knowing what we know: supporting knowledge … - knowing what we... · knowing what we know: ......

21
Knowing What We Know: Supporting Knowledge Creation and Sharing in Social Networks ROB CROSS ANDREW PARKER LAURENCE PRUSAK STEPHEN P. BORGATTI Crafting an Answer “So the call came in late on Thursday after- noon and right away I wished I hadn’t answered the phone. We had received a last-second oppor- tunity to bid on a sizable piece of work that the Partner on the other end of the line really wanted to pursue. I had no clue how to even begin looking for relevant methodologies or case examples, so my first move was to tap into my network to find some relevant info and leads to other people or databases. And I relied pretty heavily on this group over the next couple of days. Seth was great for pointing me to other people and relevant in- formation, Paul provided ideas on the technical content of the project while Jeff really helped in showing me how to frame the client’s issues in ways that we could sell. He also helped navigate and get buy-in from the client given his knowl- edge of their operations and politics... I mean the whole game is just being the person that can get the client what they need with [the firm’s] re- sources behind you. This almost always seems to mean knowing who knows what and figuring out a way to bring them to bear on your client’s issue.” —Anonymous Interviewee T he way in which this manager relied on his network to obtain information and knowledge critical to the success of an im- portant project is common and likely reso- nates with your own experience. Usually when we think of where people turn for information or knowledge we think of data- bases, the Web, intranets and portals or other, more traditional, repositories such as file cabinets or policy and procedure manu- als. However, a significant component of a person’s information environment consists of the relationships he or she can tap for various informational needs. For example, in summarizing a decade worth of studies, Tom Allen of Massachusetts Institute of Technol- ogy (MIT) found that engineers and scien- tists were roughly five times more likely to turn to a person for information than to an impersonal source such as a database or a file cabinet. In other settings, research has con- sistently shown that who you know has a significant impact on what you come to know, as relationships are critical for obtain- ing information, solving problems and learn- ing how to do your work. Particularly in knowledge intensive work, creating an informational environment that helps employees solve increasingly complex and often ambiguous problems holds signif- icant performance implications. Frequently such efforts entail knowledge management initiatives focusing on the capture and shar- ing of codified knowledge and reusable work products. To be sure, these so-called knowledge bases hold pragmatic benefits. They bridge boundaries of time and space, Organizational Dynamics, Vol. 30, No. 2, pp. 100 –120, 2001 ISSN 0090-2616/01/$–see frontmatter © 2001 Elsevier Science, Inc. PII S0090-2616(01)00046-8 www.organizational-dynamics.com 100 ORGANIZATIONAL DYNAMICS

Upload: dinhanh

Post on 31-Jul-2018

228 views

Category:

Documents


0 download

TRANSCRIPT

Knowing What We Know:

Supporting Knowledge Creationand Sharing in Social Networks

ROB CROSS ANDREW PARKER LAURENCE PRUSAKSTEPHEN P. BORGATTI

Crafting an Answer

“So the call came in late on Thursday after-noon and right away I wished I hadn’t answeredthe phone. We had received a last-second oppor-tunity to bid on a sizable piece of work that thePartner on the other end of the line really wantedto pursue. I had no clue how to even begin lookingfor relevant methodologies or case examples, somy first move was to tap into my network to findsome relevant info and leads to other people ordatabases. And I relied pretty heavily on thisgroup over the next couple of days. Seth was greatfor pointing me to other people and relevant in-formation, Paul provided ideas on the technicalcontent of the project while Jeff really helped inshowing me how to frame the client’s issues inways that we could sell. He also helped navigateand get buy-in from the client given his knowl-edge of their operations and politics. . . I mean thewhole game is just being the person that can getthe client what they need with [the firm’s] re-sources behind you. This almost always seems tomean knowing who knows what and figuring outa way to bring them to bear on your client’sissue.”

—Anonymous Interviewee

The way in which this manager relied onhis network to obtain information and

knowledge critical to the success of an im-portant project is common and likely reso-nates with your own experience. Usually

when we think of where people turn forinformation or knowledge we think of data-bases, the Web, intranets and portals orother, more traditional, repositories such asfile cabinets or policy and procedure manu-als. However, a significant component of aperson’s information environment consistsof the relationships he or she can tap forvarious informational needs. For example, insummarizing a decade worth of studies, TomAllen of Massachusetts Institute of Technol-ogy (MIT) found that engineers and scien-tists were roughly five times more likely toturn to a person for information than to animpersonal source such as a database or a filecabinet. In other settings, research has con-sistently shown that who you know has asignificant impact on what you come toknow, as relationships are critical for obtain-ing information, solving problems and learn-ing how to do your work.

Particularly in knowledge intensive work,creating an informational environment thathelps employees solve increasingly complexand often ambiguous problems holds signif-icant performance implications. Frequentlysuch efforts entail knowledge managementinitiatives focusing on the capture and shar-ing of codified knowledge and reusablework products. To be sure, these so-calledknowledge bases hold pragmatic benefits.They bridge boundaries of time and space,

Organizational Dynamics, Vol. 30, No. 2, pp. 100–120, 2001 ISSN 0090-2616/01/$–see frontmatter© 2001 Elsevier Science, Inc. PII S0090-2616(01)00046-8www.organizational-dynamics.com

100 ORGANIZATIONAL DYNAMICS

allow for potential reuse of tools or workproducts employed successfully in other ar-eas of an organization, and provide a meansof reducing organizational “forgetting” as afunction of employee turnover. However,such initiatives often undervalue crucialknowledge held by employees and the webof relationships that help dynamically solveproblems and create new knowledge.

As we move further into an economywhere collaboration and innovation are in-creasingly central to organizational effective-ness, we must pay more attention to the setsof relationships that people rely on to accom-plish their work. Certainly we can expectemerging collaborative technologies to facil-itate virtual work and skill profiling systemsto help with the location of relevant exper-tise. However, as was so poignantly demon-strated by reengineering, technology alonecan only accomplish so much in the pursuitof business performance. Improving effi-ciency and effectiveness in knowledge-inten-sive work demands more than sophisticatedtechnologies—it requires attending to the of-ten idiosyncratic ways that people seek outknowledge, learn from and solve problemswith other people in organizations.

