When we think of where people turn for information, we usually think of
databases, the Internet, intranets and portals, or more traditional
repositories such as file cabinets or manuals. What we may not think of
is one of the most crucial sources of all: other people.
Research in sociology, social psychology, communication, and management
has consistently shown that who you know has a substantial impact on
what you come to know; personal relationships are critical for obtaining
information, solving problems, and learning how to do your work. As we
move further into an economy in which collaboration is central to
organizational effectiveness, we must pay more attention to the sets of
relationships that people rely on to accomplish their work.
Emerging collaborative technologies will certainly facilitate virtual
work, and skill-profiling systems can help people locate relevant
expertise in an organization. As was so poignantly demonstrated by
reengineering efforts, however, technology alone can accomplish only so
much in the pursuit of business performance. Improving efficiency and
effectiveness in knowledge-intensive work demands more than
sophisticated technologies. It requires attending to the often
idiosyncratic ways that people learn from and solve problems with other
With this in mind, we initiated a research program designed to help
managers promote knowledge creation, sharing, and learning in
strategically important networks of employees. In the first phase of our
research, we interviewed 40 managers from an organization that most
industry analysts heralded as an exemplar of knowledge management
because of its investment in technology and the organizational roles it
had created to collect, screen, and update content in distributed
databases. We asked the managers to reflect on a recent project that was
important to their careers and indicate where they had obtained crucial
information. Eighty-five percent of them said that they had received
information critical to the success of the project from other people,
not from impersonal sources such as computer archives, the Internet, or
the organization’s knowledge- management database. And the 15 percent
who did rely on impersonal sources were all relatively new to the
organization and so had not yet developed sufficient networks.
We also asked the managers to identify the people most important to the
project in terms of information or knowledge and to describe those
relationships. Four characteristics emerged that distinguish effective
from ineffective relationships.
Knowledge: Being aware of what other people know—even if initial
perceptions turn out to be inaccurate—is a precursor to seeking a
specific person for help.
Access: Knowing what someone else knows is useful only if that person is
available in time to help. Access is heavily influenced by physical
proximity, organizational design, and collaborative technology.
Engagement: People who encourage true learning are those who think along
with the seeker and participate in problem solving. Rather than dumping
information on the seeker, these people first understand the problem as
experienced by the seeker and then shape their knowledge to that
problem. Safety: Feeling safe with a person, that is, being able to
admit a lack of knowledge or to diverge in a conversation, often results
in creativity and learning.
The managers we interviewed recounted numerous times when learning or
knowledge sharing did not happen because one of the above
characteristics did not exist in a relationship. For instance, a
colleague had knowledge that would have been useful to other people, but
he or she was not accessible. A separate quantitative study that we
performed demonstrated that these characteristics consistently determine
whom people seek out for information, even after controlling for
similarities in age and education, physical proximity, time in the
organization, and formal hierarchical position.
Social Network Analysis
With the importance of these four characteristics established, we moved
on to the second phase of our research, in which we used social network
analysis to map the flow of information in organizations as well as the
relational characteristics among strategically important groups. Social
network analysis provides a rich and systematic means of assessing
informal networks by mapping and analyzing relationships among people,
teams, and departments, or even throughout an organization.
Though managers often insist that they know their organizations, studies
show that in fact they understand the networks around them with varying
levels of accuracy. By virtue of their position in the hierarchy,
managers are frequently far removed from the day-to-day interactions
that generate an organization’s informal structure, thus many of them
have an inaccurate picture of the actual patterns of relationships. And
the potential for misperceptions is increased by the transition to a
world of virtual work and telecommuting, in which employees’
relationships are increasingly invisible to superiors. Social network
analysis can provide an x-ray of the way in which communication is or is
not occurring in these informal networks.
Consider the social network analysis we conducted of the top 20
executives in the exploration and production division of a large
petroleum company. This group was in the midst of implementing a
distributed technology to help transfer knowledge across drilling
initiatives and was also interested in assessing its ability to create
and share knowledge.
