Davenport, sistemas de información

A new approach to information promises business benefitsthat few managers could conceive of when focusing strictly

on technology.

Saving IT'sSoul:Human-Centered

nformation Managementby Thomas H. Davenport

Information technology has a polarizing effect onmanagers; it either bedazzles or frigbtens. Tbosewho are afraid of it shun it, while bedazzled IT de-partments frequently become prisoners of theirown fascination, constructing elaborate teebnologyarcbitectures and enterprise information modelsto guide systems devel-opment. Senior executiveswbo buy into tbis viewpromote technology as thekey eatalyst of businesscbange. But sucb techno-cratic solutions often spee-ify the minutiae of ma-chinery while disregardinghow people in organiza-tions actually go about ac-quiring, sharing, and mak-ing use of information. Inshort, they glorify infor-mation technology and ig-nore human psychology.

It sbouldn't surpriseanyone tbat buman na-ture, good and bad, cantbrow a wreneb into thebest-laid IT plans, yet tecb-nocrats are constantly

caugbt off guard by the "irrational" bebavior of"end users." In fact, people who are afraid of infor-mation technology may have good reason to feelthat way. Companies that ballyhoo their latestmanagement information systems or groupwareusually spend little time training employees to use

them. Even those who likecomputers can find them-selves hobbled by the rig-id structure and rules ofmany IT shops.

Obviously, people han-dle information in anynumber of ways, from ba-sic data processing to gen-

People handle information in myriad ways-fromdata processing to exchanging E-mail worldwide.

Thomas H. Davenport is apartner and director of re-search at Ernst &> Young'sCenter for Information Tech-nology and Strategy in Bos-ton and an adjunct profes-sor at Boston University'sSchool of Management. Heis the author of Process Inno-vation: Rcenginecring WorkTbrougb Information Tech-nology and two previousHBR articles.

DRAWINGS BY PAUL MEISEL 119

INFORMATION MANAGEMENT

erating sophisticated accounting documents to ex-changing informal E-mail messages around theworld. For the many diverse information users inlarge organizations, only one thing is certain: effec-tive information management must begin- bythinking about how people use information-notwith how people use machines. While it's impossi-ble to account for all tbe unforeseen consequencesof information expansion and use in today's com-panies, tbe following three observations exemplifyhow a human-centered approach to informationmanagement contrasts with the standard IT view:D Infonnation evolves in many directions, taking

Too many managers still believethat once the right technology is

At one large pharmaceutical company, for ex-ample, IT managers tried to implement shareddatabases and otber new technologies to speed upR&D, only to bave their efforts foiled by significantcultural barriers. In tbis case, managers assumedtbat researcbers involved in the development of adrug would pass along all information about it tothe people conducting its clinical trial; if re-searcbers had found early on that, say, the drug's ef-feet diminished when taken with certain foods,then patients in the clinical trial could be instruct-ed not to take tbe drug at meals. Sucb early releaseof data, however, rarely happens at this pharma-

ceutical company. Clinical studiestherefore often have to be redone,delaying tbe drug-approval processsometimes for years.

In tbis company, managementpusbed tbe new databases and soft-

p l a c e , a p p r o p r i a t e i n f o r m a t i o n ware, but researcbers were eithersharing will follow.013 multiple meanings. While IT specialists aredrawn to common definitions of terms like cus-tomer or product, most information doesn't eon-form to such strict boundaries. Forcing employeesto come to one eommon definition, as some tech-nologies require, only truneates tbe very conver-sations and sbaring of perspectives tbat tbe tech-nology is supposed to ensure. Rather than forcingemployees to simplify information so that it will fitinto a eomputer, a human-centered approach to in-formation calls for preserving the rich complexitywe prefer in our information diets.D People don't share information easily. Assumingthat different departments, professionals, or lineworkers will want to use technology to share infor-mation is one of the biggest mistakes executivesmake. Yet it is one of the fundamental assumptionsmade in planning any IT system. That is, if youbuild it, people will use it.nChanging an IT system won't change a com-pany's information culture. The presence of tech-nology, in and of itself, eannot wholly transforma corporation. Changing a company's informationculture requires altering the basic behaviors, atti-tudes, values, management expectations, and in-centives that relate to information. Changing thetechnology only reinforces the bebaviors tbat al-ready exist. Yet in most companies, many managersstill believe tbat once tbe rigbt tecbnology is inplace, tbe appropriate information-sharing behav-ior will inevitably follow.

hostile or apathetic. The IT depart-ment was so focused on the technol-ogy that they had failed to under-stand the rigid rules of scientific

exploration that govern how scientists think aboutinformation. Different departments couldn't agreeon what constituted a "drug" or a "clinical trial"-or even what font they should use for research re-ports. In this case, the rate of teehnologieal changefar outstripped the paee of ebange in tbe culture asa wbole. Instead of instituting new technologies,executives should have instituted a program of cul-tural change to convince highly competitive scien-tists that they wouldn't be penalized for sbaringearly and perbaps incomplete results.

