planning methods.Figure 3 illustrates the actual planning processfor two companies that weretypical of the other casestudy sites and, in contrast to Figure 2, showsan altered sequenceof steps. In these two cases, as in most of the other firms studied, the companies skipped or abbreviated the "left side" or top-downportion of the top-downplanning, bottom-updesign process. In doing so, they followed the alternate planning processesto be discussed in the next sections. As shownin Figure 3a, SpectrumElectronics choseto developmanufacturingsystemswithout havingusedanytop-down,data- orientedplanning methodology to arrive at that decision. Thecompanystarted with the objectiveof implementing a particularset of applications (1), thendeveloped logicaldatamodel (2) for this application set’s data, andfinally designed the physicaldatabase andthe applications (3). SierraEnergy (Figure3b) startedwith a particular user’sview(1)---that of the corporate controller-anddesigned a data model(2) andphysical databases(3) from that perspective. Methodsfor "bridging" data fromexisting applications were then developed (4).

Figure3a. Spectrum

Figure 3b. Sierra

Application Programs

Existing Applications

Data Model

Data Model

User Views

Corporate Controller UserView

Figure 3. The PlanningReality

382 MIS Quarterly~September 1988

DataResource

Thereappearto be a number of reasonsthat help explainwhymorefirms are not successfullyusing strategic data planningmethods.Table2 lists some of the reasons whyit is difficult to gaincommitmentto the processand to manage expectations regarding the outcomeof strategic data planning.Thebottomline is that, for many firms, the approach is too expensive, its benefitsare too uncertain,andit is organizationallydifficult to implement. TargetingHigh-ImpactAreas Mostcorporationsthat skip or abbreviatetopdownplanningmethods do not act without a plan. Thereare a variety of alternative planningprocessesthat canbe used. Themostcommon process(seenin 15of 31cases)is the"targeting"of particular functionor business area.In some companies, important problemor opportunity areas are evidentwithoutextensiveanalysis. At SierraEnergy,the corporate controllerknew fromhis experience that heneeded, but did not have,consistent, accurate datafor corporate reportingpurposes. Heinitiated thedevelopment of aninformationdatabase to be usedby the company’smany decentralized business units.

When top management at LDI Electronicsmade improved product qualitya corporate priority, the onlyquality-related information easilyaccessible waswarranty accounting data.It wasclearthat additional dataneeded to be collectedandmade available.A corporate-wide information database for product qualityis beingdesigned. NeitherSierra nor LDI useda rigorousdataplanning method.But in both comapnies there were key line managers whocould visualize the benefits of accessiblequality data. In eachcase,the data management programwaslimited in scope but waseffectiveandfeasible. "80/20" PlanningMethods In some firms, thereis a desireto get the major benefitsof global dataplanningwithouthavingto investthe resources necessary to carryout a fullscalestrategicdata planningprocess.Theaimin thesecasesis to zeroin quicklyonthe key"products" to be implemented (bottom-up),while reducing the amountof effort spent in a global planning(top-down)phase.Thistype of planning, foundin four of the 31 cases,can be termedan "80/20"approach,after the adagethat for many undertakings,80 percentof the benefits canbe achieved with 20percentof the total work.

Table2. WhyStrategicDataPlanningApproaches Are Difficult to Implement ¯ Thestrategic dataplanningprocess,donein detail andwith a widescope,canbevery timeconsuming andexpensive.For the processto be successful,keyoperatingmanagers mustcommitsignificant time andeffort. This commitment is often difficult to obtain (and keep)from these busy individuals. ¯ Because of the up front effort needed, organizationsface a longerandmoreexpensive development processfor the initial systems developed with dataplanningmethods. Linemanagers do not like to seeproject schedules lengthened. Similarly, IS managers, whohaveincentivesto deliver quicklyand to containcosts,mayresist theadditionaleffort involved. ¯ Themethodologies requirenewIS skills and, therefore,maynot be easily adopted by IS personnel. ¯ Oftenthe businesswill changewhile the plan is beingdeveloped andimplemented. ¯ Total implementation of a wide-scope data planningeffort canbe extremelyexpensive.Thereis a tendencyto avoidthesenewcosts, especiallyif manyof the existing systems,whichrepresenta hugeinvestment are still effective. ¯ When implementing only a subsetof the plan, it canbe difficult to bridgethe gapfromthe top-down planto bottom-up design.If proposed andexistingsystems interface alongdifferent boundaries, it maybehardto isolate andreplacea subsetof existing systems with a subsetof proposed systems. Theuseof applicationpackages also createsinterface andboundary problems. ¯ It is not alwaysclear to the plannersor top management whether a strategicdata model is beingdevelopedto produce a systems plan, createanarchitecture,or to designnewdatabases. It is difficult to manage the expectations of thoseinvolvedregardingthe results andbenefitsof the process.

