Embed Size (px)
Citation preview
Copyright © 2010 by Bradley R. Staats, Melissa A. Valentine, and Amy C. Edmondson
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It may not be reproduced without permission of the copyright holder. Copies of working papers are available from the author.
Using What We Know: Turning Organizational Knowledge into Team Performance Bradley R. Staats Melissa A. Valentine Amy C. Edmondson
Working Paper
11-031
Using What We Know: Turning Organizational
Knowledge into Team Performance
Bradley R. Staats University of North Carolina at Chapel Hill
Chapel Hill, NC 27599-3490
Tel: 919.962.7343
Melissa A. Valentine Harvard Business School
Boston, MA 02163
Tel: 617.852.8644
Amy C. Edmondson Harvard Business School
Boston, MA 02163
Tel: 617.495.6732
December 31, 2010
Acknowledgments
We thank Sambuddha Deb, Venkatesh Hulikal, Sowmya Narayanan, Ved Prakash, Alexis Samuel, and
many other individuals at Wipro for their significant commitment of time and effort, which made this
project possible. Martine Haas, Sirkka Jarvenpaa, Jackson Nickerson, Lamar Pierce, and Anita Tucker
provided valuable comments on earlier drafts of this paper. This work has been improved significantly by
the comments and suggestions of Jesper Sorenson, the Associate Editor and the three reviewers. This
material is based on work supported by the National Science Foundation under Grant No. 0943210. Any
errors are our own.
Using What We Know
- 1 -
Abstract
This paper examines how teams draw on knowledge resources in the firm in the production of novel
output. We theorize positive effects of team use of an organizational knowledge repository on two
measures of team performance (quality and efficiency), and argue that these effects will be greater when
teams face structural characteristics (team geographic dispersion and task change) that intensify the
challenge of knowledge integration. Drawing on information processing theory, we distinguish between a
team’s knowledge repository use and concentration of use (the extent to which use is limited to a few
members versus more evenly distributed within the team). Using objective data from several hundred
software development projects in an Indian software services firm, we find that knowledge repository use
has a positive effect on project efficiency but not on project quality. Concentration of repository use, a
form of within-team specialization, is negatively associated with project efficiency and positively related
to project quality. Finally, as predicted, we find that in some cases the effects of both repository use and
concentration of repository use are greater when teams are dispersed geographically or encounter
changing tasks. Our findings offer insight for theory and practice into how organizational knowledge
resources can improve knowledge workers’ productivity and help build organizational capability.
Key Words: Knowledge Management, Knowledge Work, Learning, Software, Virtual Teams
1. Introduction
An organization’s success in a competitive market often depends on its ability to create and apply
new and valuable knowledge (Wernerfelt 1984; Barney 1986; Teece, Pisano and Shuen 1997). In settings
ranging from consulting to product development, telecom, and software services, organizations use teams
to create novel output (Haas 2006a; Edmondson and Nembhard 2009; Huckman, Staats and Upton 2009),
a process that typically involves combining existing pieces of knowledge in new ways (Henderson and
Clark 1990; Nonaka and Takeuchi 1995; Fleming and Sorenson 2004). Therefore, the ability of teams to
draw on stores of relevant knowledge that exists within the boundaries of their organization presents an
important area for continued research (Cohen and Levinthal 1990; Kogut and Zander 1992).
Previous research showed that teams in the same organization confront varying conditions that
may limit their ability to leverage existing organizational knowledge in ways that enhance performance
(Hansen 1999; Cummings 2004; Haas 2006b). One way managers attempt to overcome the challenges
teams face in accessing existing organizational knowledge is by creating a computer-based repository to
store organizational knowledge as a resource that can be used by others. A knowledge repository might
include case studies from prior projects, white papers on related topics, or even reusable components
(Davenport and Prusak 1998; Alavi and Leidner 2001; Hansen and Haas 2001). Thus, team members can
Using What We Know
- 2 -
use a knowledge repository to identify relevant knowledge, acquire it, and then apply it to a task. This
paper draws on the team, knowledge management and information systems literatures to study how team
use of an organizational knowledge repository affects team performance (quality and efficiency), and how
these relationships are affected by structural characteristics of the team and the context.
First, building on prior work examining knowledge use in organizations (e.g., Haas and Hansen
2005; 2007), we examine whether differential use of a knowledge repository explains performance
variation in teams. Though all teams have access to the same knowledge repository, teams may vary in
how much they use it (Gibson and Vermeulen 2003). We examine this question using unique archival
data that measures per-click use of the repository by individuals on teams.
Second, drawing on information processing theory (Galbraith 1974; Tushman and Nadler 1978),
we explore how the structure of repository use within teams affects performance. Because teams are open
systems confronted with many types of uncertainty (Katz and Kahn 1966; Thompson 1967), how team
members structure repository use is likely to influence team performance (Tushman and Nadler 1978;
Tushman and Katz 1980). In our context, repository use can be concentrated – left to one or a few
members – or distributed more evenly across members. More concentrated use, holding the volume of use
constant within a team, may result in less redundancy in knowledge accessed from the repository and may
yield more efficient searches, as frequent users learn the technical language of the repository (Arrow
1974). Alternatively, more distributed use may increase the quality of information processing within a
team because team members’ direct interaction with the repository could increase the salience of found
knowledge (Gino 2008) and shape the search path in productive ways (Fleming 2001). Below, therefore,
we theorize how concentration of use may differentially affect team efficiency and quality performance.
In addition to examining the amount and concentration of knowledge repository use in teams, we
examine two properties of teams that pose significant challenges to identifying, accessing, and applying
stored knowledge. First, some teams are comprised of workers who are geographically dispersed (Gibson
and Gibbs 2006; O'Leary and Cummings 2007), which creates obstacles to recognizing and
communicating about the knowledge that exists within the team or within the team members’ networks
Using What We Know
- 3 -
(Cramton 2001; Haas 2006a; Cummings, Espinosa and Pickering 2009). Second, due to factors ranging
from customer co-production in services (Frei 2006) to technological uncertainty in knowledge work
(Staats, Brunner and Upton 2010), tasks are often fluid or changing. We thus explore how both
geographic dispersion of team members and level of task change moderate the relationships between the
amount and distribution of knowledge repository use and team performance.
To test our hypotheses we obtained data on software teams, team members, and team member
repository use by bringing together multiple, internal databases from Wipro Technologies, a global
provider of software services. These databases include not only objective performance outcomes and
project characteristics for more than three hundred software development projects completed during 2008
and 2009, but also comprehensive human-capital data on the more than nine-thousand individuals
constituting the project teams. Individual use of the knowledge repository was tracked on a per-click
basis, allowing us to examine the relationship between repository use and team performance at a fine-
grained level. Because the value of repository use may vary with the type of performance sought (Haas
and Hansen 2007), we theorize and empirically test the effect of team repository use on both project
efficiency and project quality. Figure 1, below, summarizes our research model.