Tecbnology, after all, is neitber the savior norarchdemon of the information age. At its worst, itdistracts and misleads us. But at its best, new sys-tems can support tbe kind of information use thatresults in real business change.

What's Wrong with the View from IT?Since the first business applications of computers

in tbe mid-1950s, planning and eontrol bave domi-nated systems development in large companies. Inparticular, the concept of "information architec-ture" bas oversbadowed a buman-centered view ofinformation. IBM created tbe first structured ap-proacb in tbe 1960s and bas defined tbe field eversince. Originally named "business systems plan-ning" (BSP), later versions came to be called "strate-gic data planning" and "information architecture."

Tbe analogy to an arcbiteetural blueprint, inwhich the location and uses of different rooms are

120 HARVARD BUSINESS REVIEW March-April 1994

specified, works as far as it goes. But informationarchitecture was invented to specify computer sys-tems and databases unambiguously. Systems plan-ners believed that information environments couldbe designed for the entire organization, without ref-erence to particular individuals. Many plannersstill assume that organizations have a core of in-

computer-based informationto make decisions.

ably consistent: managers get two-thirds of theirinformation from face-to-face or telephone con-versations; they acquire the remaining third fromdocuments, most of which come from outside theorganization and aren't on the computer system.^

When technical approaches to information plan-ning are applied broadly, not only do they fail to en-

compass all of a company's informa-1 ^ 1 tion, they also undercut business

Most managers don t rely on change Rank xerox u K , for m-^ ^ stance, began a major effort in the

late 1980s to redesign its businessprocesses with the help of informa-tion architectural techniques andcomputer-based modeling tools. Theidea behind this was that once the

new business processes were designed, then thevery same models could be used to generate codeautomatically for a new set of supporting informa-tion systems and databases.

After several years, however, a new managing di-rector asked for a simple model that could describethe old and redesigned processes. Not one could befound; all tbat existed were very detailed data mod-els that reflected the status quo. The technicianshad lost the objective of business cbange in thedetails of modeling. Now Rank Xerox uses simplerapproaches to do process modeling, such as flowcharts and cost buildup charts, and has made somesuccessful changes; for example, it has saved $11million annually in sales-order processing by elimi-

variant pieces of information-such as customers,products, and business transactions - around whichkey systems can be developed.

This approach has several potential strengths.Such blueprints attempt to structure the sharing ofdata across multiple computer applications. In ad-dition, since information storage has been a scarceresource until recently, executives hoped that in-formation architecture would belp minimize re-dundant data. And one nontechnical benefit hasbeen widely touted: after a successful planning ex-ercise, executives can supposedly make decisionsbased on common information.

But information architecture has never achievedits promise. Enterprise models of informationtypes, uses, and responsibilities aretoo broad and arcane for nontechni-cal people to comprehend-and theycan take years to build. One study ofenterprisewide BSP efforts foundthat few of the systems projectsidentified in the plans were everimplemented; another concludedthat most strategic data plans wereshelved without implementation.'Given today's rate of businesschange, even if an enterprise model is finished in ayear or two, it's likely to be outdated.

The primary reason for information architec-ture's failure, however, is that few companies haveundertaken such planning with any concern forhow people actually use information. [See the in-sert, "The Information Facts of Life.") For onething, most approaches have addressed only a smallfraction of organizations' information-that foundon computers. Yet evidence from research conduct-ed since the mid-1960s shows that most managersdon't rely on computer-based information to makedecisions. The results of these studies are remark-

Information managers mustbegin by thinking about how

people use information, not howpeople use machines.

nating approval steps and better integrating thesales force with the entire order-management pro-cess. Now the company uses information architec-ture only to design specific systems.

As at Rank Xerox and other large companies, in-formation architects have assumed that commoninformation is created through the development ofa computer model instead of through the long andoften arduous process of reaching a shared under-standing. They haven't identified, trained, or moni-tored the desired behaviors for information usersand providers, both of whom must cooperate ifcommon information is to be developed. Most im-

HARVARD BUSINESS REVIEW March-April 1994 121

The information Facts of Life1. Most of the information in organizations - and mostof the information people really care about-isn't oncomputers.

2. Managers prefer to get information from peoplerather than computers; people addvalue to raw information by inter-preting it and adding context.

3. The more complex and detailedan information management ap-proach, the tess likely it is tochange anyone's behavior.

4. All information doesn't have tobe common; an element of flexi-bility and disorder is desirable.

5. The more a company knows andcares ahout its core business area,the less likely employees will be toagree on a common definition of it.

6. If information is power and money, people won'tshare it easily. J

7. The willingness of individuals to use a specified in-formation format is directly proportional to how

much they have participated indefining it or trust others who did.

8. To make the most of electroniceommunications, employees mustfirst learn to communicate face-to-face. I

9. Since people are importantsources and integrators of infor-mation, any maps or models of in-formation should include people!