MIS Quarterly~September1988 383

DataResource

Windsor Products, having carriedoutits first round of datamanagement in a quicktargetedmanner, founditself uncertain as to the nextinformation databases to implement.Management decided, however, that a full strategicdatamodel wasnot necessary for its purposes. It therefore developed its ownabbreviated planningapproach, whichit calls"Datacharting." Thisinvolved identification of about100majordataaggregates (i.e., high-level dataentities)andthegroups in thecorporation that usedthem.Thismethod tookless thansix manmonths of effort, andinvolved reviews byline managers in all partsof thebusiness. Thedatachartis serving asthebasisof planning for thenextround of information databases. The problemwith strategic data planning approachesis the investmentof time and dollars needed to obtain results. Drawbacks of "targeting" includethe probableinconsistencies that will arise frommultiple targetedprojectsandthe fact that in somecompanies,the most important targets maynot be evident. An 80/20approach, while not providingthe detail of strategic data planningor the quick hit of targetedapproaches, doesappearto offer the majorbenefits of both prev!ous approaches. No Planning Process

In 10cases,the scopeof the planningeffort was corporate.In only three casesdid the scopeincludeall of the corporation’sdata; twoof these werecorporate-wide strategic data planningefforts for whichtheresultswerenotas usefulas th_e organizationhad expected.Thethird wasWindsor Product’s80/20planningeffort to developa corporate-wide data chart. In the othersevencases,the effort involveda limited subsetof the total datausedby the corporation. For example,one effort focusedon key corporate-widesubject areas, suchas customer or product.Twofocusedon a small set of important datadefinitions: Othersprovideduserswith adhocaccess to existingdatafiles. In additionto functionsanddivisions,othersuborganizationssuchas.groups,geographic districts, andproductlines exist within a firm. Some corporations maychooseoneor anotherof theseunits as a locus for data management efforts: As the nextsectionwill show,the scopeof the datamanagement effort is substantiallydetermined by the businessobjective.

Businessobjectives, not conceptual justifications

Bounded scope

Theproponents of data management far too often basetheir arguments on either the conceptual soundness of viewingdataas a resourceor the rationale underlyingdata-centered systemsdesign. Theyassert that processeschangewhile data is relatively stableandthat datashouldthereforebe the keyelementin IS planning.Ortheyarguethat global data management is essential because oneneedsa global plan beforedeveloping the individual pieces. Whilethese arguments are appealing, they rarely engenderaction in the pragmatic,cost-conscious world of the business manager.

In this study, nofirm attempted to manage all the datausedby the corporation;all limited the focus of theeffort in oneor moreways.Animportantfactor in the successof data management efforts is that thescope(i.e., the part of theorganization to beincluded in the effort) becarefullyselected.Althoughthe scopeof eight data management efforts wasdivisional, 13 cases,strikingly and logically, focusedon a functional areasuchas manufacturing or finance. This limited scopewas usually dominatedby a single manager whowas ableto visualize the results of a datamanagement effort in a segment of the business.

Most successful data management processes observedhavebeenaimedat solving a clear and specific businessproblemor exploiting an opportunity. In a some cases,the data management effort occurredbecause of a line manager’s needto respondto changes,suchas increasedcompetition, in his or her 6nvironment.IS departments havealso hada keyrole in initiating many of the efforts andin educatingline managers as to how improvingthe management of data cancontribute to the business.Th!ssectiondiscussesthe business-relatedreasonswhyfirms are motivatedto considermoreproactive management of data.

Thereare also data management actions that can be taken without anydata-orientedplanning.For example,if a decisionis madeto providebetter access to existing data without addressing changesin the form of that data, then no data planning methodology is needed.Rather, data can be madeavailable as it is requested.This wasthe caseat Matrac,with its data services approachdescribed earlier and at six other companies.