************************** INSERT FIGURE 1 ABOUT HERE **************************
This study makes several contributions to theory. First, we show that how a team organizes its use
of a knowledge repository affects performance outcomes. We demonstrate that there are consequences to
a team concentrating all its repository use within certain team members versus more evenly distributing
such use. This finding underscores the importance of team structure in learning and performance.
Second, we establish two contingent conditions in the linkage between knowledge repository use
and performance. We develop and test theory on moderators that are both practically and theoretically
compelling because they reflect the changing nature of work. We thereby contribute to the literature
addressing challenges that virtual teams face, by showing that organizational knowledge repositories can
help overcome the challenges of dispersion. We find that the performance effects from team repository
use are magnified when the team is more geographically dispersed. Additionally we show that repository
Using What We Know
- 4 -
use may mitigate the negative performance effects of task change. Together these findings suggest that
treating teams as uniform entities for theory development and analysis, rather than including unique
characteristics of a team’s task, work process and context (e.g., Bunderson and Sutcliffe 2002; Bunderson
2003; Zellmer-Bruhn and Gibson 2006), may miss important determinants of performance.
Finally, this paper also implements, to our knowledge, the first wide-scale evaluation of objective
performance value of an organizational knowledge repository using archival data. In 2007, U.S.
companies spent approximately $73 billion on knowledge management initiatives (Murphy and Hackbush
2007). As lamented by a senior manager at our field site, “We believe that our knowledge management
initiative has value, but we have no empirical evidence to support that view.” By building on concepts
validated in previous research, and by precisely measuring actual, as opposed to perceived, use of the
system, we are able to make generalizable comparisons across project teams. In so doing we show a
demonstrable difference (for some performance measures) due to how much a team uses the knowledge
repository. Altogether our work speaks to the question of how to organize to create knowledge: providing
teams access to knowledge generated throughout the organization and then matching the use of the
knowledge resources to the team’s problem enhances the probability of successful knowledge creation
and performance.
2. Team Knowledge Repository Use
Knowledge influences performance in many ways, including through the development and
execution of core organizational processes, and through the ability to change and adapt those processes
when requirements shift (Kogut and Zander 1992; Nonaka and Takeuchi 1995; Teece et al. 1997).
Knowledge-driven performance relies on individuals’ effective decisions and behaviors to create, retain,
transfer, and apply knowledge, so knowledge as a resource can easily be underleveraged, misapplied, or
wasted (Davenport and Prusak 1998; Argote 1999; Argote, McEvily and Reagans 2003; Singh, Hansen
and Podolny 2010). In the next section we discuss how team use of a knowledge repository affects team
performance, on average, and in particular when the team is geographically dispersed or faces changing
Using What We Know
- 5 -
tasks. We then investigate how the team’s organization of repository use (distributed across the team or
concentrated within a few members) affects performance. We also consider how geographic location and
task change moderate the relationship between the concentration of team use and performance.
2.1 Effect of Knowledge Repository Use on Team Performance
Before knowledge can be stored in a knowledge repository it must first be codified. Codified
knowledge refers to knowledge that is transmittable in formal, symbolic language, and may be of several
types: declarative – knowledge about something; procedural – knowledge about how something occurs
or is performed; or causal – knowledge about why something happens (Garud 1997). Much prior research
on codified knowledge has emphasized predictors of its creation and use in organizations (Alavi and
Leidner 2001). For example, the creation of knowledge for storage in a repository may be hampered by
misalignment of individual with organizational incentives (Cabrera and Cabrera 2002; Bock, Sabherwal
and Qian 2008), and the process of codifying knowledge may be prohibitively difficult (Szulanski 1996).
However, researchers have argued that organizational norms, culture, or incentives can mitigate these
challenges (Davenport and Prusak 1998; Bock et al. 2005). Impediments to the use of knowledge
repositories also have been identified. One of the most important drivers of repository use is the perceived
quality of the knowledge stored there (Kankanhalli, Tan and Wei 2005; Bock et al. 2008). Additionally, if
knowledge repositories are overcrowded or difficult to search, then use decreases (Hansen and Haas
2001; Garud and Kumaraswamy 2005).
A second stream of research on knowledge repositories examines how knowledge influences
team performance. Prior work finds that the existence of a knowledge management system can lead to
increased team learning though existence does not guarantee use (Gibson and Vermeulen 2003; Zellmer-
Bruhn and Gibson 2006). There are several reasons why use of a knowledge repository may prove
valuable for team performance. First, if managed effectively, repository use can reduce search costs for
identifying relevant information (Hansen and Haas 2001). Thus, a knowledge repository can facilitate
knowledge transfer from one part of an organization to another (Argote and Ingram 2000). By
implementing best practices developed by other teams and stored in the repository, teams can avoid
Using What We Know
- 6 -
inefficiencies that would arise from recreating knowledge (March and Simon 1993; Haas and Hansen
2007). Moreover, many knowledge management initiatives at least partially screen contents for accuracy
and value, meaning the knowledge available on the repository can improve work quality and save time.
In many cases, knowledge from the repository may not apply exactly to the task at hand, but
instead may serve as an input for knowledge creation or learning (Henderson and Clark 1990; Nonaka and
Takeuchi 1995; Fleming and Sorenson 2004). Reflection or discussion of the contents of the knowledge
repository may lead to transformation and recombination of knowledge, resulting in the creation of
knowledge that addresses a team’s need (Edmondson 2002; Brannen 2004; Bresman 2010). Using a
knowledge repository can also help a team to build shared understanding as team members learn the
technical code or language of the organization (Arrow 1974; Monteverde 1995). Thus, we hypothesize:
HYPOTHESIS 1: Team knowledge repository use is positively associated with efficiency
and quality performance.
2.2 Geographic Dispersion, Task Change, and Knowledge Repository Use
Previous research has suggested that the relationship between knowledge behaviors and
performance is conditioned on a variety of factors (Haas and Hansen 2005; Zellmer-Bruhn and Gibson
2006). Like the software projects we studied, much knowledge work is carried out in settings
characterized by team- and task-level factors that present coordination and performance challenges
(Arrow and McGrath 1995; Cohen and Bailey 1997; Edmondson and Nembhard 2009). For example,
Haas and Hansen (2005) found that the relationship between knowledge use and performance was
contingent on team experience and task competitiveness. Here we consider a property of the team and a
property of the task that pose significant challenges to identifying, accessing, and applying organizational
knowledge: geographic dispersion and changing tasks. We chose these moderating variables because
organizations that implement teams engaged in knowledge work are increasingly relying on
geographically dispersed teams encountering changing tasks - factors that could affect the value of
knowledge repository use, but which have received little attention in prior research on this topic.
Firms increasingly create teams with members who are not co-located (Hinds and Kiesler 2002;
Using What We Know
- 7 -
Gibson and Cohen 2003). Dispersed teams provide many advantages, allowing access to distributed
knowledge and skills as well as more profitable cost structures (Jarvenpaa and Leidner 1999; Gibson and
Gibbs 2006; O'Leary and Cummings 2007). However, geographic dispersion of team members affects the
way teams work, for example, relying on electronic communication and less face-to-face interaction. This
can decrease the volume and quality of communication, and increase conflict (Hinds and Bailey 2003;
Hinds and Mortensen 2005) and coordination complexity (Cummings et al. 2009). Dispersed teams
struggle to create an environment of mutual knowledge or a collective sense of what is known within the
team (Cramton 2001), so that coordinated action to identify and apply knowledge within the team is
hampered.