If information is power and money,people won't share it easily.

10. There's no such thing as infor^mation overload; if information isreally useful, our appetite for it isinsatiable.

portant, they make the unrealistic assumption thatmost of a company's information can be organizedaccording to a few common terms.

A Natural Mess: Multiple InformationMeanings

while information architecture can't capture thereality of human behavior, the alternative is hardfor traditional managers to grasp. That's because ahuman-centered approach assumes information iscomplex, ever-expanding, and impossible to con-trol completely. The natural world is a more aptmetaphor for the information age than architec-ture. From this holistic perspective, all informationdoesn't have to be common; some disorder andeven redundancy may be desirable. (See the chart,"Human-Centered IT Managers Focus on How Peo-ple Use Information Rather than Machines.")

No matter how simple or basic a unit of informa-tion may seem, there can be valid disagreementsabout its meaning. At Digital Fquipment Corpora-tion, for example, a "sale" to the indirect marketingorganization happened when a distributor or re-seller ordered a computer; but to direct marketing,the sale occurred only when the end customer tookdelivery. Even within direct marketing, there were

differences of opinion; salespeople recorded a salewhen the order was placed, manufacturing and lo-gistics when the product was delivered, and financewhen it was paid for.

At American Airlines, there are several perspec-tives on what an "airport" is. Some managers arguethat an airport is any location to which Americanhas scheduled service; others count any airportgranted that status by the international standardsbody. At Union Pacific Railroad, there's little con-sensus on what a "train" is. Is it a locomotive, allcars actually pulled from an origin to a destination,or an abstract scheduling entity? Even U.S. Depart-ment of Agriculture officials can't agree on themeaning of "farm."

These multiple meanings make the job of infor-mation management treacherous at best. At one oilexploration company, for example, information ar-chitects worked for years on ineffective modelsbecause people assigned different meanings to "oillocation." Some users defined it as the originalgeographic coordinates in the ground; othersthought it was the well from which oil sprang; stillothers used the term to refer to the oil's current lo-cation in a tank farm or pipeline. Each definitionfound its way into computer databases. As a result,it was difficult to share even the most basic infor-

122 HARVARD BUSINESS REVIEW March-April 1994


mation on the production of different sites. Amongmany other problems, the company couldn't accu-rately monitor the performance of specific wells orfigure the taxes it owed states and counties wherethe oil was pumped.

In this case, the CEO finally dictated to the en-tire management team what "oil location" wouldhenceforth mean: an official corporate algorithmthat reflected drilling location, well angle, and drilldepth. Those managers or other employees whoused alternative meanings would lose their jobs.Although this solution is extreme, it did achievethe desired result: consensus on the meaning of oillocation and better information on production thatcould be shared.

But while multiple meanings can create prob-lems for organizational integration and informa-tion sharing, they shouldn't always be eliminated,especially in large companies with diverse busi-nesses. In fact, given the importance of informationto the success of individuals and groups withinorganizations, managers should expect pressures todefine information in ways that are useful to thesesmaller units. There will always be a healthy ten-sion between information globalism, which seeksto create meanings that apply to an entire organiza-tion, and information particularism, in which indi-

viduals and small groups define information inways that make sense to them.

Another large computer company exemplifiesthe natural tension between particularism andglobalism. This company is renowned for grantingautonomy to product and geographical units. Thatautonomy extends to information; wben it comesto financial information, for example, there are 103general ledgers. Divisional, geographical, and prod-uct executives can therefore count costs, revenues,and profits in ways that are most meaningful fortheir particular products or businesses. To dealwith aggregation, this company maintains a corpo-rate-level ledger to consolidate results across com-mon financial categories.

Undoubtedly, such particularism turns aggrega-tion and information sharing into a challenge. Eventhough there is a corporate-level informationstream, managers are often evaluated by compar-ing their financial results against that corporatestream. Much effort goes into reconciling and ex-plaining how the local stream relates to the corpo-rate stream. Finance managers keep trying to re-move as many entries from ledgers as possible andcoaxing local executives into using corporate-levelinformation when they can. Some top managers areactively trying to get rid of the local ledgers alto-

How People Use Information Rather Than Macnii

Information orchitectures:

Focus on computerized data

Emphasize information provision

Assume permanence of solutions

Assume single meaning of terms

Stop when design is done or when system is built

Build enterprisewide structures

Assume compliance with policies

Control users' information environments

Human-centered approaches:

Focus on broad information types

Emphasize information use and sharing

Assume transience of solutions

Assume multiple meanings of terms

Continue until desired behavior is achieved enterprisewide

Build point-specific structures

As5ume compliance is gained over time through influence

Let individuals design their own information environments

D O


No unit of information is too basic to prevent disagreement about its meaning:USDA officials can't even agree on what a farm is.

which the entire husiness shouldbe run. The executives debatedthe issue on several occasions hutweren't ahle to reach a consensus.They did agree, however, thattheir main priorities were cus-tomer, financial, and product in-formation - in that order.