384 MIS Quarterly~September1988

Data Resource

Operational Coordination A major objective for data management action is to better coordinateoperational activities, either within specific functions or business units, or across them. This objective often arises when competitive pressures cause a firm to focus on cost reduction. Improvedcoordination requires an enhancedability to communicate within or among organizational units. In practical terms, this implies the ability to readily sharedata. Thereare six clear examplesof coordination as a motivation in the casestudies. Both SpectrumElectronics and WaverlyChemicals felt the needto standardize their manufacturing systems so that manyplants could be coordinated moreeffectively. In both cases,their efforts havemeantsignificant benefits. For example, the standardization of datahasled to reduced in-processandinter-plant inventoriesandto the coordination of sparepartsavailability, whichhas reduceddowntimes. Common data has also facilitated coordinatedpurchasing,whichhas enabled special arrangements with vendorsto be made.

OrganizationalFlexibility A secondtype of objective for data management, seenin three cases,is the desire for greater organizational flexibility to allow either an internal restructuring of the organization, or a refocusing of the organization due to changes in the environment, WaverlyChemicals hasrestructuredits divisional organization severaltimesoverthe pastdecade. It merged two large manufacturing divisions in the mid-1970s,andfaced major problemswith accountingsystemsthat hadbeendesignedandimplemented separatelyin the original divisions. In the late 1970s,Waverly combined five old divisions into twonewones.It wasquite clear to seniormanagement that therewouldbeother reorganizations in futureyears.Aslongas eachdivisionhadits own accountingsystemswith muchincompatibledata, the problems wouldpersist. Changingan organization’s strategic focus can also require more effective data management. Several companiesstudied have been faced with important changesin the marketplacethat put intense competitive pressures on them to change from a product focus to a marketor customerfocus. Organizationalflexibility is often hinderedby data structures that have beendesigned to support particular applications or suborganizations

but which are not flexible enoughto provide new strategically important "views" of the business. In 1984, DobbsInsuranceCompanies determined that the basicdatastructureusedby its Group Insurance Division,whilestill appropriate for 90percentof its currentbusiness, wouldnot supportits future needs.For example, the "flexible benefits" products the company wasconsideringoffering wouldrequiremajorchanges in the wayit keptdata aboutpolicies andpolicy holders.Giventhe competitive pressures in the insurance industry,Dobbs felt it hadnochoicebut to move to a moreflexible datastructure.

ImprovedInformation for Managers The dominantobjective for more effective data management, seen in 16 of 31 cases, was to improve information for senior managers,middle managers,and key staff personnel. The needfor better data is recognized, for example,whenmanagement wants to analyze changing market trends or moreclosely monitorprofitability. These information consumerswant two things: improved accessto data andimproveddata quality. VanBurenBankwantedto manage customerrelationships better, especially withits largestclients. Because eachbankbranchassignedits owncustomeridentifier to eachcustomerwith whom it dealt, therewasno easywayto aggregate data for a single customer acrossthe entire bank.Thebank has nowspentover $1 million to developa customerinformationdatabase to allow accountmanagersto retrieve information aboutthe bank’s1000 largest customers. Whenproblems such as these becomeimportant enoughto management, strong motivations arise to improvedata quality and access. IS Effectiveness As information and information technology becomemore important to firms and information systemsbudgets grow, there is strong pressure on IS groups to not only develop systemsfaster (while controlling costs andimprovingquality) but also to be moreproactive in addressingbusiness needs. Improved data managementcan potentially contribute to the effectivenessof IS planning anddelivery by linking the data requirementsof the businesswith IS plans, by increasing systems developmentproductivity, or by reducing systems maintenance costs. Data managementactions mayalso be motivatedby the recognition of a lack

MIS Quarterly~September 1988 385

DataResource

of integrationamong existingsystems andthe difficulties this causesas demand for multi-function or multi-organization systemsor information grows. In six of the 31 cases,the primarymotivatorfor data management efforts seemedto be general IS effectiveness,without anymorespecific businessgoals. Fourof thesesix werestrategic data planningefforts, whiletwowereefforts to identify and standardize key corporate data element definitions. Manyfirms felt that the data management actions they wereundertakingwouldinvolve greater, not lower,IS costsin the shortrun. Onthe otherhand, twofirms, actingwith a clear business objectiveas the primarymotivator,hadIS effectivenessas a secondaryobjective and said that they had achievedreductions in development or maintenancecosts.

rial information,in a functionalareaby targeting key businessopportunitiesandby building informationdatabases.Twoother firms followed almost this samepath, but with a-corporate or divisional scope.Management’s needto be able to seeconsolidatedfinancial information, comprehensivedata aboutits customers or profitability of productsis a majorimpetusfor data management action and can often be addressed throughan informationdatabase. Thesecondprominentpattern (dashedarrows), with four cases,representsfirms seekingbetter operationalcoordinationin a functional areaby targetingkeybusinessopportunitiesandbuilding either operationalSADB’s or common systems.If operationalcoordinationis the goal, then these two productsare the mostappropriate.