For teams facing communication and coordination challenges due to dispersion, access to a
knowledge repository might prove more valuable than for face-to-face teams. For example, when
accessing documents, team members learn the technical language used within the firm that might aid
problem solving (Arrow 1974; Monteverde 1995). The efficiency losses from searching the repository
and reworking the stored knowledge for the specific context might be less costly than searching within a
virtual relationship for help on how to execute a certain task. Similarly, as access to knowledgeable team
members becomes increasingly difficult, adapting best practices from other teams for new projects may
also become increasingly valuable for producing a quality deliverable. Finally, if team members are
accessing similar documents in the repository, this process may help the team build shared context and
mutual knowledge. Therefore, we hypothesize:
HYPOTHESIS 2: Knowledge repository use has a greater effect on both efficiency and quality
for teams that are more geographically dispersed.
As global markets grow more competitive the tasks that teams undertake grow increasingly fluid
(i.e., change in-process). Whether due to customer co-production in services (Frei 2006) or technological
uncertainty in knowledge work (Staats et al. 2010), in-process changes are likely to harm team
performance (Fisher and Ittner 1999). However, certain team properties have been shown to help teams
cope with in-process changes: for example, teams with greater intrapersonal team diversity (Bunderson
Using What We Know
- 8 -
and Sutcliffe 2002) are more able to respond to changing tasks (Huckman and Staats 2010).
We propose that knowledge repository use will be more valuable for teams facing changing tasks.
When the task faced by a team changes, the team must adjust its work plan, meet the new requirements,
and integrate previous work with the new task (Wallace, Keil and Rai 2004; Rai, Maruping and
Venkatesh 2009). Each of these processes requires additional time and effort, and may introduce
problems or sub-tasks that the team does not have the knowledge to solve. By using the repository, the
team may reduce the time expended at each step by applying the knowledge captured in the repository to
plan, execute, and integrate the new work. The repository may provide examples to help the team learn
how to address problems and sub-tasks that they would not otherwise know how to solve. Given that a
team facing dynamic tasks must coordinate their response (Weick and Roberts 1993), the common
language and shared representations gained through repository use may be particularly valuable.
Therefore, we hypothesize:
HYPOTHESIS 3: Knowledge repository use has a greater effect on efficiency and quality for
teams that encounter more task change.
2.3 Effect of Concentration of Knowledge Repository Use on Team Performance
Recent work highlights the role that within-group structure and individual interactions have on
team performance (e.g., Guzzo and Shea 1992; Sparrowe et al. 2001; Bunderson 2003). This has
important unexamined implications for how knowledge repository use within teams relates to team
performance. What we term knowledge repository use is a measure of how much a team is using the
repository relative to the team size. However, within a team, use of the knowledge repository may be
concentrated within just a few members or distributed among team members. We posit that for the same
level of team use, a concentrated or distributed pattern of use may affect performance differently.
Why should how team members vary in accessing a knowledge repository affect team
performance? To conceptualize this variable, we draw on prior work viewing organizations as
Using What We Know
- 9 -
information processing systems (Galbraith 1974; Tushman and Nadler 1978).1 According to this
perspective, organizations face many types of uncertainty (Katz and Kahn 1966; Thompson 1967) and
those that structure their information processing resources to reflect crucial characteristics of the task and
environment will exhibit better performance (Tushman and Nadler 1978; Tushman and Katz 1980).
Teams can similarly be thought of as information processing systems, in that they face uncertainty and
must effectively structure their information processing resources to reflect the demands of their task and
environment (Hinsz, Tindale and Vollrath 1997; De Dreu, Nijstad and van Knippenberg 2008).
By applying this theoretical lens to team repository use patterns, we can identify two primary
reasons why a concentrated or distributed pattern of use within a team may differentially affect
performance. First, as suggested by empirical work on information processing systems, communication
challenges may arise that inhibit transferring knowledge within and across intra-organizational boundaries
(Tushman and Katz 1980). Organizations develop coding schemes for technical material to facilitate
information transfer (Arrow 1974; Monteverde 1995); however, these schemes often vary across units
requiring expertise to be conversant in the varying technical languages within the organization.
Therefore, a certain level of expertise is required to be conversant in the technical language of the
knowledge repository, which stores information from a variety of units and contexts. Knowledge search
is an uncertain process and the objective of the individual is to rapidly and accurately identify a high-
value solution (Simon 1962; Nickerson and Zenger 2004). With repeated experience an individual learns
how to navigate the system (e.g., by structuring queries effectively) in a manner that increases both the
efficiency and the quality of the search process (Hansen and Haas 2001). Thus, more concentrated use,
holding the total volume of use constant within a team, may result in more effective search and therefore
may improve team performance. In addition, a concentrated pattern of knowledge repository use may
yield less redundancy in knowledge accessed from the repository. In other words, in two searches, one
with an expert accessing two knowledge artifacts from the repository and the other with two individuals
1 Early work in this tradition considered control rights of different structures at the firm level (e.g., Burns and
Stalker 1961). We draw on later studies that concentrated on the access and flow of knowledge within units
(Galbraith 1974; Tushman and Nadler 1978).
Using What We Know
- 10 -
accessing one artifact each, it is unlikely that the artifacts from the former search will be the same.
Although more concentrated use may be more efficient for a team because fewer individuals’ time
is consumed, it may lower the quality of information processing. In other words, with fewer of the
team’s members (the information processors) actively searching for relevant knowledge the potential
benefits of learning and problem solving gained from direct interaction with the knowledge will be
correspondingly limited. The learning and problem solving benefits are expected because direct
interaction with useful knowledge increases an individual’s understanding of the salience of that
knowledge, as compared with receiving a summary or translation from a colleague (Gino 2008).
Similarly, because problem solving is an uncertain process, individuals often do not know which
knowledge resources will lead to a high-value solution (Fleming 2001). Conducting a search individually
may change the path that the search takes. Individuals who directly access the repository may encounter
previously unconsidered possibilities, which they can combine with their existing knowledge, possibly
playing an important role in an eventual solution. Another consequence to concentrated use of the
repository is that knowledge held by only one or just a few team members may be left out of a team
discussion, even when it is directly relevant to the team’s task (e.g., only common knowledge may be
shared, Stasser and Titus 1987).
To study differences in team repository use patterns, we introduce the construct of concentration
of knowledge repository use. The construct is grounded in the literature discussed above, but to
understand how it operates in a particular kind of team knowledge work, we conducted interviews at our
field site, Wipro Technologies, a leading global provider of software services. During three separate visits
to Wipro’s operations in India we interviewed over thirty individuals involved in the knowledge
management initiative, senior executives at Wipro, and project team leaders and members. The interviews
offered qualitative support for the idea that varied use patterns would have a differential relationship with
performance and that teams did in fact use the repository in different ways. Although Wipro encouraged
teams to use the repository, the company did not have a specific policy on how the teams should use it.