Xerox's IT department thentook another tack. From aroundthe world, 15 marketing and salesmanagers, accompanied by theirIT counterparts, met to agree onthe set of common customer in-formation the company woulduse. As usual, people disagreedahout what "customer" meant.But these managers eventuallyagreed to define customers as cor-porations that had already pur-chased products or services fromXerox and to refer to them with acommon worldwide numher; theyalso reached consensus on 11 oth-er customer-oriented terms, in-cluding customer-satisfactionmeasures. This coordinated ap-proach allowed country managersto then create customer informa-

gether. But while dual information streams aremessy and hard to control, they seem realistic forthis diverse company.

A larger managerial barrier, however, remains:operating with multiple meanings also requires ha-sic changes in behavior-not only for informationproviders, who categorize and collect the informa-tion, but also for users. The CEO who is annoyedwhen told there's no quick answer to how manycustomers (or employees or products) tbe companyhas is just as guilty of oversimplifying informationas the datahase designer who insists on one defini-tion of customer.

And when it is necessary to define commonmeanings, the process requires much more man-agement participation and time than many assumeor want to allot. For instance, Xerox did data mod-eling and administration for 20 years, hut in thewords of the director of information management,"We got nowhere." These initiatives were drivenby IT rather tban by senior husiness managers; tbeywere always abandoned in favor of specific develop-ment projects like the new order-processing orbilling system, which yielded obvious henefits.

Finally, Xerox's IT department asked senior exec-utives to identify the key pieces of information on

tion that tbe IT department has now combined intoa global data warehouse.

The Trouble with InformationSharing

In today's competitive business environment, itmakes sense to give information particularism itsdue; hut as Xerox's experience with customer in-formation illustrates, executives must also decidewbicb aspects of a company's information are glob-al. More to the point, executives must determinebow sucb information is to be shared effectively-one of the trickiest management issues for today'scompanies. While information architecture canspecify who controls information, such rigid mod-els don't account for tbe unpredictable growth ofinformation or human nature.

Some managers are quick to point out tbe ob-vious difficulties with information sharing, espe-cially wben it's driven hy new tecbnologies likeelectronic mail. If sharing makes it easier for acompany's employees to get at critical information,it also opens the way for any interested externalparties-competitors, attorneys, even computerhackers. Given the many recent and highly visible

124 HARVARD BUSINESS REVIEW March April 1994


cases of departing employees allegedly taking withthem reams (or diskettes) of proprietary informa-tion, many executives wonder whether or not suchinformation should he widely disseminated in thefirst place.

Paranoia ahout external opportunists has itsroots in practical information issues. For informa-tion to he shared, it must first be structured and

Paranoia about disseminationhas its roots in practicalinformation issues.compiled, which makes it easier to steal or sub-poena. For example, wben Otis Elevator hegan tocompile information on elevator reliability and per-formance-wbich would enable sharing amongmanagers, service personnel, and new product de-signers-the company's internal counsel fearedbaving to produce this information if the companywere sued for an elevator-related accident. Thishasn't happened so far, hut it's all too easy to under-stand this attorney's concerns. Ironically, his re-sponse exposes some of tbe old-line corporate atti-tudes about controlling information tbrougbsecrecy and ambiguity.

Indeed, the internal problems that arise with in-formation sharing bave the most impact on compa-nies and are much less obvious than externalthieves and ex-employees witb a grudge. Mergersproduce some of tbe most visihleclashes, since managers from compa-nies with sometimes very differentattitudes toward information use of-ten find themselves thrown togeth-er. For example, a numher of con-tentious issues surfaced at CbemicalBank sbortly after it merged withManufacturers Hanover.

Tbe two banks bad very differ-ent information cultures. ChemicalBank favored sharing information across depart-ments and product groups. Manufacturers Hanoverbelieved tbat each group owned its information andcould choose not to share it. To help integrate bank-ing operations, senior executives decided to createa basic set of information management principles, aprocess that allowed managers of both hanks to dis-cuss which policy sbould prevail.

One draft principle stated tbat if a business areahad a legitimate need for information, it sbould get

it. But managers from tbe different banks first dis-agreed ahout access to sensitive information-would they he breaching botb customer securityand trust? And what was a "legitimate need," any-way? For instance, should the private bankinggroup furnish information on wealthy customers tothe capital markets division, which could tben pro-mote a bond offering to them? If so, wbich of the

groups were responsible for identify-ing likely prospects, notifying theappropriate managers, and output-ting tbe customer information intbe correct format for the capitalmarkets division?

Otber Chemical Bank principlesaddressed tbe need for a clear ownerfor each major piece of information

and clarified responsibilities and priorities for sup-plying information to other parts of the bank.Tbese information management principles aren'tmagic, hut they've hastened the integration of thetwo banks and limited disagreements about impor-tant information issues. As with so many human-centered information management techniques, theprocess of developing principles-of hammeringout how information is defined and distributed-was more important than any fixed result.