Spectrum Electronicseliminatedseparateprograms andprogramming staffs in its plantsby centralizing all data processing andby usinga centralized database andcommon software for all manufacturing systems. Thecompany also claims to havereduced maintenance costsby eliminating notjust redundant data,buttheprograms that updatedredundant data andthe programmers who maintained thoseprograms.

Thethird pattern(dottedarrows),also with four cases,is made up of firms seekingimproved management information,at either a corporate or divisional level, whichput in place improveddata accessservices usingno explicit data planning process.Thisis animportantoption,especiallyin companieswith manycommon systems extensive data definition and codingstandards.This approachrapidly placesdata whereit is valuab~c~n the handsof end users--anddoesso at minimalcost.

Crockett,the Canadian subsidiaryof a major American computer manufacturer, claimsa 40percentreduction in development costsbecause of a newdevelopment processfocusedon data, the useof anactivedatadictionary,andusers’generationof their own reports.

Twoof thesethree majorpatterns(and11 of the 31 cases)involve functional scopeandtargeted planning. This reflects the importanceof managersin specific functionalareaswhoare ableto seeconcretewaysin whichbetter datacanassist themin their areasof responsibility.

Majorpatternsin the cases Thefour critical components for data managementaction--businessobjectives, scope, data planning process, and "product".--provide a roughframeworkfor thinking aboutthe 31 data management efforts. By placing those efforts in the appropriatecategoriesof the framework,as shownin Figure4, wecanseewhichchoicespredominate,and whatcombinationsemergeas re4 currentpatterns. Threemajorpatterns,shown by the arrowsin Figure 4, emergewhenthe sets of choicesmadeby the casestudyfirms are analyzed.Themoststriking pattern(solid arrows),followedin six cases, representsfirms that soughtimprovedmanage4Examining the patternsby industrygroupshowed no majordifferences; seetheAppendix.

386 MIS Quarterly~September1988

If welook at the numbers of occurrences in each boxof Figure4, it becomes apparent that for three of the components of data management action, onealternative dominates the others. Sixteenof the 31 efforts weremotivatedby a businessneed for better managerial information.Fifteen of the efforts addressed a targetedneedwithoutusinga moreformal planningprocess.Elevenof the 31 productswereinformationdatabases.Thatthese alternativesare well represented is notsurprising, but the fact that theyare so predominant suggests they are "high-action areas" that ought to be carefully consideredby practitioners of data management. Finally, eventhoughthere wereonly four instancesof 80/20planningprocesses in this study there is a needto emphasize the importanceof this approach.In manycorporations, after the moreobvioustargets are addressed, animportant shift is expected toward80/20planningprocesses

Data Resource

Business Objectives

Coordination6...............

Scope

-~1

Function ........... 13

Planning Process

Product ......

~’~ I Targeted 15

Divisional 8

80/20

Corporate 10

Strategic DM 5

4

None

~ I Common Systems 2

I

Operational SADBs 7

Information Databasels 1J

Data Access

4

Architectures 7 Note: The numbersshowhowmanydata management efforts fell

into each category.

Figure 4. Major Patterns in the Framework that identify strategic opportunities without the major investmentof morerigorous strategic data planning approaches.

framework’s categories represent a continuumof data management actions that can be directed towardany of three aspectsof a corporation’s data: infrastructure, content, or delivery.

Organizational issues Affecting Data Management Implementation

Infrastructure encompasses the standards that force systemsdevelopment efforts to conformto a coherentdata architecture. Actions to build a data infrastructure (e.g., developinga strategic data model)tend to be both difficult andexpensive,and the benefits fromsuchactions are mostoffen realized only in the longerterm.

In addition to the contingencymodelof data managementchoices presentedabove, the interviews with IS andline managers havesuggestedfive organizationalissuesthat affect the ability of firms to implementdata management efforts. This section briefly presentsthese issues.