Our analysis revealed a striking contrast between the division’s formal
and informal structures. (See the exhibit.) It turned out that certain
midlevel managers were critical to the flow of information within the
group. A particular surprise came from Cole’s central role: over time,
his reputation for expertise and responsiveness had made him a critical
source for all sorts of information. Indeed, he was the only point of
contact between members of the production division and the rest of the
network. Through no fault of his own, the number of informational
requests he received and the number of projects in which he was involved
had become excessive, which not only overburdened him but also turned
him into a bottleneck.
The social network diagram made it very clear that if Cole were hired
away, the company would lose both his knowledge and the relationships he
had established, which were holding the network together. As a result of
the analysis, the organization decided to categorize the requests that
Cole received and then allocate some of these informational or decision
domains to other executives. This simple solution unburdened Cole and
made the overall network more responsive and robust.
Just as important, the social network analysis helped identify
peripheral people who represented untapped expertise. In particular, it
became apparent that the most senior person, Jones, was one of the most
peripheral in the informal network. This is a common finding. As people
move higher within an organization, their work begins to entail more
administrative tasks, which makes them both less accessible and less
knowledgeable about the day-to-day work of their subordinates. In this
extreme case, the social network diagram helped turn what could have
been a difficult confrontation with this executive into a constructive
discussion, which led him to commit more of his time to the group.
Finally, the social network analysis also demonstrated the extent to
which the production division (the subgroup at the top of the diagram)
had become separated from the overall network. Several months prior to
this analysis, this division had been moved to a different floor. After
reviewing the network diagram, many executives realized that this
physical separation had resulted in fewer serendipitous meetings. The
executives decided that they needed to introduce more structured
meetings, and they adopted an instant messaging system to promote
Beyond Mapping Information Flow
In addition to using social network analysis to map the flow of
information in an organization, we have also used it to assess networks
of knowledge, access, engagement, and safety among various groups. If we
map only an information network and find that certain people are not as
connected as they should be, it can be difficult to know what to do.
Simply proposing more or better communication is the oldest consulting
recommendation in the book—and no one today really needs more meetings.
By analyzing the dimensions of relationships that lead to effective
knowledge sharing, we can offer more precise ways to improve a network.
Let’s consider some case examples of how we have used social network
analysis to assess the four characteristics.
Do we know what we know?
Assessing the knowledge of who knows what at a network level reveals how
well members of a network can reach others with relevant expertise. For
example, we analyzed a network of immunologists in a Fortune 250
pharmaceutical company. This group had the potential to provide
strategic advantage to the organization, but there were many impediments
to collaboration. One telling view emerged when we mapped the knowledge
network. The network we uncovered was very sparse, indicating that group
members were not aware of one another’s knowledge. Two characteristics
of this group seemed to result in the sparse pattern. First, the group
was physically dispersed, which precluded the serendipitous interactions
that help people learn about their colleagues’ expertise. Second, the
group housed two subgroups of deep specialists. One set of specialists
often did not know enough about the other to be able to involve it in
The Formal Organizational Structure of a Petroleum Company's
Exploration and Production Division
The Division's Informal Network
As a result of lengthy discussions about what the social network
analysis had revealed, the organization made some changes. First, a
variety of internal projects—ranging from process improvements to a
project-tracking database—were staffed with people from each subgroup,
which forced them to work together and so to develop an appreciation of
one another’s skills and knowledge. Second, several communication forums
were created, including weekly status calls and a short weekly e-mail.
Finally, simple changes in project management practices and a
restructuring of the project leaders’ responsibilities helped people
connect more effectively.
In more staid times, working relationships developed as a product of
interaction over longer time periods. Given the rapid turnover in many
companies today, people must find ways to become better connected. One
way is to focus on how new people are brought into a group. Most
organizations offer courses that teach new hires about the computer
system, benefits, and, perhaps, the company’s culture and history. It is
rare to find practices that teach employees what the newcomers know.