In this case, bank executives were well aware ofwbat made information sharing such a touchy suh-ject. But consider a less successful example: the ITmanagers of a large telecommunications companygenerated an admirable set of their own informa-tion management principles. Tbey addressed theneed to estahlisb "enterprise information" and tbe

Many people suffer from far toomuch noninformation rather

than the "information overload"they complain about.

ways sucb corporate information should be man-aged and shared. But while corporate senior man-agers reviewed these principles, divisional headsweren't consulted. As a result, several divisions de-cided they were separate "enterprises" and couldtherefore define their own information.

Such natural power plays, malicious or not, arelegion. The will to power-wbether that applies toCEOs, separate divisions, line supervisors, or indi-vidual prof essionals - is the main reason wby new



information technologies don't inevitably lead toflattened hierarchies and empowered employees.Working out information issues in a company witha monolithic culture-instead of wrestling withtwo competing information cultures that resultfrom a merger-often involves digging out en-trenched attitudes toward organizational control.

In such companies, technolo-gies that promote informationsharing can end up controllingemployees rather than empower-ing them. When lower level work-ers are ordered to "share" infor-mation with those higher up thecorporate ladder, a cutthroat infor-mation culture of meddling micro-management can result. At therefining and marketing division ofa large oil company, for example,the division president delighted inheing ahle to use his computer topeer electronically over the shoul-ders of oil traders-and occasion-ally to override or initiate a deal.

On the other hand, Xerox's ex-ecutive support system has beenlimited to accessing data two lev-els below the user-precisely toavoid this type of excessive con-trol. Such human-centered tech-nology implementations are stillrare, but they indicate the waymanagers must think about theissues that information sharingbrings to the surface.

Populist exhortations to thecontrary, unlimited informationsharing doesn't work. In fact, in-creased information sharing caneither improve or actively harmcompany morale. Sharing infor-mation about actual corporateperformance is usually good formorale-even when performanceis poor, since uninformed employ-ees often assume that it's worsethan it really is. Sharing rumors,however, can be demoralizing.

An information systems man-ager at a New York bank, forexample, created a Lotus Notesbulletin board that he called the"Rumor Mill." The system al-lowed employees in his depart-ment to share rumors easily; the

When Chemical Bank andManufacturers Hanover merged,

two information cultures clashed.

manager could then quash false ones on-line. Thisexperiment worked just fine-until rumors wereposted about the manager's own departure from thebank. When he refused to comment, employeescorrectly surmised it was true. They became cyni-cal about this attempt to share information throughtechnology, since the manager hadn't communicat-

ed with them on this particularpiece of information. Needless tosay. Rumor Mill was not contin-ued by his successor.

Sharing rumors in this fashionunderscores the distinction be-tween information and noninfor-mation. Many people suffer fromfar too much noninformation-which companies seem to gener-ate with ease and at the expense ofuseful information - rather thanthe "information overload" theycomplain about. Any heavy E-mail user can testify to the junkmail problem. Right now I havemore than 160 messages in myelectronic mailbox, some of whichinform me that one colleaguelost his appointment book or thatanother wanted to be includedin last Thursday's pizza run. Ishould never have received them,and now I don't have the timeto delete them.

Technologists are working onpersonalized filters or "agents"that can separate real informationfrom junk. But it's likely that goodmarketers of electronic informa-tion will find ways to circumventfilters-just as direct mail nowlooks like a tax refund or personalcheck. In fact, some communica-tion technologies just exacerbatethis problem.

At Tandem Computers, for ex-ample, a combination E-mail/hul-letin board allows field-servicepersonnel to send a "has anyoneseen this problem?" message to alltechnical people in the company.The service technician may get ananswer, but is it really necessaryfor everyone to read this message?As in so many other cases, simplyimplementing an electronic-mailsystem-without any guidelines


for how to use it-won't resolve the complicated is-sues of information sharing and management.

If some companies generate noninformationthrough E-mail, others rely on it too much to com-municate real information. Although such tech-nologies can improve organizational communica-tions, they have their limits. Several researchershave argued persuasively that the organization-al trust and interpersonal context necessary toachieve a true network organization aren't hasedsolely on electronic networks.' Rather, relation-ships must be initially constructed through face-to-face meetings.

Symantec Corporation, for instance, found thatelectronic mail is not all it's cracked up to he. AtSymantec, a California software company thatgrew rapidly through acquisition and ended up withrelatively autonomous product groups, there wassubstantial use of electronic mail. Indeed, seniorexecutives helieved E-mail was the fastest way toforge connections in this virtual corporation. Butsenior managers soon realized their diverse organi-zation still didn't communicate very well. Theyconcluded that people in geographically far-flungproduct groups just didn't see each other enough.