Issue1: Short-term andlong-term trade-offs in resource allocation Managersconsidering data management actions must decide how to allocate limited resources amongactivities that will produceimmediatebenefits and"infrastructure" efforts that often do not have a quick payback.This tension betweenlong andshort term is reflected in our framework.The

Contentrefers to the choice of whatdata to maintain, andalso to policies that addressthe accuracy of that data. Systemsto capture.newor more detailed data and decisions to purchaseexternal data are examples of actions that affect data content. Theseefforts tend to be moderatelyexpensive, with benefits in the middleterm. Delivery refers to makingexisting data available to managerswho need it. Data consulting services, extract policies, andthe provisionof fourth generation reporting tools are examplesof mechanismsto improvedelivery. Actions in this area tend to be less expensive and have short-term benefits.

MIS Quarterly~September 1988 387

DataResource

Thechoiceof wherea firm shouldbestallocateits resources--among infrastructure, content, and delivery~ependsvery muchon the willingness and ability of senior management to invest currently for futurebenefits.A characteristic of Americanbusinesses has beenthe short-termfocus of line managers. Decisionsto allocate resourcesto dataarchitectureare in direct conflict with the pressuresto achievedemonstrable results, neartermearningsper share,andthis year’s return on investment. Notsurprisingly, mostof the successful efforts reportedin this article havetendedtowardthe short-termendof the spectrum.Onthe other hand,firms that haveput in placestronger data architectural foundationsappearless likely to face problemsintegrating and sharing data acrossthe organization.

Issue 2: Thecentralizing tendency of data management Underlyinganyeffort towardmoreeffective data management is the reality that improveddata management canlead to greater centralization of decision makingin an organization. Increased standardization of datafacilitates increased central control. For example,standarddata definitions maybe establishedas common systemsare developed.If the resulting data is made accessible to senior executives,they will havean enhancedability to compare operationaldetails of business units undertheir jurisdiction.Thereis indeeda tendency to act onthis kind of data;several of the systemsdescribedherewereinstituted to facilitate "central" coordination andcontrol.Even whereincreasedcentralizationis not a designobjective, it maybea result.

Issue 3: Impacton’the IS culture A third organizationalissueis the impactof data management on the IS culture. As Durrell (1985) points out, "data administrationchallengesthe basic process-orientedapproachthat has been employed during the last 20 years. This can be disconcerting and sometimes insulting to many long-timeDPprofessionals"(p. ID/29). Thenewdata-centered systemusedat Crockett has hada majorimpacton the workof programmers andanalysts.Themajoractivity of these peoplenowis workingwithbusiness managers to definethebusiness rulesgoverning themeaning of dataelements andentering thoserulesinto theactive datadictionary. Notsurprisingly, this hashada

388 MIS Quarterly~September1988

significantimpact ontheattitudesandturnover of theprogrammer/analyst staff there.While thelevel of enthusiam of currentpersonnel is high, many programmers foundtheir traditionalskills of no valuein the newenvironment andleft duringthe transition period. Theproblem is not only oneof teachinginformation systemsprofessionalsnewskills. Theremust also be changesin organizationalprocesses and managerialpolicies to support the movetoward data-oriented design. For example,incentive mechanisms must be changedto reward programmersfor conformingto a corporate data modelevenif it involvesadditionaltimeandresources. Without these changes,whensystem deadline pressuresbecomehigh, programmers will havea strongtendencyto developtheir own local datastructuresrather thanembark on negotiations with the dataadministrator for changes to the corporatedata model.

Issue 4: Newresponsibilities for user management If datais to servebusiness needs,to whatextent mustline managers, not only IS professionals,assumenewresponsibilities? In two firms studied user organizations assume almost completeresponsibility for data management. At Sierra Energy, a groupwithin the corporatecontroller’s organizationprovidesgeneralpolicies and support for systemsand data management actions relatedto thefirm’s financialfunctions.At Foothill Computer,a data management group within the corporate-levelcustomer administrationfunction is responsiblefor managing data subject areas suchas customer andprice. This groupviewsitself as a datacustodian,responsible for obtaining data from the appropriate sourceorganization (e.g., price datafromthe productline managers), for maintaining the integrity of the dataandfor delivering datato the userorganization’s transaction or informationsystems. Otherfirms focusedon increasinguser involvementin the processof managing data. At a minimum,manyfirms hadtask forces with user (and IS) representation to establishdatadefinitions. It seems obviousto saythat the effective managementof data as a corporateresourcerequiresthe participationof busi .hessmanagers. Butthe exact nature andscopeof line management’s responsibilities arenotwell defined.