With this in mind, we initiated a researchprogram to determine means of improvingemployees’ ability to create and share knowl-edge in important social networks. In thefirst phase of our research, we assessed thecharacteristics of relationships that 40 man-agers relied on for learning and knowledgesharing in important projects. In the secondphase, we systematically employed socialnetwork analysis to map these dimensions ofrelationships among strategically importantnetworks of people in various organizations.Working with a consortium of Fortune 500companies and government organizations,we developed empirical support for rela-tional characteristics that facilitate knowl-edge creation and sharing in social networksas well as insight into social and technicalinterventions to facilitate knowledge flow inthese networks.

Rob Cross is an assistant professor at the

McIntire School of Commerce (University of Vir-

ginia) and a research fellow within IBM’s Insti-

tute for Knowledge Management. He currently

leads a stream of research for the Institute on

social capital, and has consulted with over 40

public and private organizations on issues re-

lating to social networks, organizational design,

and knowledge management. Rob holds a doc-

torate from Boston University’s school of man-

agement and has published extensively in ac-

ademic and practitioner publications. He

earned both his B.S. and M.B.A. from the Uni-

versity of Virginia, and lives in Charlottesville

with his wife and daughter.

FALL 2001 101

SUPPORTING KNOWLEDGECREATION AND SHARING INSOCIAL NETWORKS

In the first phase of our research we asked 40managers to reflect on a recent project thatwas important to their careers and indicatewhere they obtained information critical tothe project’s success. As can be seen in Fig. 1,these managers overwhelmingly indicated(and supported with vivid stories) that theyreceived this information from other peoplefar more frequently than impersonal sourcessuch as their personal computer archives, theInternet or the organization’s knowledgemanagement database. And we found this inan organization that most industry analystsheralded as a knowledge management ex-emplar because of its investment in technol-ogy. This is not to say that the firm’s leadingedge technical platform and organizationalpractices for capturing, screening and ar-chiving knowledge were not helpful. Just topoint out that “impersonal” informationsources were primarily leveraged only afterthe managers had been unsuccessful in ob-taining relevant knowledge from colleagues(or when directed to a point in the databaseby a colleague).

We also asked the managers to identifythe people most important to them in termsof information or knowledge acquired forthat project, and had them carefully describethese relationships. Four features emergedthat distinguished effective from ineffectiverelationships: (1) knowing what another per-son knows and thus when to turn to them;(2) being able to gain timely access to thatperson; (3) willingness of the person soughtout to engage in problem solving rather thandump information; and (4) a degree of safetyin the relationship that promoted learningand creativity. An in-depth review of thesedimensions is beyond our scope here; how-ever, a summary of these relational featuresand representative quotes can be found be-low in Table 1.

The managers we interviewed indicatedthat these four dimensions were key charac-teristics of relationships that were effective

Andrew Parker is an associate consultant with

IBM’s Institute for Knowledge Management. He

is currently conducting social network research

with 30 member companies to develop insight

into knowledge creation and sharing activities.

This research includes work on mergers and

acquisitions, top leadership teams and commu-

nities of practice. Parker has co-authored sev-

eral articles on network analysis and knowledge

intermediaries. He earned his M.Sc. from the

London School of Economics.

102 ORGANIZATIONAL DYNAMICS

for acquiring information, solving problemsor learning. In contrast, they also recountednumerous times when learning or knowl-edge sharing did not happen because of oneof the above dimensions not existing in therelationship (e.g., someone knew what theyneeded to know, but did not make himself orherself accessible). Further, a separate quan-titative study demonstrated that these di-mensions consistently predict whom peopleseek out for informational purposes, evenafter controlling for such features as educa-tion or age similarity, physical proximity,time in organization, and formal hierarchicalposition. With the importance of these fourrelational characteristics established, the sec-ond step of our research was to use socialnetwork analysis to map information flow aswell as these relational characteristics amongstrategically important groups to improveknowledge creation and sharing.

SOCIAL NETWORK ANALYSIS

Social network analysis (SNA) provides arich and systematic means of assessing infor-mal networks by mapping and analyzing re-lationships among people, teams, depart-ments or even entire organizations. Thoughmanagers are often adamant that they knowtheir organization, studies are showing thatthey have different levels of accuracy in un-derstanding the networks around them. Byvirtue of their position in the hierarchy, man-agers are frequently far removed from theday-to-day work interactions that generatean organization’s informal structure, and somay have a very inaccurate understanding ofthe actual patterns of relationships. And thepotential for inaccurate perceptions is onlyincreased by our transition into a world ofvirtual work and telecommuting, where em-ployees are engaged in work relationshipsincreasingly invisible to superiors. Social net-work analysis can provide an X-ray of theway in which work is or is not occurring inthese informal networks.

Larry Prusak is executive director of the Insti-

tute for Knowledge Management. He has ex-

tensive research and consulting experience,

within the U.S. and internationally, in helping

organizations leverage and optimize their infor-

mation and knowledge resources. He has co-

authored two books with Tom Davenport: Infor-

mation Ecology (Oxford University Press, 1997),

and Working Knowledge (Harvard Business

School Press, 1997). Working Knowledge has

sold over 60,000 copies and has been trans-

lated into 12 languages. His most recent book

is In Good Company: The Role of Social Capital

in Organizations (Harvard Business School

Press, 2000).

FALL 2001 103

Mapping Information Flow amongExecutives

We conducted a social network analysis ofexecutives in the exploration and productiondivision of a large petroleum organization.This group was in the midst of implementinga distributed technology to help transferknowledge across drilling initiatives andwas also interested in assessing their abilityas a group to create and share knowledge. Asa result, we were asked to conduct a socialnetwork analysis of information flow amongthe top 20 executives within the Explorationand Production Division. As can be seen inFig. 2, this analysis revealed a striking con-trast between the group’s formal and infor-mal structure.

Three important points quickly emergedfor this group in relation to sharing informa-tion and effectively leveraging their collec-tive expertise. First, the social network anal-ysis identified mid-level managers who werecritical in terms of information flow withinthe group. A particular surprise came fromthe very central role that Cole played interms of both overall information flowwithin the group and being the only point ofcontact between members of the productiondivision and the rest of the network. A facil-itated session with this executive team re-vealed that over time Cole’s reputation forexpertise and responsiveness had resulted inhis becoming a critical source for all sorts ofinformation. Through no fault of his own,the number of informational requests he re-ceived and the number of projects he wasinvolved in had grown excessive, which notonly caused him stress but also frequentlyslowed the group as a whole, because Colehad become a bottleneck.