To address the prohlem, executives organizedthe first companywide meeting. Managers hegancommunicating about important issues through sev-eral different routes: letters to employees' homes,face-to-face conversations, as well as E-mail mem-os. In some cases, they made the same announce-ment across all media to make sureall employees heard essential news.The company's executives noticedsubstantial improvement in the . prohlem thereafter; employees com- l l l l Oplained less about communicationproblems, and those in the field W 3 . Vtalked ahout Symantec's overall "^strategic directions with greater un-derstanding.

New communication technologieswill certainly support information sharing whenphysical proximity isn't a possibility. But asSymantec's story shows, the proliferation of thesetechnologies has created a new problem: how tochoose among all the alternatives. A sales rep whowants to communicate with a customer can usefirst-class mail, express mail, voice mail, electronicmail, a fax, an electronic hulletin hoard, videocon-ferencing, or the telephone - not to mention a face-to-face meeting.

Few of us have a clear sense of which alternativeis most appropriate for a given communication. Butwhile using a suboptimal medium is not yet a cor-

porate crime, managers should at least acknowl-edge the confusion. And regardless of the technicalform of communication, managers must bear inmind that employees who work together still needregular personal contact.

Preparing the Cultural Ground for ITIf companies as diverse as Symantec, Cbemical

Bank, Xerox-even the oil company with the con-trolling division president-are all struggling withinformation-sharing problems, it's hecause such is-sues are unavoidable in today's global economy.What many have discovered, however, is that theirsolutions do not turn out to he particularly "scien-tific." Indeed, the solution that most reliably leadsto successful IT implementation is also the hardestone to carry out: changing an organization's infor-mation culture.

Nonetheless, preparing the cultural ground is es-sential. Two professional services companies,which I'll call Company A and Company B, illus-trate the impact information culture has on tech-nology implementation. Both companies imple-mented the same new tecbnoiogy for the samepurpose. But while one had an existing informationculture that fit the management's objectives for thetechnology, the other did not.

Company A' hadn't had a successful informationorientation in the past, and now managers decidedit was time to lead with technology. They acquired

Changing the company'srmation culture is the bestimplement IT, but it's also

the hardest to carry out.both a large number of workstations and an organi-zationwide license for a new software program thatcombined electronic mail, conferencing, and docu-ment distribution. But the company's professionalsreceived little training on how to use the new sys-tem. They also had no incentives to share informa-tion-only disincentives, especially the fear of giv-ing away their hest ideas to others, who would thenuse them to get promoted in this company's up-or-out culture.

The average professional at Company A workedwith few other employees outside his or her officeand had little knowledge of anyone else's informa-


tion requirements. The company recruited newemployees based on their willingness to work hardand their training in specific disciplines rather thanany demonstrated ability to generate new ideas andpackage them for use by others. As a result. Compa-ny A's fancy new software program was ignored andmisunderstood. Even the company's IT sponsor fortbe program now admits that professionals use thenew system mostly for E-mail, a limited applica-tion that hasn't solved the main information issues.

Company B, on the other hand, had a long historyof hiring people who were good at generating ideasand expressing them in written and verbal form.Managers showed an interest in sharing informa-tion long before tecbnoiogy was in-vented to support this task; the com-pany published regular journals andsummaries of press mentions and en-couraged its professionals to puhlishbooks and articles externally. Com-pany B also had an up-or-out culture,but a key criterion for promotionwas whether or not an individual hadcreated and disseminated new ideas in the form ofpractice bulletins, articles, or books. Most impor-tant, information managers at Company B are justthat: in addition to software and hardware, tbey fo-cus on incentives, organizational structures, hu-man support, and presentation formats as facilita-tors of good information behavior in tbe company.

As for information tecbnoiogy. Company B onlyrecently invested in a new system comparable toCompany A's. Before that, however. Company Bhad set up a simple database for key practice andclient documents; it had also created a system formeasuring the documents that were most com-monly accessed, which tben counted toward thepromotions of individual authors. I never heardanyone at Company B utter the words "informationculture"; hut hy the time an IT platform had beenimplemented, tbis company could build on andsupport a program of information sharing that wasalready in place. Now its professionals use the ex-panded software capabilities to facilitate electronicdiscussions and bave created new databases at arapid clip.

As Company A and Company B reveal, valuabletools are still tools; new technologies, no matterbow advanced, won't change anyone's behaviorwithout human intervention. In fact, we have yetto address fully the role of people in informationmanagement work, though some research has fo-cused on how information itself affects humans atwork.^ It's not even a matter of "implementing" theright information culture at the right time. The

specific solutions to information problems de-scribed below demonstrate how information cul-tures can evolve to match new organizationalneeds, becoming more human-centered, flexible,and cost-effective in the process.

Information Maps. Most large companies nowhave plenty of databases. But precisely because ofthe vast amounts of information circulating aroundorganizations, few employees know where to findwhat they really need. As obvious as it may seem,few companies have an information map that de-scribes the location and availability of the mostwidely used information. Even at IBM, founder ofbusiness systems planning and steeped in the

Valuable tools are still just tools;new technologies alone won't

change anyone's behavior.rhetoric of information architecture, executives on-ly recently realized the need for information maps.