Our interviews revealed two distinct patterns of use. The first, which we term a divide and
Using What We Know
- 11 -
conquer approach, consisted of only some members using the knowledge repository. Describing this
approach, a project manager noted, “I have been on projects where two or three people will become
formal experts on KM [knowledge management]. People know that these experts really know the system.
This can help improve project performance since they [team members] can get the right knowledge very
quickly. …So if someone has a question, they may go to the expert to ask about structuring a query or
more commonly, they may just ask the expert to get the information for them.” In follow-up questioning
the manager clarified that individuals were informal experts, as their roles were not formally assigned.
The second use pattern, what we term, share and share alike consisted of many team members
using the knowledge repository. Describing the benefits of this approach, a manager noted, “Finding a
solution through KM can quicken the pace. However, sometimes in searching for the solution, you can
examine the process the other person used to solve the problem. If you see the process yourself, then you
not only understand the solution better, but you could also improvise or find a better way.” A team
member added, “Unless you go and look yourself you may not find the answer [to a specific question]. I
can look in one area and find nothing, but I may then get an idea or find something completely different
that solves my problem.” By way of analogy, consider conducting a literature review for an academic
paper. Although an assistant might efficiently conduct the review, the quality of insight is likely to be
enhanced by doing it oneself, allowing new ideas to be identified and disparate threads to come together
in the process. This might hinder efficiency, but could improve the quality of the problem solving.
Together prior theory and qualitative evidence suggest that concentration of knowledge
repository use within a team may have differential effects on project efficiency and quality. Namely,
when knowledge repository use is concentrated, project efficiency may be increased. Given two teams
with similar numbers of downloads, the team with more concentrated use may have faster searches that
result in improved project efficiency, as compared to a team in which most members are searching,
downloading, and holding information that needs to be integrated. As a result, we hypothesize:
HYPOTHESIS 4a: Concentration of team knowledge repository use is positively associated
with efficiency performance.
Using What We Know
- 12 -
The opposite may be true with respect to project quality. For two teams with the same level of
use, a more evenly distributed use-pattern might mean that individuals are accessing knowledge that helps
their individual work for the project. Although increasing team-level search costs, each individual’s work
may be slightly improved by greater exposure to external knowledge, such as examples of prior work,
generally promoting quality performance. Work quality in software development, often defined as lack of
defects, requires an integrated, holistic view of the unique output and the ability to flawlessly combine
individual components to create a code base that is new and distinct (Faraj and Sproull 2000). Mistakes
in any one area can severely compromise an entire project. Thus, we hypothesize:
HYPOTHESIS 4b: Concentration of team knowledge repository use is negatively associated
with quality performance.
2.4 Geographic Dispersion, Task Change, and the Concentration of Knowledge Repository Use
As mentioned earlier, while geographic dispersion offers benefits to teams, such as the ability to
access distributed knowledge or situated expertise (Sole and Edmondson 2002), it also increases task-
related uncertainty due to decreased communication, increased process difficulties, and higher
coordination complexity (Hinds and Bailey 2003; Cummings et al. 2009). Similarly, task change initiates
alterations to the established work plan, regardless of the cause, and thus introduces meaningful task-
related uncertainty. Greater task uncertainty increases the need for knowledge and for information
processing in a team (Tushman and Nadler 1978). With a more even distribution of knowledge repository
use, more individuals can use their knowledge to engage in cycles of action and reflection, promoting
team learning (Edmondson 2002). Additionally, the increasing demands for knowledge associated with
task uncertainty, give rise to a risk of capacity constraints from concentrated use when the primary users
are overloaded. Thus for both efficiency and quality we offer the following two hypotheses:
HYPOTHESIS 5: Concentration of knowledge repository use has a lower effect on efficiency
and quality in teams that are more geographically dispersed.
HYPOTHESIS 6: Concentration of knowledge repository use has a lower effect on efficiency
and quality in teams that encounter more task change.
Using What We Know
- 13 -
3. Setting, Data, and Empirical Strategy
3.1 Setting
As of December 31, 2008, Wipro, the setting for this study, employed more than 75,000
employees worldwide and had annualized revenues of more than four billion dollars. To test our
hypotheses about the relationship between knowledge repository use and team performance, we focus our
attention on development projects. Development projects are challenging to coordinate and to deliver
successfully (Faraj and Sproull 2000; Boh, Slaughter and Espinosa 2007; Huckman et al. 2009)
Development projects are ideal for our purposes because objective performance measures and control
variables are available for all development teams, allowing for comparisons across projects.
Wipro employs an active knowledge management strategy, which has played a key role in the
company’s winning seven consecutive Asian MAKE (Most Admired Knowledge Enterprises) Awards
and three consecutive Global MAKE awards (Wipro 2010). A key part of the company’s knowledge
management initiative is its knowledge repository, and employees are encouraged to upload and
download content. In addition to inviting input from the workforce, a knowledge management team also
develops content. We focus our analysis on the consumption behavior (i.e., downloads) of individuals
within teams, as we are interested in how knowledge repository use affects team performance. In 2007 the
company changed its knowledge management website (named KNet), and as part of the changes included
advanced analytic technology to permit detailed tracking of system use.
3.2 Data
The empirical analysis draws on three sources of data: knowledge repository use, project
outcomes and characteristics, and human capital information. With the restructuring of its knowledge
repository in 2007, the company gained ability to track person-level use – the number of unique visits
(i.e., to different URLs) a person makes on a day. This data consists of person-day observations from
January 1, 2008, to December 31, 2009. Each URL corresponds to a unique knowledge artifact in the
knowledge repository. Although we know the number of unique visits each person made on any day, we
Using What We Know
- 14 -
have no information on the specific content a user views or how long they view a knowledge artifact.
Describing benefits of the knowledge repository, one project manager noted, “There are many
different types of documents on KNet. For example, there are case studies of prior projects that talk about
benefits, problems, customer value, new innovations, and best practices. There are also documents that
explain how a specific aspect of a technology or domain works. These include details about solving
particular problems such as the flow of development, steps to follow, and examples. Also, there are
reusable forms where someone can post code or an object.” Another project manager added, “Early in a
project, we can go to KNet and use it to find best practices. We can look at case studies and see lessons
learned and what issues different projects faced. All of this helps the team deliver better.” We note that
while there is some reusable code in the knowledge repository, the company reports that this is a small
percentage of the knowledge artifacts and that most artifacts are documents.
Our second data set consists of information about the 487 development projects that started on or
after January 1, 2008, and finished by December 31, 2009. We restrict our sample to these projects as we
wish to examine only those projects for which we have complete data on knowledge repository use. The
number of projects in our final sample for analysis is reduced further by the elimination of seventy
projects that do not use kilolines of code (KLOC) as the unit of measurement2 and eighty-seven projects
from customers who submitted only one project in the sample, as our models control for customer effects.