The social network analysis also re-vealed the extent to which the entire networkwas disproportionately reliant on Cole. If hewere hired away, the efficiency of this groupas a whole would be significantly impactedas people in the informal network re-estab-lished important informational relation-ships. Of course, people would find ways toreconnect to obtain necessary information.

Steve Borgatti is an associate professor of

organizational behavior at the Carroll School of

Management, Boston College. He is the princi-

pal author of UCINET, the leading software

package for social network analysis, and past

president of INSNA, the professional associa-

tion for social network researchers. Dr. Borgatti

has published more than 50 journal articles in

the area of social network theory and

methodology.

104 ORGANIZATIONAL DYNAMICS

However, the social network diagram madeit very clear that if Cole left, the companywould lose both his valuable knowledge andthe relationships he had established that inmany ways were holding the network to-gether. As a result, a central intervention that

came from this analysis was to reallocatemany of the informational requests comingto Cole to other members in the group. Sim-ply categorizing various informational re-quests that Cole received and then allocatingownership of these informational or decision

TABLE 1 RELATIONAL QUALITIES THAT PROMOTE EFFECTIVEKNOWLEDGE SHARING

RelationalDimensions Impact on Knowledge Seeking Representative Quote

Knowledge Knowing what someone else knows (evenif we are initially inaccurate and calibrateover time) is a precursor to seeking aspecific person out when we are facedwith a problem or opportunity. For otherpeople to be options we must have at leastsome perception of their expertise.

“At [Company X] we had access to backgroundinformation and you know lots of case studiesand approaches that were really well writtenup. I had no experience though of actuallyapplying this approach on an engagement. Sowhat was specifically useful to me was to talkwith Terry who I knew had done several ofthese engagements. He helped me work someof the content in the database into a workableapproach. I was lucky I knew him and couldleverage some of his experience . . .”

Access However, knowing what someone elseknows is only useful if you can get accessto their thinking in a sufficiently timelyfashion. Access is heavily influenced bythe closeness of one’s relationship as wellas physical proximity, organizationaldesign and collaborative technology.

“I have gotten less frustrated the more I haveworked with him because I know how to getahold of him. It took me a while to figure outthat he was a phone guy and not an e-mailguy. And I have also learned how to ask himfor help and what I can expect. It wasimportant to learn what I could rely on him forand how to get his attention to make therelationship, which was initially frustrating, animportant one for me . . .”

Engagement People who are helpful in learninginteractions actively think with the seekerand engage in problem solving. Ratherthan dump information, these people firstunderstand the problem as experienced bythe seeker and then shape their knowledgeto the problem at hand.

“Some people will give you their opinionwithout trying to either understand what yourobjectives are or understand where you arecoming from or be very closed in their answerto you. [She] is the sort of person who firstmakes sure she understands what the issue is. Ihave been around people who give you a quickspiel because they think they are smart andthat by throwing some framework or angle upthey can quickly wow you and get out of thehard work of solving a problem. [She], for allher other responsibilities and stature within thefirm, is not like that.”

Safety Finally, those relationships that are safe areoften most effective for learning purposes.Being able to admit a lack of knowledge orto diverge in a conversation often resultsin creativity and learning.

“[he] is always looking for the positive spin onsomething. I mean even if he thinks that isgarbage and if he really thought that, he wouldmake this known but in a positive way. So hemight say “Well I think we might be a little offtrack on that and here’s why” and then saywhy and of course there is learning that comesfrom that.”

FALL 2001 105

domains to other executives served to bothunburden Cole and make the overall net-work more responsive and robust.

Just as important, the social networkanalysis helped to identify highly peripheralpeople who essentially represented un-tapped expertise and thus underutilized re-sources for the group. In particular, it be-came apparent that many of the seniorpeople had become too removed from theday-to-day operations of this group. For ex-ample, Fig. 2 reveals that the most seniorperson (Jones) was one of the most periph-

eral in the informal network. This is a com-mon finding. As people move higher withinan organization, their work begins to entailmore administrative tasks that makes themboth less accessible and less knowledgeableabout the day to day work of their subordi-nates. However, in this case our debrief ses-sion indicated that Jones had become tooremoved, and his lack of responsiveness fre-quently held the entire network back whenimportant decisions needed to be made. Inthis case, the social network diagram helpedto make what could have been a potentiallydifficult conversation with this executivenon-confrontational, and resulted in more ofhis time being committed back to the group.

Finally, the social network analysis dem-onstrated the extent to which the production

1 Total sources exceed the number of interviews be-cause several respondents indicated more than one crit-ical source of information. In this case, prior materialincluded computer or paper files or archives that theinterviewees had used in prior projects.

FIGURE 1 WHERE PEOPLE GO FOR INFORMATION.1

106 ORGANIZATIONAL DYNAMICS

FIGURE 2 FORMAL VS. INFORMAL STRUCTURE IN A PETROLEUMORGANIZATION.2

2Names have been disguised at the request of the company.

FALL 2001 107

division (the subgroup on the top of the di-agram) had become separated from the over-all network. Several months before this anal-ysis, these people had been physicallymoved to a different floor in the building.Upon reviewing the network diagram, manyof the executives realized that this physicalseparation had resulted in loss of a lot of theserendipitous meetings that occurred whenthey were co-located. In this case, the exec-utives decided that they needed to introducemore structured meetings to compensate forthis recent loss of serendipitous communica-tion (and they also adopted an instant mes-saging system to promote communication).

Beyond Information Flow

In addition to mapping information flow, wealso use social network analysis to assess therelational characteristics of knowledge, access,engagement, and safety among a group. Some-times, if we have only mapped an informa-tion network and find that certain people arenot as connected as they should be, it isdifficult to tell what to do. Simply proposingmore or better communication is the oldestconsulting recommendation in the book—and no one today really needs more meet-ings. By analyzing the dimensions of rela-tionships that precede or lead to effectiveknowledge sharing, we can offer more pre-cise ways to improve a network’s ability tocreate and share knowledge. For example, ifit is discovered that the knowledge network issparse, it might make sense to consider askill profiling system or new staffing prac-tices—technical and social interventions de-signed to help a network know what itknows. In contrast, if the access network issparse, then it might make sense to considerpeer feedback or technical means of connect-ing distributed workers (e.g., video confer-encing or instant messaging) to make surethat people within the network have accessto each other in a timely fashion.