DataResource

Issue5: Theprocessof effectively introducinginnovationsinto the organization Theimplementationof initial data management efforts canbeusefullyviewed as theprocess of effectively introducinginnovations,i.e., newmethods or tools of unprovenvalue. Rogers(1962) suggests that in generalthe diffusion of innovations is dependent on, among other things, five characteristicsof theinnovation.These are (1) the relative advantage of the innovation overits alternatives;(2) theobservability of theresults;(3) compatibilityof the innovationwith existingvalues, past experience,and perceivedneeds;(4) the complexity of theinnovation; and(5) its "trialability," or the extentto whichthe innovationcan be experimented with on a small-scale,low-risk basis. Rogers’(1962)researchcanhelp explain why hasbeendifficult to implement data management actions, especiallylarge-scaleefforts, in many corporations.First, the relative advantage of most data management actions comparedto current practiceis not known.Also, in mostorganizations there havebeenfewresults to observe.Where results areavailable,it is extremely hardto separate and quantify the impactof data management from otherrelated(or unrelated)actions. As notedpreviously, a data-focused approachto systemsis not compatiblewith the existing process-oriented focus,where the goalis to build individual systems to specificationon time, rather than to createa dataarchitectureto meetcurrent andfuture needs.Certainly, data management involvesa greatdeal of complexity as the walls between applications are torn downand the interrelationships betweensystems,functions, and organizationalunits are examined. Finally, very often data management actions have not beenpresented as testable, small-scalelow-risk efforts. Instead, data resourcemanagement has beensold onthe basisthat a majorfinancial investmentand top-to-bottom commitment in the organizationwill be needed to achieveresults. Viewingdata management as an innovation also helpsto explain the successfulefforts studied here. FromRogers’(1962)perspective,successful implementation of data management efforts is morelikely for morelimited approaches wherethe relative advantageis clearer, the impactsare moreobservable,andthe complexityis lessened. In addition, limited approaches canbe viewedas trials or experiments. When the first trial is successful,the organization will probablybereadyfor

a moreambitioussecondtrial andultimately for significant investmentin data management.

Conclusion Theseexploratory casestudies in 20 organizations suggest that thereis nosingleclear-cutapproachto improvingthe management of the data resource.A widerangeof optionsexists that can beselectedto fit the needsof a particular business. Theappropriate planningprocessto use andthe best"product"to deliver depend heavily onthe particular business objectiveandorganizationalscope.However, in spiteof this varietyin approaches,there are sevenimportantconclusions that canbedrawnfromthis research. Businessbenefits can result from improvementsin data management. Manyof the companiesstudiedhaverealizedsignificant benefits in their attemptsto improvethe management of data, as reportedby their ownevaluations. To highlight a fewof the benefits mentioned in the casevignettes:Wavedy Chemical attributes a reductionof 20percentin spareparts inventoryto the fact that its dozenplants nowusea common system;Sierra Energyhasreducedthe amount of effort it takesto consolidate financialreportsfrom a six personeffort takingtwoweeks to four people taking four days; Consumer Publicationshasreducedcostly errors causedby the manualtranscriptionof dataandhassimplifiedthe process of managing a promotionalmailing; management at Crockett,usingits newoperationalsubjectarea databasesand a special query system,is now able to get previously unavailable answersto wide-ranging adhocquestions,suchas the actual effects of price changes on quarterly revenues. For the mostpart, the companies that realized majorbenefits weremotivatedby specific businessgoals rather than by conceptualarguments for data resource management or a desire to improvegeneralIS effectiveness. Lackof data standardization is a key managerial problem today.A great portion of the data maintainedby corporationstodaywasoriginally designedto meetthe needsof isolated applications, developed in dispersedor autonomous suborganizations. The resulting lack of data standardizationis a majorunderlying problem withdata,oftenmaking it difficult or impossible to shareor interpret dataacrossapplicationsystem boundaries. In the study, mostof the data-related problemsencountered by business managers surfacedwhenthey neededto combinedata from severaldifferent functionsor fromseveraldifferent organizationalgroups.Almostall of the suc-