We control for customer effects in order to account for time-invariant aspects of customers that could
affect our performance measures (e.g., customer processes and legacy systems, difficulty in gaining
access to systems, etc.) Finally, we also collect complete human capital information on the 9,554
individuals involved in these projects. This information includes both demographic information and
individual project assignments since 2000 (individuals usually work on only one project at a time).
3.2.1 Dependent Variables. We use two classes of dependent variables: effort deviation, to
measure efficiency, and post-delivery defects, to measure quality. Table 1 includes summary statistics.
2 Projects that use either project-specific measures (e.g., a database project) or client-specific measures are not
comparable to those using KLOC, so are not included.
Using What We Know
- 15 -
****************************INSERT TABLE 1 HERE****************************
Project Efficiency. To evaluate the efficiency of a project, we examine whether the project meets its
estimated number of hours of work to be completed. We construct a variable, effort deviation, by
subtracting the project’s estimated effort (in person-hours) from its actual effort required (in person-
hours), then dividing that difference by the estimated effort to normalize for project size. Effort estimates
are initially made by sales and pre-sales personnel at Wipro. The estimate may be changed during the
course of a project, usually as a result of the customer’s changing a project’s scope. To make sure
estimates are not altered for inappropriate reasons (e.g., because a project is falling behind), a project
manager must receive both customer and internal Wipro management approval for the change. We use the
revised estimates to calculate effort deviation, since these estimates most accurately capture the final
goals and objectives of a project. Prior work examining software development also includes schedule (i.e.,
on-time) performance, to evaluate project efficiency. In 2008 and 2009, Wipro delivered over ninety
percent of its projects on time. Given this high rate of schedule performance, variation is insufficient to
analyze projects’ schedule performance. We therefore focus our analysis on effort performance.
Project Quality. In addition to examining project efficiency, we also examine post-delivery defects as a
measure for project quality. Following completion of development projects, customer acceptance testing
takes place. During customer acceptance testing, the customer or a third party tests the code against the
project’s pre-specified metrics. The output of this process is a count of the number of defects, or post-
delivery defects, a commonly used quality metric in software engineering (Boehm 1981). As a process
check, in cases where zero defects are recorded, the internal auditing group confirms testing took place.
3.2.2 Independent Variables. Our study has two classes of independent variables. The first class
comprises measures of knowledge repository use: mean use and concentration of use. The second class
consists of the interactions between the two measures of knowledge repository use and geographic
dispersion and task change. We standardize all independent variables by subtracting the mean and
dividing by the standard deviation to aid interpretation and limit multicollinearity of interaction effects
(Aiken and West 1991).
Using What We Know
- 16 -
Knowledge repository use. We measure a team’s repository use by counting the number of times a team
member accesses a unique URL on any given day during the duration of a project, summing this total for
all team members, then dividing by the number of team members. We calculate the mean to normalize for
the effects of team size.
Concentration of knowledge repository use. In addition to examining the amount of use by a given team,
we also wish to examine the distribution of the team’s use. To do this, we utilize the Herfindahl measure,
commonly used to capture the distribution of a characteristic across team members (Harrison and Klein
2007). The measure is calculated by first computing the percentage of downloads made by each
individual on the team (i.e., each individual’s downloads / total team downloads), then squaring and
summing these values for the entire team. Thus, in a project with highly concentrated use (e.g., only one
person downloaded from the knowledge repository), the Herfindahl measure would equal one.
Alternatively, for a project with equal use across all team members, the measure would equal 1 / N, where
N equals the size of the team.
Geographic dispersion. Project teams deploy individuals across two locations: Wipro’s Indian facilities –
“offshore” – and clients’ offices – “onsite” (31% of projects locate individuals at only one site). To
measure team members’ dispersion, we first calculate the percentage of the team’s hours spent offshore.
However, because we are interested in how work is dispersed rather than how much of the work is
completed offshore, we then create a variable, geographic dispersion, equal to the offshore percentage if
less than half the work is completed offshore, or equal to one minus the offshore percentage if more than
half the work is completed offshore. We calculate the variable in this manner since a team that is 65%
offshore and 35% onsite is as equally dispersed as a team that is 35% offshore and 65% onsite. We use
hours of effort, rather than individual’s location, to construct the variable since hours more precisely
measures the distribution of project effort. However, substituting a variable constructed using an
individual’s location (coded one or zero for offshore or onsite) yields similar results for all hypotheses.
Task Change. Development project requirements are delineated before a project begins. During the
project the requirements may be changed by the customer. To capture this we construct a measure, task
Using What We Know
- 17 -
change, which is equal to the percentage of requirements that are changed during a project. When a
project’s requirements are changed the effort estimates are often changed. Since the development of
software is a complex, interconnected process, when tasks are changed in-process the complexity of the
project significantly increases (Wallace et al. 2004; Rai et al. 2009). Thus, even after taking into account
the revision in the estimates, we expect a negative effect of task change on performance.
3.2.3 Control Variables.
Team familiarity. Prior work shows that team familiarity, defined as members’ prior shared work
experience, may enable performance in a team setting (Reagans, Argote and Brooks 2005; Espinosa et al.
2007; Huckman et al. 2009). Therefore, we control for this effect by calculating a team familiarity
variable equal to the sum of the number of times each unique dyad on the team has worked together on
any project over the prior three years. We divide that sum by the number of unique dyads on the team, to
scale the variable (Reagans et al. 2005).
Role experience. Prior work has established that role experience is a useful measure in knowledge-work
settings (Huckman et al. 2009). Role experience averages the number of years each individual on the team
has served in their present, hierarchical role (i.e. project manager, middle manager, or project engineer).
To calculate the team value, we average across individuals, weighting each individual’s calculated value
by the number of days he or she was on the team.
Task complexity. We measure task complexity using kilolines of new source code (KLOC). As KLOC
grows, complexity increases due to increasing project scale and interactions within the code. KLOC is a
commonly used measure to evaluate software complexity (MacCormack, Verganti and Iansiti 2001) and it
is the measure that Wipro uses to evaluate project complexity (the company does not widely track
function points). The company uses a regimented approach for counting lines of code. As software has
been shown to exhibit scale effects, we use the log of KLOC in our models (Banker and Kemerer 1989).
Estimated effort. Projects involving more hours of effort may in turn be more difficult to complete.
Therefore, we include the log of the estimated total person-hours for a given project. We use the
estimated value since a project that is delivered over budget (i.e., takes more hours) would have a larger
Using What We Know
- 18 -
actual effort value than a project that meets its estimates.
Team size. If a team becomes too large, then adding members could create coordination or integration
challenges (Heath and Staudenmayer 2000). Alternatively, if a team is small, then adding team members
could prove useful, as doing so increases the knowledge resources available to the team (Hackman 2002).
Therefore, we control for team size by including the log of total personnel who worked on the project.
Including the variable without a log transformation generates the same pattern of results.