Throughout our research we have foundorganizations employing various practices topromote these relational dimensions in im-portant networks. We have summarized

some of these initiatives in Fig. 3, and nowturn to specific case examples where we usedsocial network analysis to assess these rela-tional dimensions.

Knowledge Dimension: Do We Know WhatWe Know? Other people can only be usefulto us in solving problems if we have someawareness of their expertise. Even if we arewrong, our initial perception will determinewhether and how we turn to them for infor-mation when faced with a new problem oropportunity. The managers we interviewedin the first phase of our research reportedthat people they turned to for informationprovided a critical extension to their ownknowledge when the manager had at least asemi-accurate understanding of her or hiscontact’s expertise. As a result, assessing thisrelational knowledge of “who knows what”at a network level provides insight into thepotential for members of a network to beable to tap others with relevant expertisewhen faced with a new problem or opportu-nity.

Supporting New ProductDevelopment

For example, we analyzed a network of im-munologists in a Fortune 250 pharmaceuticalcompany. By virtue of effectively integratinghighly specialized knowledge in the drugdevelopment process, this group of peopleheld the potential to provide strategic advan-tage to the organization. However, they alsodealt with many impediments to effectivecollaboration, in that they were dispersedacross five geographic sites and four hierar-chical levels and attempting to integrate verydifferent kinds of expertise. One telling viewof this network emerged when we mappedthe knowledge relation to get a better under-standing of who understood and valuedother people’s expertise in this group.

What we found was that the knowledgenetwork was very sparse compared withothers that we had seen, indicating that animpediment to this group effectively creat-ing and sharing knowledge was that they did

108 ORGANIZATIONAL DYNAMICS

FIGURE 3 INITIATIVES PROMOTING KNOWLEDGE SHARING IN HUMANNETWORKS.

FALL 2001 109

FIGURE 3 CONTINUED.

110 ORGANIZATIONAL DYNAMICS

not know what each other knew. Two char-acteristics of this group seemed to result inthe sparse pattern. First, the group was phys-ically dispersed, which precluded serendip-itous interactions that help people learn col-leagues’ expertise and skills. Second, thegroup housed deep specialists who oftenstruggled to find overlap with their col-leagues. Stories emerging in interviews indi-cated that even when there were opportuni-ties to incorporate each other’s expertise, thiswas often not done, because one group ofspecialists did not know enough about whatanother group did to be able to “see” a wayto involve them in projects.

Conducting the social network analysisprovided several intervention opportunities.A facilitated session with leaders allowedthem to assess and discuss the relative isola-tion of the specialties, as well as morepointed concerns about certain members’ ex-pertise not being tapped while other mem-bers appeared to be bottlenecks in sharinginformation. As a result of the discussionaround this social network, various changeswere made to the group’s operations. First, avariety of internal projects—ranging fromprocess improvement to a project trackingdatabase—were jointly staffed with peoplefrom various locations. This forced people towork together and so begin to develop anappreciation of each other’s unique skillsand knowledge. Second, several new com-munication forums were created—includingweekly status calls, a short update e-maildone weekly and a project tracking databasethat helped each person keep up to date onwhat other members of the group were do-ing. Finally, some simple changes in projectmanagement practices and restructuring ofthe project leaders’ responsibilities helpedpeople to connect around them.

Facilitating Merger Integration

In another scenario we assessed the top lead-ership network of a Fortune 250 organization(i.e., top 126 executives of this conglomerate).This was an organization that had grown by

acquisition over the course of several years,with the strategic intent that acquired com-panies would combine their expertise in de-veloping and taking to market new productsand services. The chief executive officer(CEO) of this organization had becomeacutely aware of the need to create a leader-ship network that knew enough of what oth-ers in the conglomerate knew to be able tocombine the appropriate resources in re-sponse to new opportunities. As there wassome evidence that this was not happening,he asked us to conduct a social network anal-ysis of his top executives across these ac-quired organizations.

Mapping information flow amongthese executives showed that there wasonly limited collaborative activity in pock-ets of the organization, and that in generalthis lack of collaboration was a product ofpeople not knowing what other peopleknew. In fact, we found that the problemwas so significant that a key executivewould not only indicate that he or she didnot know what a specific person in anotherdivision did, but also that the executive didnot even know what that division did. De-spite alignment of the organization’s for-mal structure, this lack of collective aware-ness of “who knows what” was having asignificant impact on the organization’sability to execute strategically.

Two interventions were undertaken tobegin helping to integrate this group. First,on a technical front, a customized technologywas introduced for this group that combineda skill-profiling system with a new collabo-rative environment where executives postedproject information. This system was quicklyused, as the CEO pushed people into adopt-ing it and also made it the primary forum bywhich these executives began to get informa-tion they needed to run their business. Inaddition, action learning sets were employedon internal projects. People from across thesedivisions were staffed together on smallteams that each attacked a given project, butdid so with reflective exercises, as the pointof the initiative was both to solve problems

FALL 2001 111

for the company and to create connectionsacross the executive team.

Creating Awareness of “WhoKnows What”

Overall, we are finding that it is importantfor organizations to pay attention to howstrategic networks of employees develop anunderstanding of their collective knowledge.In more staid times, working relationshipsdeveloped as a product of interaction overlonger time periods. This is not so in today’sbusiness environment. Given the rapid turn-over many companies experience today, it isimportant to find ways to help people be-come better connected so the organizationcan get the true benefit of their expertisemore quickly. This is often a process that canbe improved by focusing on the way thatnew people are brought into a group. Gen-erally what most organizations do when hir-ing a new person is to hold orientationcourses that teach the person about the com-puter system, benefits and, perhaps, somehomilies about the culture and history of thecompany. It is rare to find practices thatteach the group what the newcomers know.This is a critical shortcoming in increasinglyproject-based work, where people will bebrought into the center of the network pri-marily as a result of what other people un-derstand about their expertise and so how totap them when new problems or opportuni-ties arise.

However, knowingly or unknowingly,some organizations we worked with wereemploying different mechanisms that builtthis awareness of “who knows what.” Forexample, on a technical front many organi-zations are implementing skill-profiling sys-tems or corporate yellow pages. On an orga-nizational front, organizations such as theWorld Bank have organized their employeesinto thematic groups that have Help Deskswhom anyone connected with the organiza-tion can contact. The individuals staffing theHelp Desks are able to route people to otherswithin the thematic group who have exper-tise on a particular subject. Other companies

and government organizations have regularKnowledge Fairs where teams, communitiesor departments can set up a booth and dis-tribute information about the expertise thatthey have. Although this has limited scope, ithas proven effective in increasing awarenessof the projects and knowledge activities tak-ing place within the different departmentsand communities of the organization.