Many organizations we worked with are employing various mechanisms to
build an awareness of who knows what. For example, some are implementing
skill-profiling systems or corporate yellow pages. Others, such as the
World Bank, have arranged employees in thematic groups that have help
desks that anyone in the organization can contact to locate relevant
expertise. Other companies and government organizations hold knowledge
fairs at which teams or departments distribute information about what
Can we reach the right people?
Knowing that someone else knows something relevant does little good if
we cannot gain access to that knowledge in a timely fashion. And
consider how difficult such access can become during a reorganization.
We worked with a commercial lending organization in transition from a
functional to a cross-functional team-based structure. Prior to the
transition, each function had been housed on its own floor, so
functional employees were able to tap one another’s knowledge with
relative ease. With the redesign, it was far more difficult for
inexperienced people to learn the basics of their function and for
experienced lenders and analysts to engage in collaborative
Social network analysis showed that four months after the transition to
teams, several people were heavily sought out by their functional
colleagues as well as their new team members. In particular, we found
that the people who were reputed experts in their area were asked for
advice to such an extent that they were falling behind on their own
work. In the functional department, these interactions were more
controlled and observable, but in the team-based environment it was
difficult for management to see the importance of these opinion leaders.
As a result of these findings, the organization developed new staffing
practices and better orientation materials to help bring new people up
to speed more effectively, and it re-allocated tasks within teams
Through the course of our research, we saw many companies try to manage
accessibility with technical solutions: e-mail, collaborative
environments, video conferencing, and instant messaging. But the
obstacles that keep one person from learning from another may be deeply
rooted in an organization. Performance management systems promoting
individualistic behaviors seem to be one of the primary drivers of
sparse, disconnected networks. And we often find that hierarchy has a
marked impact on who is able to access whom. Again, this is a telling
indicator for organizations trying to become more flexible and effective
at information sharing.
Some organizations have taken interesting steps to promote access across
hierarchy, such as making knowledge sharing a part of the mission or
code of ethics. At Buckman Laboratories, all associates are empowered to
speak with any associate at any level, and this is supported by a
communication technology that gives each employee access to all other
employees. Other companies are beginning to use physical space to
promote both intentional and serendipitous interactions among high-end
knowledge workers. For example, Chrysler has gone full circle (from
dispersion back to co-location) by bringing all the people involved in
new car development into one building so that they can have face-to-face
access to one another.
Are knowledgeable people willing to engage?
One of the most interesting findings from our interviews in the first
phase of our research was the importance of true engagement in problem
solving. People who effectively share their knowledge tend first to
ensure that they understand the problem (as experienced by the person
seeking information from them) and then to shape what they know to that
We conducted a network analysis of a specialist group supporting the
internal knowledge-management efforts of a global computer manufacturer.
This group of 18 people was a virtual team that had been formed to
combine expertise in both the technical and organizational and strategic
aspects of knowledge management. Although members of the group claimed
to know and have access to one another’s expertise, a quick review of
the engagement network showed that in fact the group was having little
success integrating their expertise. What became apparent was a strong
split in the network caused by divergent skill bases.
Despite people knowing at a high level the skills and knowledge of
people in the other discipline, there were only two connections between
the groups on the engagement network. Interviews revealed that the
members of this virtual team had few, if any, face-to-face interactions.
Aside from the leader of the group, who had experience with both
subgroups, there was little language and few occupational values that
the two sides shared.
Several organizational learning interventions have been undertaken to
help this group increase engagement and build trust. As always, a key
component of these interventions has been the use of various network
diagrams to help the group make sense of productive and unproductive
dynamics. Further, the organization has made a shift in performance
measurement to encourage joint problem solving and de-emphasize
individual project metrics. While in the midst of these initiatives, the
group plans to assess periodically the engagement network and intervene
as appropriate to improve its operations over time.
Many organizations have implemented technologies in an effort to
increase the level of engagement among employees, but such applications
cannot fully substitute for face-to-face interactions. Videoconferencing
between people in different locations does seem to help. This has been
particularly important at British Petroleum, where experts have been
able to assist technicians working on oil rigs thousands of miles away.