Estimated duration. We control for project duration since a relatively long project may be more difficult
or face greater likelihood of employee attrition (Ethiraj et al. 2005) than a shorter project. We use the log
of the estimated value (in days) to avoid the same endogeneity concern as with effort. Estimated duration
and estimated effort are correlated variables (Pearson coefficient = 0.62). We include both variables as
they capture additional information about a project’s characteristics. However, we note that dropping
either variable generates the same pattern of results that we present.
Contract type. Development contracts use either a time-and-materials (i.e., cost-plus) structure or a fixed-
price structure (Banerjee and Duflo 2000). In the former case, a customer pays the negotiated rate for
number of hours worked on the project while in the latter case a set payment is agreed to prior to the start
of the project. We use an indicator variable, contract type, to control for these differences, such that the
variable equals one if the contract is for a fixed price, and zero if it is for time and materials.
Software languages. We control for both the number and type(s) of software languages used in a project.
For the former, we include an indicator set to one if a project uses more than one software language
(which is the case for 53% of projects). For the latter, we include indicator variables for the different
types of languages used (C, C++, Java, query (e.g., SQL), markup (e.g., HTML), BASIC).
Technologies. Just as projects may include multiple software languages, they also might cover multiple
technology classes (e.g., client server, e-commerce). Therefore, we include an indicator equal to one if a
project includes more than one technology, and zero otherwise (10% and 90% of projects, respectively).
3.3 Empirical Strategy
Since effort deviation is a continuous variable and post-delivery defects is a count variable, we use
Using What We Know
- 19 -
different regression models for each. In each case, we wish to control for time-invariant attributes of
customers that can affect performance (Greene 2003). For effort deviation, we use a fixed-effects linear
regression model since a Hausman test rejects the null hypothesis that the random effects model is
consistent (p<0.05). For post-delivery defects, we use a conditional fixed effects negative binomial
model, as the data exhibits overdispersion (Cameron and Trivedi 1998). Since the model conditions on
the customer, as opposed to just including a customer fixed effect, we lose all customers in which the
defect variable does not vary from zero. This yields a final sample for defects of 255 projects. We
compare the projects used in the quality models to the seventy-five projects used only in the efficiency
models and find no statistically significant differences in the two samples for the variables of interest.
4. Results
Table 2 shows results for the study’s regression models. Columns 1 and 2 correspond to the models for
effort deviation, while Columns 3 and 4 report results for post-delivery defects. In Column 1, we report
the main effects of the knowledge use and concentration variables. First, as Hypothesis 1 predicts, we see
that knowledge repository use is related to improved performance (specifically, a negative coefficient
corresponds to lower deviation). A one-standard-deviation increase in repository use is related to a
reduction in average effort deviation of 27%. Examining concentration of use, we find that higher
concentration of use is related to lower effort deviation. This finding supports Hypothesis 4a. A one-
standard-deviation increase in the concentration of knowledge repository use is related to a 40% decrease
in average effort deviation.
****************************INSERT TABLE 2 HERE****************************
Moving to Column 2, we add the interaction variables. Although Hypothesis 2 predicts that
repository use will have a greater effect on efficiency performance when teams are more geographically
dispersed, the coefficient on the interaction terms is not statistically significant, failing to provide support
for the hypothesis. Examining the interaction of repository use and task change, we find the coefficient to
be negative and statistically significant. This result supports Hypothesis 3, indicating that repository use is
Using What We Know
- 20 -
related to higher performance when tasks change. Inspecting the coefficient of concentration of repository
use and geographic dispersion (Hypothesis 5), we find that instead of a positive coefficient, the
interaction is negative and significant at a ten percent level. We explore this result further below. Finally,
the coefficient for the interaction of concentration of use and task change is not statistically significant.
In Figures 2a and 2b, we investigate the significant interaction results in more detail by plotting
the effects for low and high values of each variable (one standard deviation below and above the mean for
geographic dispersion, and zero task change and one standard deviation above the mean for task change).
Figure 2a plots the interaction of repository use with task change. As expected, increasing knowledge use
is related to improved performance in both the high and low cases, however, the improvement is greater
when task change is high. Figure 2b plots the results for the interaction of concentration of repository use
and geographic dispersion. This plot highlights how the result differs from our expectation in Hypothesis
5. Rather than concentrated knowledge use providing less value for highly geographically dispersed
team, as compared to teams with low geographic dispersion, the result is just the opposite. Put another
way, highly geographically dispersed teams benefit even more from concentrated knowledge use than do
teams that have little geographical dispersion. We explore this finding further in the discussion section.
**************************INSERT FIGURES 2a and 2b HERE**************************
In Column 3 of Table 2, we report the effect of knowledge repository use and concentration of
use on post-delivery defects. First, we find that knowledge repository use is not significantly related to the
quality outcome, which result fails to support Hypothesis 1. However, in support of Hypothesis 4 we see
that more concentrated knowledge repository use is related to worse performance (i.e., higher expected
defects). A one-standard-deviation increase in concentration of use is related to a 35% increase in
expected defects. In Column 4, we add the four interaction terms. We find that the interactions of
repository use with geographic dispersion and task change are negative and significant (related to fewer
expected defects), supporting Hypotheses 2 and 3, respectively. Also the coefficient for concentration of
repository use and geographic dispersion is positive and significant, providing support for Hypothesis 5.
Finally, the coefficient for the interaction of the concentration of repository use and task change is not
Using What We Know
- 21 -
statistically significant, failing to support Hypothesis 6. Figure 3 plots the statistically significant results
for low and high values of the variables and provides graphical support for Hypotheses 2, 3, and 5.
**************************INSERT FIGURES 3a, 3b, and 3c HERE*************************
Several limitations to our work bear mentioning. First, while our dataset provides detailed
information on when and how many times individuals download information, we have no detail on what
they download. If possible, future work should explore how similarities and differences in the materials
downloaded by team members affect performance. Second, while we measure knowledge repository use
we do not know how (or if) the knowledge found in the repository was eventually used in the team’s
work. Although we hypothesize and find effects of knowledge repository use, future research should seek
to understand in more detail how knowledge accessed in the repository is used to improve team
performance. Third, although we are able to measure how much knowledge stored in the repository
comes into the team, we do not have information on how that knowledge is subsequently shared within
the team. Given the important role of knowledge sharing in teams (Bunderson and Sutcliffe 2002), future
work may wish to examine how the sharing of knowledge affects the relationships we study. Fourth, our
variable that operationalizes geographic dispersion only measures whether individuals on a team are
located at Wipro’s facilities in India or at the customer’s location, not whether they are spread out across
Wipro facilities in India. Wipro managers report that teams are typically located in one place within India,
and the level of detail we have on individuals’ locations is more specific than that for much other work on
this topic. Even though it is not clear whether this unmeasured variation would increase the likelihood of
finding the relationships we identify, it would be preferable to have the additional information. Finally,
while the empirical dataset we use is both large and detailed, it comes from one organization. This is a
necessary, but less than ideal, consequence of gaining access to a research site and collecting such data.