Access Dimension: Can We Access What WeKnow in A Sufficiently Timely Fashion? Ofcourse knowing that someone else knowssomething of relevance does little good if wecannot gain access to their thinking in atimely fashion. Critical issues on which wemay turn to others for help or advice oftenrequire turnaround within increasingly tighttime frames. As with the knowledge dimen-sion, we have found it helpful to map theaccess relation at a network level to under-stand who is able to reach whom in a suffi-ciently timely fashion.

Assessing Access in a GlobalConsulting Practice

We conducted a network analysis of theglobal consumer goods practice within a ma-jor consulting firm. One of the more tellingnetworks in this analysis was the diagramreflecting who was sufficiently accessible towhom among this group of 46 people spreadthrough Europe and the U.S. Despite the en-tire practice reporting to one overall partnerand being subject to a common strategy, per-formance measurement and reward prac-tices, we found significant clustering in thenetwork when we assessed who was acces-sible to whom. The social network analysisof accessibility showed three tightly knitgroups rather than one integrated network—two in North America and one in Europe—that were all highly centralized around dif-ferent partners. In fact, only three employeesserved to bridge these fiefdoms, and thesewere not the people in charge of the group.Rather, they had been through rotating workassignments and so developed relations withmany others in the network.

112 ORGANIZATIONAL DYNAMICS

A first intervention for this organizationwas reconsidering staffing practices to helpintegrate people from the different locationson both client projects and internal initia-tives. A key concern lay with developingrelationships throughout the overall practiceto improve knowledge sharing and the loca-tion of relevant expertise for both sales ef-forts and client engagements. Further, in-creasing overall connection within thenetwork also reduced the extent to which thepractice was exposed by the potential for anyof these three central people leaving. In thisand many other examples, we have consis-tently found that a network view makes itclear that, should certain central people in anetwork leave, they take more than just whatthey know—they also fundamentally affectthe connectivity of the entire group.

The two groups in the U.S. representedanother challenge for management. It turnedout that the majority of people in these twogroups not only had offices in the samebuilding, but also were interspersed alongthe same corridor. What we discovered ininterviews was a political problem that hademerged and resulted in tensions betweentwo subgroups. While management had sus-pected there were problems, the visual rep-resentation of the network diagram clearlyshowed the extent to which these issues wereimpeding the ability of the overall group toeffectively leverage the expertise of its mem-bers. Various steps were taken to help re-solve the problem including: executivecoaching, revised performance managementpractices and an extensive off site planningsession and organizational development(OD) interventions to help the group inte-grate.

Accessibility after a Transition toTeams

Reorganizations often shift the location andconcurrently the accessibility of specific ex-pertise. For example, we worked with onecommercial lending organization in a transi-

tion from a functional to a team-based struc-ture. To minimize inefficiencies resultingfrom cross-functional hand-offs in the com-mercial lending process, the organizationshifted to a team-based structure that co-located lenders, analysts, and servicers in in-dustry teams. Before the transition, thesegroups had been housed together on differ-ent floors and so were able to tap into eachother’s functional knowledge with relativeease. With the redesign, it was far moredifficult for inexperienced people to learnthe basics of their function and also forexperienced lenders and analysts to engagein collaborative problem-solving efforts onthe more creative aspects of commerciallending (e.g., structuring a specific trans-action).

Social network analysis showed thatfour months after the transition to teams,several key people had become significantlyoverburdened, as they were heavily soughtout by both their past functional colleaguesas well as their new team members. In par-ticular, we found that the people who werereputed experts in their area were tapped foradvice to such an extent that they were fall-ing behind on their own work. While in thefunctional department these interactionswere more controlled and observable, in theteam-based environment it was difficult formanagement to see how instrumental theseopinion leaders really were to the success ofthe whole system. In fact, from a cursoryreview of their individual performance met-rics (e.g., loans serviced or loans booked)these people experienced a fairly significantdecline in productivity. Further, the longerhours that these people were working, intandem with declining individual perfor-mance metrics that influenced their bonuscalculations, served to undermine their ownmorale. As a result of these findings, severalsteps were taken—such as new staffing prac-tices, better orientation materials (to helpbring new people up to speed more effec-tively) and a reallocation of tasks withinteams.

FALL 2001 113

Managing Accessibility

Through the course of our research we havefound that many organizations struggle withthe notion of accessibility as people increas-ingly work from diverse locations. By andlarge, most solutions that companies consid-ered were technical in nature and includedsuch things as e-mail, asynchronous and syn-chronous collaborative environments, videoconferencing and instant messaging. How-ever, we generally find that organizationaldesign considerations and cultural norms arethe more powerful indicators of who is ac-cessible to whom.

Performance management systems pro-moting individualistic behaviors seem to beone of the primary drivers of sparse, discon-nected networks. Further, though more of atrait of organizational culture, we often findthat hierarchy has a marked impact on whois able to access whom. Again, this is a tellingindicator for organizations trying to becomemore flexible and effective at informationsharing. Some organizations have taken in-teresting steps to promote access across hi-erarchy, such as making knowledge sharinga part of the mission or code of ethics. AtBuckman Laboratories, all associates are em-powered to speak with any associate at anylevel, and this is supported by a communi-cation technology that gives each employeeaccess to all other employees. Others are be-ginning to turn to creative uses of physicalspace to promote both intentional and seren-dipitous interactions among high-endknowledge workers. For example, Chryslerhas gone full circle (from dispersion back toco-location) by recently bringing all the peo-ple involved in new car development intoone building so that they can have face toface access to each other.

Engagement Dimension: How Do We Im-prove Engagement in Problem Solving? Oneof the most interesting findings from ourinterviews with the forty managers in thefirst phase of our research was the impor-tance of the person sought out for informa-tion being willing to cognitively engage with

the information seeker. People who werewilling to engage in problem solving helpedseekers to create knowledge with sufficientunderstanding and clarity that they couldtake action on it. And when we say engagingin problem solving, we do not mean a sig-nificant time investment on the part of theperson sought out. Rather, we mean a simpletwo-step behavior whereby those contactedfor information first ensured that they un-derstood the other person’s problem, andthen actively shaped what they knew to theproblem at hand. In short, these peopletaught rather than dumped information onthe seeker—a behavior that if developedamong a network can improve the effective-ness with which people learn from eachother.