Pointers to information in a computer or on a li-brary shelf alone are useful; but pointers to the peo-ple who own or oversee particular information areespecially vaiuahie. These people can interpret theinformation, describe its intended uses and limi-tations, and direct information seekers to otbersources. At IBM, a task force studied the use of mar-ket-oriented information throughout the entirecompany. This task force found that market infor-mation in regular computerized reports was some-times ignored by managers-something that othernon-IBM research has also suggested. What tbesemanagers really wanted was fast answers to their adhoc questions.

As a result, IBM's task force created the "Guideto Market Information," an internal catalog. Theinsert "IBM's Catalog of Information" displays asample page. This guide not only lists availablemarketing information at IBM but also the peopleor organizations responsible for that informationand how to contact them. It includes proprietarymarket research, internal and external databases towhich IBM has access, electronic bulletin boards,libraries, and internally and externally produced re-ports. In 1992, IBM printed 5,000 copies, charged in-ternal huyers at cost, and sold out.

Even so, the task force and managers still had towhittle away at old attitudes about informationsharing. Some information "owners" were initiallyreluctant to have tbeir names listed, since they

HARVARD BUSINESS REVIEW March April 1994 129


were afraid that answering questions about infor-mation would be too time-consuming. In practice,however, the extra time involved hasn't really in-terfered with anyone's job. Many of these informa-tion owners now say they learn from the questionsand comments of others. More important, IBM hassaved millions by avoiding duplication in the pur-chase of external market information.

Information Guides. Along with maps, informa-tion users need people to guide them to the rightkind of information in the first place. Librarianshave often performed this role in the past. Butwhile information owners at IBM can answer spe-cific questions, few companies have general guidesto the vast information resources availahlethroughout an organization. Once again, includingnew kinds of human support for teehnology canhelp change a company's information culture.

In 1991, Hallmark Cards's MIS managers realizedthat the company's information users were con-fused about how to access necessary data. The prob-lem was both technical and behavioral. Financial,customer, supplier, product, and other data wereburied in many different databases. In addition, ex-isting applications were hard to use and provided noinformation ahout how the data were created.

Hallmark's MIS managers therefore establishedin each business unit a new full-time position: the"information guide." These individ-uals are the primary point of con-tact for anyone at Hallmark seekingcomputer-based information. Theytranslate between user informationrequests and the IT staff who canquery datahases and get the comput-erized information that users need.Hallmark's information guides havehelped improve data access so muchthat there are now 10 guides aroundthe company. They have substantially reduced thetime it takes for employees to find the right infor-mation and to compare information across busi-ness units.

Business Documents. The form in which infor-mation is presented is also eritieal to its under-standing and use. After all, raw data is not informa-tion; and aecumulating data is not the same asinterpreting it and putting it in a usable form. Com-pany B's emphasis on documentation and presenta-tion, demonstrates how such an attitude shapes theoverall information culture. In that case, promo-tions and other financial incentives were tied to thekinds of documents professionals produced.

In general, business documents provide organiza-tion and context, and they exclude enough informa-

tion so that what remains is digestible. Focusing onwhich documents an organization needs often leadsto a more fruitful discussion than looking at broadinformation requirements or trying to pin downa term like "customer."

Several companies have begun to identify criticalinformation needs in the form of doeuments. AtDean Witter, for instance, information managers,particularly those in the central lihrary, were frus-trated hy their inability to address brokers' infor-mation needs effieiently. They advoeated hiringmore librarians, but financial executives were re-luctant to take on additional workers.

With the help of a consultant, finance managerstalked to brokers about what information theyneeded. Instead of phrasing their questions in termsof information and systems, they asked which keydocuments brokers required. As it turned out, al-most all used the same doeuments over and over.Their needs were categorized into a set of "coredoeuments," most of which were regulatory and re-porting documents from U.S. companies.

By separating the documents into three or fourindustry groups, 90% of the information needed bya typical broker fit on one CD-ROM disk. DeanWitter then created a "perfect information platter,"which was updated monthly and kept on a localarea network server. By defining common informa-

Hallmark has established"information guides"

translators between informationusers and the IT staff.

tional needs and implementing technology to sup-port what brokers were already doing. Dean Witterwas able to reduce its library staff-rather than in-creasing it as originally suggested-while greatly fa-cilitating information use.

Groupware. Groupware like Lotus Notes, NCR'sCooperation, and Digital Equipment's TeamLinksare excellent examples of less structured informa-tion-sharing technologies. This new technology al-lows teams in different locations to share docu-ments electronically, to discuss issues on-line, andto capture and distribute key information easily.