Nevertheless, future work may seek to examine these findings in other companies and settings.
5. Discussion
In this paper we hypothesize and find that team use of a knowledge repository, together with the
Using What We Know
- 22 -
appropriate organization of that use within the team, enhances team performance. We also find that teams
that are geographically dispersed and face higher levels of task change see differential performance
effects with their use of a knowledge repository, at least in some cases. We find that repository use is
related to improved efficiency but not improved quality. Our interviews at Wipro revealed perspectives
consistent with these results, and offer insight into underlying mechanisms. For example, one project
manager noted that by using the knowledge repository, team members can gain access to improved
solutions, thus shortening or even eliminating the need for some of their own problem-solving cycles,
improving efficiency. He explained, “On one project we had a development activity that took two to three
hours. We had to deploy code onto servers at a customer location so it could be tested. I realized that by
reordering the steps, I could create a simplified process that would do the same work in an hour and
fifteen minutes. I posted a document explaining this, and then others [either on his project or other
projects] could download it and save time.”
We did not find a main effect of repository use on the quality of code developed. This is
consistent with results found by Haas and Hansen (2007), in a study examining a relationship with
performance, rather than with our dependent variable of conformance quality. Delivering defect-free
code requires individuals to both complete and integrate their assigned tasks. Knowledge about how to
integrate system components may not be easily codifiable, but instead may require trial-and-error problem
solving (Pich, Loch and Meyer 2002). If so, then the amount of stored knowledge accessed may not be
sufficient to explain performance, because not all relevant knowledge can be codified. Alternatively,
while our sample of 255 projects is large for a study of teams, it is still possible that insufficient power
contributed to the lack of a statistically significant finding and that a larger sample might support such an
effect.
A second contribution of our study is to show that team and task-level factors -- in particular,
geographic dispersion and task change -- moderate the relationship between knowledge repository use
and team performance. Our results show that knowledge use is related to improved quality performance
for more dispersed teams, but did not show a similar effect on such teams for project efficiency.
Using What We Know
- 23 -
Illustrating how dispersed teams can gain more value from knowledge use, a project manager noted,
“When a team is spread out, then KM can be very beneficial. For example, the onshore team can go to
KNet to check a code review checklist instead of having to contact someone on the offshore team. This
makes code reviews faster and more accurate.” Stored knowledge resources create the opportunity for
dispersed team members to generate a shared mutual understanding, such as with respect to the quality
processes and checklists, which can improve performance (Cramton 2001). This project manager’s
perspective also may help explain why we see an effect for quality but not efficiency. The mutual
knowledge problem may create more issues for dispersed teams for quality performance than for project
efficiency. Quality requires integration across multiple parts of a project in a way that efficiency may not.
Future work is needed to investigate in more detail how dispersed teams’ ways of processing knowledge
stored in a repository affects performance outcomes.
With respect to the moderating effect of task change, we find that repository use is related to
improved project efficiency and quality for teams faced with increasing task change. As task change
increases, additional work may be created that may require knowledge that is not present within the team.
A knowledge repository offers teams a method to complete new work more efficiently and a means to fill
knowledge gaps. Thus, ensuring access to stored knowledge may be an important managerial lever for
those organizing teams faced with uncertain and changing environments.
Our third contribution builds on the idea that groups are information processing systems, by
investigating how the concentration of knowledge repository use within a team affects performance.
Holding knowledge repository use constant, concentrated use within the team was related to better project
efficiency but worse project quality. In other words, in this study, a divide and conquer approach – where
only some members of the team use the knowledge repository – is associated with efficiency, but not
quality, performance. Concentrating knowledge repository use creates an opportunity to develop
specialized skills and to build expertise; these experts (those who become facile in a knowledge
repository’s specific coding schemes and technical language) can presumably use the system quickly and
efficiently, contributing to project efficiency overall.
Using What We Know
- 24 -
In contrast, when teams appeared instead to share and share alike – with everyone on the team
using the knowledge repository – the quality of team performance was higher. When more individuals do
their own searching and downloading, they may attend more to entire knowledge artifacts, rather than
focusing only on the answer in a transactive manner; this more inclusive attention may also allow them to
identify additional related information that can improve the quality of their own code, thereby promoting
project quality (Leonardi 2007). However, the additional time expended in this way could hurt project
efficiency. This finding suggests that concentration of knowledge repository use might be used as a
strategic lever, based on a project’s performance objectives.
This study highlights the need to consider how knowledge repository use is structured within
teams. However, while the use of archival data allows us to make inferences about the objective
statistical relationships, we are limited in our ability to identify precisely the mechanisms underlying
these relationships. Future work should seek to elucidate the mechanisms within this black box. At
Wipro, senior management did not dictate patterns of use within teams; rather, a combination of project
level management and team member mutual adjustment led to these patterns. Examining these dynamics
directly in future work will aid in the theory building process.
In addition to the main effect of concentration of knowledge repository use on team performance,
we examined the moderating role of task change and geographic dispersion. Our hypotheses related to
interaction effects of task change and concentration of knowledge repository use were not supported. We
expected but did not find that a more distributed use pattern would aid teams facing task change; it is
possible that simply getting the relevant knowledge into the team (i.e., knowledge repository use)
dominates other effects in this high-velocity work context. Future research is needed to understand the
consequences of task change on knowledge repository use, as well as how best to respond to them.
Finally, our results show that geographic dispersion amplifies the main effect for concentration of
knowledge repository use found above (although only at a ten percent significance level for project
efficiency). While the relationship for quality performance is consistent with our hypothesis that
increasing geographic dispersion requires more distributed use, the efficiency result is not. Although
Using What We Know
- 25 -
geographic dispersion increases the coordination complexity that a team faces (Cummings et al. 2009), it
is possible that our results can be explained by the fact that dispersed teams are less able to share
knowledge about how best to use a knowledge repository (e.g., what structuring queries work best), such
that efficient usage from a concentrated structure is especially valuable. This would mean that informal
relationships within the team could play a key role in transferring such knowledge. Future work should
explore both knowledge use patterns and knowledge transfer mechanisms within geographically dispersed
teams in more detail. Overall, an important contribution of the present study came from considering how
different knowledge use behaviors within a team influence performance, rather than viewing the team as
an undifferentiated entity.
In practice, teams draw not only on knowledge stored within a repository but also on tacit
knowledge from within and outside the team (Nonaka and Takeuchi 1995). This paper focused on the use
of a knowledge repository to gain insight into how repository use is translated into project performance,
without attention to tacit knowledge. Our use of archival data as opposed to surveys gave us only limited
information on team members’ use of tacit knowledge. The lack of measures of tacit knowledge access is
a limitation of the present study; however, our decision to examine the effect of knowledge repository use
on performance is consistent with Haas and Hansen’s (2007) finding that different types of knowledge are
not substitutes for one another. Teams may differ in the ways they use tacit knowledge, but variation in
the use of the knowledge repository likely has distinct performance effects, worthy of focused study. As a
robustness check on this perspective, we ran a test of whether repository use and team familiarity (a
source of tacit knowledge) were substitutes for one another, with respect to performance.