Integrating Specialized Expertise inProblem Solving

We conducted one network analysis of a spe-cialist group supporting the internal knowl-edge management efforts of a global com-puter manufacturer. This group of 18 peoplewas a virtual team that had been formed tocombine expertise in both the technical andorganizational/strategic aspects of knowl-edge management. While members of thegroup claimed to know and have access toeach other’s expertise, a quick review of theengagement network showed that in fact thegroup was having little success in integratingtheir expertise. Rather, what became appar-ent was a strong split in the network becauseof unique skill bases.

Despite people technically knowing at ahigh level what the skills and knowledge ofpeople in the other discipline were, therewere only two connections between the twogroups on the engagement relation. This clus-tering was a significant concern, as it is inengagement in problem solving that truelearning takes place and people effectivelyintegrate specialized expertise in projects—rather than just doing what they know orhave done before. Interviews revealed thateach group’s depth of specialization and thefact that they were virtual and so had little

114 ORGANIZATIONAL DYNAMICS

slack, face-to-face time to interact made itdifficult for them to find common ground.Aside from the leader of the group, who hadexperience with both subgroups, there waslittle common language or occupational val-ues that existed between the two subgroups.

Several organizational learning interven-tions have been undertaken in this group tohelp build engagement and trust. As always,a key component of these interventions hasbeen the use of various network diagrams infacilitated sessions to help the group createcommon awareness and make sense of pro-ductive and unproductive dynamics. Fur-ther, a shift in performance measurementwas made to encourage joint problem solv-ing and de-emphasize individual projectmetrics. While in the midst of these initia-tives, the group plans to periodically assessthe engagement network and intervene as ap-propriate to improve their operations overtime.

Supporting Engagement ofSpecialists in New ProductDevelopment

In another scenario, we conducted a networkanalysis of 78 members of a drug develop-ment community of practice. The commu-nity, which was geographically dispersedacross eight sites in the U.S. and Europe,included people from the drug discoverystage all the way to clinical development.The analysis indicated that within this highlydispersed community there were many peo-ple who did not know each other. It alsobecame apparent that, although the peoplewithin each functional unit engaged witheach other on matters relevant to the com-munity, there was little engagement betweenthe functions. This was a critical problem forthis group, given the need to combineunique expertise to effectively develop andmarket a specific drug.

Further, the network proved to be highlycentralized around a few individuals. The sixmost central people resided in the main U.S.site. Although they had many connections topeople within the site, they did not engage as

often with community members in the Euro-pean locations. There were also several peo-ple who were totally disconnected from thegroup, which resulted in their skills and ex-pertise being lost to the community. In thisinstance, a new collaborative technology wasintroduced that had both synchronous andasynchronous features. Second, differentproject management practices and a new rolewithin the community were initiated to helpbridge functional areas of expertise. Finally,the network diagrams were used to convincemanagement to support staged face-to-faceforums focusing on specific problems. Theseforums helped the different functional areasfind common ground, while solving prob-lems critical to the success of a project.

Engagement in Human Networks

Overall, as with the access dimension, wefound that many of the things organizationswere doing that had an impact on engage-ment were technical in nature and includedsynchronous technologies such as VPBuddy, Same Time, or white boarding appli-cations that allow for dispersed engagementin a common problem. In many ways, instantmessaging does seem to support the seren-dipitous kinds of interactions that are lostwhen employees are not co-located. How-ever, there are limitations in the ability ofthese applications to richly convey knowl-edge across media that provide relativelyfew cues in comparison to face-to-face inter-actions. Videoconferencing for visual inter-action between people in different locationsdoes seem to help. This has been particularlyimportant at British Petroleum, where ex-perts have been able to assist technicianswho are working on oil rigs thousands ofmiles away.

British Petroleum is also unique in itsrecognition of the importance of engagementin problem solving early in projects wherelearning from others’ experiences can have adisproportionate impact on the trajectoryand success of a project. For example, BP hasinstituted a peer review process in its drillinginitiatives as an effective way to tap into

FALL 2001 115

others’ knowledge. Before engaging in anysignificant task, the individual or group in-vites peers to provide input. Because the fo-cus is performance, those with the most rel-evant knowledge and recent experiences aretapped to participate. Through this peer re-view process not only is performance on thetask at hand improved, but also people be-come much more aware of the unique skillsand abilities of others. This creates a naturalreason for meeting and developing theneeded norms of reciprocity and trust thatmake engagement and sharing of expertise anatural process.

Safety Dimension: How Do We PromoteSafety in Relationships? Finally, the manag-ers we interviewed in the first phase of ourresearch indicated that safe relationships of-fered certain advantages in problem solving.First, they provided more learning, as peoplewere not overly concerned about admitting alack of knowledge or expertise. Askingsomeone for help often requires that theseeker have some degree of trust in the per-son sought out for information. Such trustoften shapes the extent to which people willbe forthcoming about their lack of knowl-edge, as defensive behaviors can knowinglyand unknowingly block learning in criticalinteractions. Second, several of the managersindicated that in more safe relationships theycould be more creative. An important featureof these relationships was that they weremore willing to take risks with their ideasand felt that this often resulted in more cre-ative solutions.

Safety Promotes Learning in HighEnd Knowledge Work

Social network analysis provides us with ameans of understanding the extent to whichinformation and knowledge seeking is a safebehavior in important groups. For example,we assessed the safety network in the infor-mation resources group supporting a key re-search and development function of a For-

tune 500 manufacturing organization. Thisgroup of 34 people was composed of twoorganizational units that had recently beenmerged under one leader. The Safety net-work represented an interesting point of in-tervention here because, unlike many net-works we have seen, the Knowledge, Access,and Engagement networks were all very wellconnected, whereas the safety network wasnot.