MIS Quarterly~September1988 389

DataResource

cessful datamanagement efforts involvedat least someimprovement in data standardization(i.e., common definitions and common codes), evenif the improvements wereto only a few keydata elements,suchas customer,product,or vendorIDs. Totalstandardization is not the goal. Toogreat an emphasis on datastandardizationis a mistake. A detailed corporate-wide data modelis probably prematurein mostcompaniestoday becauseof importanthard-to-resolvedifferencesin the way datais defined,stored,or usedin differentpartsof the organization. Thesesignificant managerial and systemdifferences are a majorreasonwhy strategicdataplanningapproaches are so difficult to carryout. It is extremely costlyto completely revampthree decadesof embedded systems. Becauseof the expense, datastandardization efforts shouldfocusfirst onthe obvioushighpayoffareas havinga clear businessimpact.Thequestionof whether,for example,standardcustomeror vendor indentifiers are appropriatetargets depends entirely on the business situation. Theendgoalis not total standardizationbut only as muchstandardizationas makes sensefroma businesspoint of view. "80/20" processesare growingin importance in dataplanning. In some situations,it is possible to quickly target the key businessimpactareas. Wheretheseareasare not apparent,80/20planningprocesses provideanattractive alternativeto larger scale strategic data planningmethodologies. These80/20approaches lead quickly to the nexttargetsfor action andcanprovidea roughinterimdatamodelas a startingpoint for guidingand coordinating various data management efforts. Althoughthis researchnotesfar more"targeted" planningefforts, it is believedthat theadvantages of 80/20approaches will, andshould,leadto their growingusein the nextseveralyears. Informationdatabaseswill remainthe dominantproductfor the forseeabiefuture. Operational subject databases are not the only appropriate "product" for data management efforts. Aninformationdatabase canprovidea standardizedsourceof managerialdata drawnfrom non-standardized applications. Building information databases is usually less expensive andcan be donemuchfaster than rewriting the existing non-standardized applications. Theycan be developedusingtechnology(e.g., relational databases and fourth generation languages)most appropriate for managerialuse. In addition, information databasescan provide a significantelement of a defacto architecturefor future development work.

390 MIS Quarterly~September1988

Resourceallocation mustbalancelong-term andshort-termconsiderations.Organizations wishingto addressdata problems face a resource allocation choicebetweenimprovements to infrastructure (or datastandardization),developmentof specific newdatabases,and enhanced deliveryof the existingdataresource.Whilethere is a tendency in American businessto opt for the neartermpayoff (for example, addressingdelivery or implementing informationdatabases),the long termappearsto favor judicious investments in dataarchitecture infrastructures that will ensure greaterdatastandardization. A number of difficult organizationalissues mustbe addressed.No matter which resource allocation choicesare made,there are a number of organizationalissuesthat mustbe addressed for successfulimplementationof data managementefforts. Themanagerialpredilection for short-termresults, the centralizing tendencyof data management, and the impact of data management onthe IS culture, as well as on the responsibilitiesof line managers, all presentissues that, if notmanaged well, will severelyinhibit the effectivenessof data management efforts. Treating aninitial forayinto datamanagement as a processof introducinginnovationcanhelp managers understand whichefforts are practicablein their organizationalenvironment.

Acknowledgement This researchwassupportedby the MITCenter for Information SystemsResearchand the IBM Program of Supportfor Educationin the Managementof InformationSystems.

References Appleton, D.S. "BusinessRules: The Missing Link," Datamation (30:16), October1984,pp. 145-150. Appleton, D.S. "The ModernData Dictionary," Datamation (33:5), March1, 1987,pp. 66-68. Chert, P.P.S. "TheEntity-Relationship Model-Towarda Unified Viewof Data," ACM Transactions on Database Systems (1:1 ), March1976, pp. 9-36. Chert, P.P.S.(ed.) Entity-RelationshipApproach to InformationModelingand Analysis, NorthHolland, Amsterdam, 1983. Coulson,C.J. "PeopleJust Aren’t UsingDataDictionaries," Computerworld (16:33), August16, 1982,pp. ID15-22. Curtice, R.M."Gettingthe Database Right," Datamarion(32:19), October1, 1986,pp. 99-104.