British Petroleum has recognized the importance of engagement early on
in a project, when learning from others’ experiences can have a
disproportionate impact on the project’s success. For example, the
company has instituted a peer review process in its drilling
initiatives. Before undertaking any significant task, the individual or
group invites peers to provide input. Because the focus is performance,
those with the most relevant knowledge and recent experiences are tapped
to participate. Through this peer review process, performance on the
task at hand improves, and people become much more aware of the skills
and abilities of others.
Do we feel safe with one another?
The managers we interviewed indicated that safe relationships offer
advantages in problem solving. When one person trusts another, he or she
can more easily admit a lack of knowledge and thus end up learning more.
Safe relationships also let people take bigger risks with their ideas to
come up with creative solutions. We assessed the safety network in the
information resources group supporting a key R&D function of a Fortune
500 manufacturing organization. This group of 34 people comprised two
units that had recently been merged under one leader. The knowledge,
access, and engagement networks were all very well connected, but the
safety network split into two groups that reflected the two departments
that had existed before the merger. The unsurprising lesson:
relationships that are safe and therefore useful for deeper levels of
knowledge sharing and learning take time to develop.
In this specific network analysis, there were two interesting points.
First, two people who were low in the hierarchy had become ambassadors
between the groups. In fact, senior members often went to these more
junior people when they needed information from a colleague in the other
subgroup. A light-hearted but very effective intervention was created by
using anecdotes along with the network diagram in a session with the
overall group. Playfully illuminating the way in which members of each
group had stereotyped the other resulted in a productive discussion of a
potentially charged issue.
Second, there were different levels of safety between the two groups. In
part this seemed to be a product of the physical environment: the more
tightly connected group had all worked in an open environment that
allowed frequent face-to-face communication. We also found that
leadership styles differed between the two groups prior to the
restructuring. The behaviors that leaders exhibit and those they reward
shape the extent to which people will be forthcoming about their lack of
knowledge on various topics. In some organizations we studied, employees
were expected to seek out the information most relevant for the success
of a given project. In others, people were very cautious about revealing
any lack of knowledge.
• • • • •
In addition to mapping each characteristic individually, we have
analyzed networks in which pairs of characteristics exist (e.g., both
knowledge and access) and networks in which all of them exist (e.g.,
knowledge, access, engagement, and safety). Such analysis yields a full
picture of who is central to a group and who is peripheral, and it helps
bring into high relief the elusive concept of organizational learning.
By introducing social network analysis to understand how a given network
of people creates and shares knowledge, we are able to make these
interactions visible and thus actionable.
We have found it particularly important to identify points of knowledge
creation and sharing within an organization that hold strategic
relevance: senior management networks, communities of practice, and
collaborative initiatives such as new product development, R&D units, or
joint ventures and alliances. It can be particularly fruitful to map
collaborative relationships that cross boundaries, whether hierarchical,
functional, geographical, or even organizational. Understanding how
knowledge flows (or does not flow) across various boundaries can yield
critical insight into where management should target efforts to promote
collaboration that will have a strategic payoff for the organization.
Rob Cross is an assistant professor at the University of
Virginia’s McIntire School of Commerce and a research fellow at the
IBM Institute for Knowledge Management in Cambridge, Massachusetts.
He has an MBA from the University of Virginia and a doctorate in
organizational behavior and information technology from Boston
University. Contact him at email@example.com.
Andrew Parker is a research associate at the IBM Institute
for Knowledge Management and can be reached at
Laurence Prusak is the executive director of the IBM
Institute for Knowledge Management and can be reached at
Stephen P. Borgatti is an associate professor of management
at Boston College. Contact him at
A longer version of this article appeared in Organizational Dynamics
30, no. 2 (2001): 100 - 120.
(c) 2002 The Darden School Foundation.
This article was originally published in
Transforming Culture: An Executive Briefing on the Power of Learning.
June 2002. It is reprinted here with permission.