Interestingly enough, the safety networksplit into two groups that reflected the twodepartments that had been merged severalmonths before this analysis. This is a com-mon finding in both restructuring andmerger scenarios. We often have found thatcommunication networks (i.e., network dia-grams developed from asking people whothey typically communicate with) formquickly in restructuring or merger scenarios.However, what simply assessing communi-cation patterns obscures is the time and ef-fort that must be put into developing trustamong a group, if we truly want people tolearn from each other. Safety is importantand highly predictive of who is sought outwhen one engages in problem solving and soexposes a lack of knowledge or allows some-one else to shape the course of a solution.Relationships that are safe, and thereforeuseful for deeper levels of knowledge shar-ing and true learning, take time to develop.

In this specific network analysis, therewere two interesting points. First, two peo-ple who were low in the hierarchy had be-come important ambassadors between thegroups. Several amusing anecdotes were dis-covered in our interviews, whereby peoplethat were senior in this group often went tothese more junior people when they neededinformation from a colleague in the othersubgroup. A light-hearted but very effectiveintervention was created by using these an-ecdotes along with the network diagram in afacilitated session debriefing the overallgroup. Playfully illuminating the way inwhich members of each group had stereo-typed the other, and the inefficiencies that

116 ORGANIZATIONAL DYNAMICS

this caused, resulted in a productive discus-sion of a potentially charged issue.

Second, there were different levels ofsafety between the two groups. In part thisseemed to be a product of the physicalenvironment, as the more tightly con-nected group had all worked in an openspace environment that allowed frequent,face to face communication. We also foundthat leadership style differed in the twogroups before the restructuring. In general,creating a greater degree of safety withinnetworks of relationships is often a prod-uct of leadership style and organizational(or sometimes occupational) culture. Thebehaviors that leaders exhibit and thosethey reward shape the extent to which peo-ple will be forthcoming about their lack ofknowledge on various topics. This variedwidely by organization. In some, safetywas never considered a concern, because itwas an accepted norm to doggedly seekout the most relevant knowledge for thesuccess of a given project. In others, safetywas a critical concern, and employees werevery cautious about exposing a lack ofknowledge.

Just as important, our interviews indi-cated that relationships need time and somespace (physical, cognitive and social) to de-velop a sense of safety. Although communi-cation technologies such as e-mail are helpfulin maintaining relationships, when creatingrelationships we have found that it is impor-tant to increase the opportunity for face-to-face interactions between people. For exam-ple, though often chided, organizations thathave instigated a program of brown baglunches find that this process is effective forthe development of safe relationships be-tween people. One organization we workedwith encouraged face-to-face contact bymonthly meetings between different groupsof researchers. These meetings consisted of adiscussion session in the morning and aworking session in the laboratory in theafternoon and allowed for a free flow ofideas within the context of a real workingenvironment.

A COMBINED NETWORK VIEWAND ORGANIZATIONALLEARNING

In addition to looking at each of the net-works individually, it is also instructive toassess the dimensions cumulatively to get abetter understanding of a network’s under-lying learning potential. In doing this, wecan analyze networks where pairs of rela-tionships exist (e.g., both knowledge and ac-cess) or networks where all of the relation-ships exist (e.g., knowledge, access, engagement,and safety). For example, we conducted asocial network analysis of 38 employees con-stituting the telecommunications consultingpractice of a Big Five accountancy. We firstassessed the knowledge network to better un-derstand who in this network of people in-dicated that they knew and valued other’sexpertise. Though relatively sparse, wefound that the knowledge network showed ahealthy, integrated pattern without distinctsubgroups. However, the network diagramtook on added life when we also consideredthe access network, where each person ratedhis or her colleagues on the extent to whichthey were accessible in a timeframe sufficientto help solve problems. Ultimately, bothknowledge and access relations must bepresent for information sharing in a group tobe effective. By combining the networksfrom these two questions, we had a view ofthe potential of a person to obtain informa-tion from others when faced with a newproblem or opportunity.

Several things were interesting in thisnetwork. First, we noticed a fairly markeddecline in the number of connections amongthe group in comparison to the knowledgenetwork. While many central people re-mained central, several people higher in thehierarchy shifted out to the periphery of thenetwork. As people move higher in an orga-nization, their work begins to entail moreadministrative tasks, which makes themboth less accessible and less knowledgeableabout the day-to-day work of their subordi-nates. What network analysis affords in thispicture is an opportunity to assess whether

FALL 2001 117

those in positions of formal authority aresufficiently central to the flow of knowledge,as well as to identify those people that trulyare influential knowledge brokers in thegroup.

The third question asked of the 38 con-sultants was who in the group they couldcount on to actively engage in problem solv-ing. When the engage network was added wewere assessing a network where a line wasdrawn between two people only if all threedimensions of a relationship existed (know-ing what the other knows, having access totheir thinking and being willing to engage inproblem solving). With the addition of theengagement network, we found a significantdecrease in connections, which is not trivialin terms of the network’s ability to solveproblems. As outlined in the initial inter-views, it is often those people who are will-ing to engage in problem solving who helpboth create actionable knowledge (ratherthan information overload) and ensure thatwe are solving the right problem. The finalquestion we asked of this consulting practicedetermined with whom each person felt safediscussing work-related issues. With the in-corporation of the safety network there isvery little change. This is because the safetynetwork in this group was the densest of allthe networks. Ultimately, this was a soundindicator of the culture of this group forknowledge creation, and is obviously not aplace we would look to intervene. It is alsoimportant to note that based on our experi-ences, a dense safety network is not typical.

Interventions from a CombinedNetwork View

Analyzing the combined network (i.e.,Knowledge � Access � Engagement � Safety)provides a great deal of insight into who iscritical as well as who is currently less uti-lized within a group in terms of knowledgecreation and sharing. Understanding who iscentral to a group indicates people whomight either be bottlenecks or highly valuedknowledge resources upon whom the groupis reliant. Only interviews providing an in-

depth understanding of a network can tell,but these people do pose interesting ques-tions to management. Has the group becometoo reliant on these people should they de-cide to leave? Are these people hoarding in-formation and so are bottlenecks in terms ofthe group’s knowledge creation and sharingactivities? In contrast, should these people berewarded for the somewhat invisible rolethey play in supporting a group from aknowledge perspective?

If we discover that people are central inthese networks for legitimate reasons, man-agement has an opportunity to begin ac-knowledging the work that these people dofor the group. In the words of one of thepeople central in the telecommunicationspractice, “I spend about an hour and a halfevery day responding to calls and other in-formational requests. . . [and] . . . none of thattime gets seen in my performance metrics.”Network analysis makes such interactionsthat are critical to a group visible, thus pro-viding an opportunity for management toacknowledge these people and the criticalrole they play. For example, managementmight choose to better support knowledgecreation and sharing by offering central peo-ple such things as:

• Monies for efforts that might stimu-late knowledge flow in a group via face-to-face meetings, or to purchase technologiessuch as groupware.