DataResource

Date,C.J. An Introductionto Database Systems, 3rd edition, Addison-Wesley,Reading,MA, 1981. Diebold, J. "Information ResourceManagement--The NewChallenge," Infosystems (26:6), June1979,pp. 50-53. Durrell, W."ThePolitics of Data,"Computerworld (19:36), September 9, 1985,pp. ID25-36. Edelman,F. "The Management of Information Resources:A Challengefor AmericanBusiness,"MISQuarterly(5:1 ), March 1981,pp. 1727. EDPAnalyzer."MakingBetter Useof YourData," (24:8), August1986. Gillenson,M.L. "Trendsin DataAdministration," MISQuarterly(9:4), December 1985,pp. 317325. GUIDE International Corporation."Establishing the DataAdministration Function,"Chicago, IL, 1977. Holland, R.H. "Tools for Information Resource Management," presented at The GUIDEConference, NewOrleans,LA, November 9, 1983. Horton, F.W., Jr. Information ResourcesManagement, Prentice-Hall,Englewood Cliffs, NJ, 1985. IBM, Business SystemsPlanning, IBMManual #GE20-0527-3, July 1981. Kahn,B.K. "Some Realities of DataAdministration," Communications of the A CM(26:10), October 1983,pp. 794-799. Kahn,B.K. andGarceau,L.R. "A Developmental Modelof the DatabaseAdministration Function," Journalof Management InformationSystems(1:4), Spring1985,pp. 87-101. Martin, J. Strategic Data-PlanningMethodologies, Prentice-Hall, Englewood Cliffs, NJ, 1982. Rogers, E.M. Diffusion of Innovations, Free Press,NewYork, 1962. Ross,R.G. DataDictionaries and StandardData Definitions: Concepts and Practicesfor Data ResourceManagement,Amacon,NewYork, 1981. Tillman, G.D. "WhyDataAdministrationFails," Computerworld (21:36), September7, 1987, pp. 73-76. Weldon, J.L. "WhoOwns the Data?"Journalof Information SystemsManagement (3:1), Winter 1986.

Aboutthe Authors DaleL. (3oodhue is an assistantprofessorof MIS at the Universityof Minnesota’s CarlsonSchoolof Management. Hereceived a BSfrom Brown,an MBA from CarnegieMellonanda Ph.D.fromMIT. Heis currently conductingadditional data management casestudies focusedon planningmethods, data architectures, and newroles and responsibilities for managing data. Other researchinterests include measuring the impactof IS and supporting end-usercomputing.Hehas also workedas a businessanalyst for American Management Systems. Judith A.(;luillard is associatedirector of the Center for Information Systems Research (CISR),MITSloanSchoolof Management. She continuing to researchdata management issues andhasbeeninvolved in studies of the managementof end-usercomputing,including a major surveyof IS anduser managers, whichaddressed suchissuesas top priorities for IS management anduser demand for newsystems.Shereceived her master’sdegreefrom the SIoanSchooland holds a bachelor’s degreein mathematics from TuftsUniversity.Prior to joiningCISR in 1979,she worked as a softwareengineerat a high technologyfirm andas a technicalproject manager for a systemsconsultingorganization. JohnF. Rockart is directorof the Centerfor Information SystemsResearch(CISR) at the MIT SloanSchoolof Management. Hehas conducted researchandpublishedarticles on manykey informationsystems issues,includingthe "critical successfactors" concept, the management of end-user computing, the useof informationby top management, andthe role of the ISexecutive.His recently publishedbook, ExecutiveSupportSystems: The Emergence of Top Management Computer Use(with D.W.DeLong), presents the results froma studyof computeruseby senior management in morethan 30 companies.Currently heis supervisinga majorCISRproject on the impactof informationtechnology onorganizational structure andprocesses.Hereceivedhis BAfrom Princeton,MBA from Harvard,andPh.D. from MIT.

MISQuarterly~September1988 391

DataResource

Appendix Animportantquestionis whetherthe type of data management problemsexperienced or whetherthe most effective approaches differ fromoneindustry groupto another.Givenonly 20companies and31. data management efforts, it is impossible to give a definitive answer.However, the different industrygroupshave beencompared usingthe dimensions of the framework. Thetable below,for example, showsthe distribution of data management productsfor the four industrygroupsfromTable1. "Product" Operational SADBs

Common Systems

Financial Services

1

0

2

0

1

Manufacturing

3

0

5

3

2

Process

2

2

2

0

3

Other

1

0

2

1

1

Industry

Information Databases

Access Services

Architectural Foundations

Withthe possibleexceptionof the processindustrygroup,thereare no clear differencesamong the groups. Theprocessindustry grouphadthe only newcommon systemsefforts anda higher incidenceof efforts aimedat architectural foundations.Thesedifferencescouldeasily be dueto chancevariationswithin the groups,to the the smallness of the sample,or mightreflect specialproblems or concerns for that group.A similar analysisof the distribution of the threeoverall patternsby industrygroup(shown by the arrowsin Figure4) also showsnoclear differences.

392 MIS Quarterly~September1988