Even so, companies will fail to take advantage ofgroupware if they don't also provide adequate train-ing and human support. Indeed, groupware imple-mentation stands or falls on a eompany's iniorma-


tion culture. For one thing, groupware increases theappetite for information rather than controlling it;therefore, companies must provide both the timeand training for employees to get used to handlingmore information. For another, groupware requirespeople to manage the technology on a regular basis,not just a one-time implementation of the system.

Grand IT schemes that don'tmatch what rank-and-file userswant simply won't work.

Consider this investment bank, where LotusNotes has been installed to improve communica-tions and access to external information. TheNotes system was linked to several different exter-nal databases of information on companies andmarkets. Individual bankers could specify in gener-al terms the types of information they wanted, andintelligent software (aptly named "Hoover") wouldthen search all these databases and send news itemsand financial reports on particular eompanies ordeals to the individual desktop automatically. Anybanker who later sought information about a topicwould also find the results of all previous searches.

Information managers expected this facilitywould increase information demands and externalexpenses initially, but demands would taper offsince the information could be reused and sharedwithin the organization. They were wrong; de-mands and costs are still increasing. Yet it also ap-pears that this investment in information now sup-ports the bank's overall business goal: more andbetter deals. In this case, an unexpected result-in-creasing information use-led to a clear husinessbenefit that a limited focus on the technologycouldn't predict.

In addition, since the database searching and ba-sic communications features of Lotus Notes re-quire little human attention after initial setup - andthe system itself is easy to use - the bank's informa-tion managers planned a low level of human sup-port. But they didn't anticipate two labor-intensiveactivities eritieal to the successful use of group-ware. One is training, or more accurately, educa-tion: that is, the need to show bankers how the newtechnology can be used to create better deals andworking relationships with eolleagues and clients.

The other important task is the ongoing prun-ing and restructuring of the system's documentdatabases. Bank managers have found that this task

requires judgment and knowledge; if the systemmade decisions about, say, which documents todelete based on their age, some of the bank's mostwidely used documents might disappear.

As this use of Lotus Notes shows, even the bestnew technologies depend on a strong informationculture-one that is open, flexible, and expansive.

When executives introduce such po-tentially valuable new technologies,they must be prepared to support aninereasing appetite for information.Large appetites may mean further in-formation expenditures. But that'sthe reality of today's informationeconomy - a reality that can providebetter deals, investments, or product

planning, as well as new costs, technical require-ments, and all-too-human complications.

Some managers have always been distrustful ofthe information systems approaches of their com-panies, largely because they didn't understand them.In many cases, they were right to feel uneasy. Asthe diverse company experiences suggest, grand ITschemes that don't match what rank-and-file userswant simply won't work.

It certainly doesn't hurt for executives to under-stand communications networks, complicateddatabases, and the latest groupware. But preciselyhecause of the enormous financial resources in-volved, we must abandon the idea that technologyin and of itself ean solve a company's informationprohiems. No matter how sophisticated an IT sys-tem, information is inherently hard to control. It'sonly when executives stop being "technologicallycorrect"-when they start viewing information asever expanding and unpredictable-that they real-ize how little the latest computer application has todo with effective information use.

References1. Albert L. Lederer and Vi|ay Sethi, "The Implementation of StrategicSy5tem.s Planning Methodologies," MIS Quaneily, September 1988, pp.445-461; also Dale L. Goodhue, Laurie J. Kirsch, Judith A. Quillard,Michael D. Wybo, "Strategic Data Planning: Lessons from the Field, "M/SQvaneily, March 1992, pp.11-34.2. Sharon M. McKinnon and William |. Brims, The information Mosaic(Boston: Harvard Business School Press, 1992], pp. 162-164.3. Nitin Nohria and Robert G. Eccles, "Face to Face: Making Network Or-ganizations Work," in Networks and Oigcinizations, Nohria and Eccles,eds. (Boston: Harvard Business School Press, 1992). The authors also col-laborated on a Harvard case study about Symantec Corporation that illus-trates the problem.4. Thiscompany'ssituatinn was first described by Wanda ! Orlikowski in"Learning from Notes: Organizational Issues in Groupware Implementa-tion" [Center for Information Systems Research Working Paper, No. 241,MIT Sloan Schooi of Management, May 19921-5. See Shoshana Zuboff's work on "informating" jobs in in the Age of theSmart Machine (New York: Basic Books, 1988).

Reprint 94203


Harvard Business Review Notice of Use Restrictions, May 2009

Harvard Business Review and Harvard Business Publishing Newsletter content on EBSCOhost is licensed forthe private individual use of authorized EBSCOhost users. It is not intended for use as assigned course materialin academic institutions nor as corporate learning or training materials in businesses. Academic licensees maynot use this content in electronic reserves, electronic course packs, persistent linking from syllabi or by anyother means of incorporating the content into course resources. Business licensees may not host this content onlearning management systems or use persistent linking or other means to incorporate the content into learningmanagement systems. Harvard Business Publishing will be pleased to grant permission to make this contentavailable through such means. For rates and permission, contact [email protected].