• Cognitive and social space to allowroom for both individual and collective cre-ativity and bonding to occur.

• Executive focus such as rewarding orpromoting network enabling people to bothacknowledge their efforts and signal the im-portance of this kind of work to otherswithin the organization.

In addition to central or core individuals,we also find it important to better under-stand why some people are peripheral inthese networks. It might be that people inthese positions do not know what wethought they knew when hired. In thesecases they are peripheral for a legitimate rea-son and so reflect development or re-staffingopportunities. Alternatively, it might be that

118 ORGANIZATIONAL DYNAMICS

these people are peripheral because they arerelatively new and the organization’s assim-ilation processes do little to help them inte-grate into a network of colleagues. The im-portant feature of this combined networkview is that we can isolate why people areperipheral. Being peripheral because one isinaccessible is a different coaching processthan if one is not considered safe.

Finally, on a more conceptual level, thecombined network view offers unique pur-chase on the elusive concept of organiza-tional learning. Some have claimed that anorganization has learned when, through itsprocessing of information, its range of poten-tial behaviors has changed. Thus, if we areinterested in promoting an organization’sability to react to new opportunities, we needto account for the ways in which people innetworks become able to leverage each oth-ers’ knowledge. Changes in the knowledge,access, engagement, and safety relationshipsunderlying a network’s future informationprocessing behavior provide one means ofboth descriptive and prescriptive traction onorganizational learning. Organizations haveoften been claimed to be path-dependent orconstrained by what they know. Such no-tions as absorptive capacity, core rigidities orarchitectural knowledge have been claimedto lead to this path dependence over time.While critically important, this work has of-ten been done at a level of abstraction thatmakes interventions questionable. In con-trast, the combined view of these networksprovides some idea as to precisely whoseknowledge is primarily responsible for whata group is likely to learn over time.

CONCLUSION

A critical resource embedded within organi-zations is the knowledge that workers bring

to work on a day-to-day basis. However,aside from human resource policies targetedto the attraction, development and retentionof identified valuable workers, there hasbeen little effort put into systematic ways ofworking with the knowledge that is embed-ded in social networks. Given the extent towhich people rely on their own knowledgeand the knowledge of their contacts to solveproblems, this is a significant shortcoming.By introducing social network analysis tounderstand how a given network of peoplecreate and share knowledge, we are able tomake these interactions visible and so action-able.

In applying these ideas in various orga-nizations, we have found it particularly im-portant to identify points of knowledge cre-ation and sharing within an organizationthat hold strategic relevance. Typical do-mains yielding benefit include senior man-agement networks, communities of practiceand collaborative initiatives such as newproduct development, R&D units or jointventures and alliances. It is particularly fruit-ful to map collaborative relationships thatcross boundaries of some form. Such bound-aries might be hierarchical, functional, geo-graphical, or even organizational, as in jointventure or merger and acquisition scenarios.Understanding how knowledge flows (ormore frequently does not flow) across thesevarious boundaries within an organizationcan yield critical insight into where manage-ment should target efforts to promote collab-oration that has a strategic payoff for theorganization.

To order reprints of this article, please call�1(212)633-3813 or e-mail [email protected]

FALL 2001 119

SELECTED BIBLIOGRAPHY

Much of the emphasis on organizationalknowledge today (at least in terms of prac-tice) is focused on efforts to capture, screen,store and codify knowledge. To get a morepopular view of what many organizationsare doing under the rubric of knowledgemanagement we suggest some of the follow-ing publications: T. Davenport & L. Prusak,Working Knowledge (Boston, MA: HarvardBusiness School Press, 1998); C. O’Dell & C. J.Grayson, If Only We Knew What We Know(New York, NY: Free Press, 1998); T. Stewart,Intellectual Capital: The New Wealth of Organi-zations (New York, NY: Doubleday, 1997);and R. Ruggles, “The State of the Notion:Knowledge Management in Practice,” Cali-fornia Management Review, 1998, 40(3), 80–89.

Of course, our own perspective is thatknowledge embedded in human networks istoo often overlooked in these initiatives. Twostreams of literature heavily influenced ourthinking here. First is the rich ethnographicevidence accumulating within the situatedlearning and community of practice tradi-tions. This work is making clear the largedegree to which people learn how to do theirwork not from impersonal sources of infor-mation but through interactions with otherpeople. Some important work in this tradi-tion includes: J. S. Brown & P. Duguid, “Or-ganizational Learning and Communities-of-Practice: Toward a Unified View of Working,Learning and Innovation,” Organization Sci-

ence, 1991, 2(1), 40–57; J. Brown & P. Duguid,The Social Life of Information (Cambridge, MA:Harvard Business School Press, 2000); J. Lave& E. Wenger, Situated Learning: Legitimate Pe-ripheral Participation (Cambridge, UK: Cam-bridge University Press, 1991); J. Orr, TalkingAbout Machines (Ithaca, NY: Cornell Univer-sity Press, 1996); and E. Wenger, Communitiesof Practice (Oxford, UK: Oxford UniversityPress, 1998).

The second stream of literature influen-tial in our thinking came from the socialnetwork tradition, which has also shown,with very different methods, the extent towhich information that affects what we dolargely comes from other people. Some im-portant works on how social networks influ-ence information flow and diffusion in net-works include: G. Simmel, The Sociology ofGeorg Simmel (New York, NY: Free Press,1950); R. Burt, Structural Holes (Cambridge,MA: Harvard University Press, 1992); M.Granovetter, “The Strength of Weak Ties,”American Journal of Sociology, 1973, 78, 1360–1380; T. Allen, Managing the Flow of Technol-ogy (Cambridge, MA: MIT Press, 1984); P.Monge & N. Contractor, “Emergence ofCommunication Networks,” forthcoming inF. Jablin and L. Putnam (Eds.), Handbook ofOrganizational Communication, 2nd ed. (Thou-sand Oaks, CA: Sage); and E. Rogers, Diffu-sion of Innovations, 4th ed. (New York, NY:Free Press, 1995).

120 ORGANIZATIONAL DYNAMICS