We add the interaction of team familiarity and repository use (after standardizing team
familiarity) to our existing model to examine this question (detailed results available from authors). With
respect to both project efficiency and project quality, we find that the interaction term is not statistically
significant. This lends support to the view that the different types of knowledge are not substitutes for one
another because high levels of team familiarity do not correspond with the lower value of knowledge
repository use for team performance. Nevertheless, this remains an open area for future research.
Using What We Know
- 26 -
In this paper we examine how team knowledge repository use affects team performance, treating
repository use as an independent variable, however there are a number of interesting questions that would
instead examine what factors would encourage use of the repository, modeling knowledge repository use
as a dependent variable (Alavi and Leidner 2001). For example, future research can explore whether
knowledge repository use increases in smaller teams because those teams have a smaller stock of
expertise on which to draw.3
Our findings imply that knowledge stored and disseminated through information technology can
be a source of organizational learning (subject to interesting contingencies). An organization is said to
learn when organizational actions are improved in response to an increase in knowledge (Edmondson
2002). Consider, for example, a situation where a work team discovers a new way to execute a core
process. The team has created strategically valuable new knowledge, and to the extent that the team
reflectively adopts the new process, we can say that the team has learned. Has the organization learned?
By one narrow definition, in which organizational learning exists when new knowledge is created and
improves action somewhere in the organization, the organization has learned. By a broader definition of
organizational learning, however, the organization has not learned until that piece of knowledge affects
how work is done widely enough to be impactful at the organizational level.
Given variation in learning and knowledge capabilities across work units, together with the need
for widespread adoption of valuable knowledge for organizations to show performance benefits, how
should organizations actively manage this kind of learning? A formal mechanism for doing this might be
implementation of an information technology system like the one at Wipro. For knowledge contained in
the repository to be a source of organizational learning, it needs to have the potential to affect how work
is done. This paper provides insight into the value of stored knowledge when it is brought into team
projects; future research could explore how stored knowledge influences not only team performance, but
also organization-wide performance.
3 We thank an anonymous reviewer for this example and for suggesting this line of thought.
Using What We Know
- 27 -
6. Conclusion
This paper addresses questions related to how teams should organize to draw on organizational
knowledge resources to deliver novel output (Edmondson 2002; Nickerson and Zenger 2004). We
examined team knowledge repository use, taking a group-level perspective on how characteristics of the
task and team affect the relationship between knowledge repository use and team performance. Our
findings offer insight into how knowledge repository use can improve knowledge-worker productivity.
This paper makes several contributions to the study of knowledge, teams, and knowledge-work
productivity. First, we use fine-grained, objective data to show that knowledge repository use has a
positive effect on project efficiency, but not on project quality. Despite significant industry investment in
knowledge management systems (Davenport and Prusak 1998; Hansen, Nohria and Tierney 1999), there
has been, to the best of our knowledge, no large-scale archival study of these systems’ value (Alavi and
Leidner 1999; 2001). Second, we build on information processing theory to show that, holding the
amount of repository use constant, concentration of use affects team performance. Concentration has
opposite effects, helping improve project efficiency but harming project quality. Finally, we investigate
theoretically and practically compelling moderators of the relationship between knowledge repository use
and both dimensions of team performance. By examining team and task characteristics such as members’
geographic dispersion and task change, we find that teams whose conditions suggest a greater opportunity
to benefit from knowledge repository use do indeed show performance benefits.
These findings also provide insight into the microfoundations of organizational capability (Helfat
2000). In many knowledge-based organizations such as the one we studied, project work constitutes the
primary output delivered to customers (Ethiraj et al. 2005; Edmondson and Nembhard 2009). We
investigate how an organizational-level practice (e.g., creating an organizational knowledge repository)
helps build the project management capability of the organization, which is manifested at the group level
of analysis. Finally, our work suggests new managerial levers for action by identifying conditions under
which knowledge repository use can help to build an operations-based competitive advantage.
Using What We Know
- 28 -
7. References
Aiken, L. S. and S. G. West (1991). Multiple Regression: Testing and Interpreting Interactions. Thousand Oaks, CA,
Sage Publications.
Alavi, M. and D. E. Leidner (1999). "Knowledge management systems: Issues, challenges, and benefits."
Communications of the AIS 1(2): 1-10.
Alavi, M. and D. E. Leidner (2001). "Review: Knowledge Management and Knowlege Management Systems:
Conceptual Foundations and Research Issues " Mis Quarterly 25(1): 107-136.
Argote, L. (1999). Organizational Learning: Creating, Retaining, and Transferring Knowledge. Boston, Kluwer.
Argote, L. and P. Ingram (2000). "Knowledge transfer: A basis for competitive advantage in firms." Organ. Behav.
Human Decision Proc. 82(1): 150-169.
Argote, L., B. McEvily and R. Reagans (2003). "Managing knowledge in organizations: An integrative framework
and review of emerging themes." Management Sci. 49(4): 571-582.
Arrow, H. and J. E. McGrath (1995). Membership dynamics in groups at work: A theoretical framework. Research
in Organizational Behavior. B. Staw and L. Cummings. Greenwich, CT, JAI. 17: 373-411.
Arrow, K. J. (1974). The Limits of Organization. New York, W.W. Norton & Co.
Banerjee, A. V. and E. Duflo (2000). "Reputation effects and the limits of contracting: A study of the Indian
software industry." Quart. J. of Econ. 115(3): 989-1017.
Banker, R. D. and C. F. Kemerer (1989). "Scale economies in new software development." IEEE Transactions on
Software Engineering 15(10): 1199-1206.
Barney, J. B. (1986). "Strategic factor markets." Management Sci. 32(10): 1231-1241.
Bock, G. W., R. Sabherwal and Z. J. Qian (2008). "The effect of social context on the success of knowledge
repository systems." IEEE Transactions on Eng. Mgmt. 55(4): 536-551.
Bock, G. W., R. W. Zmud, Y. G. Kim and J. N. Lee (2005). "Behavioral intention formation in knowledge sharing:
Examining the roles of extrinsic motivators, social-psychological forces, and organizational climate." MIS
Quarterly 29(1): 87-111.
Boehm, B. (1981). Software Engineering Economics. New York, Prentice Hall.
Boh, W. F., S. A. Slaughter and J. A. Espinosa (2007). "Learning from experience in software development: A
multilevel analysis." Management Sci. 53(8): 1315-1331.
Brannen, M. Y. (2004). "When Mickey loses face: Recontextualization, semantic fit, and the semiotics of
foreignness." Acad. of Management Rev. 29(4): 593-616.
Bresman, H. (2010). "External learning activities and team performance: A multimethod field study." Organ. Sci.
21(1): 81-96.
Bunderson, J. S. (2003). "Team member functional background and involvement in management teams: Direct
effects and the moderating role of power centralization." Acad. of Management J. 46(4): 458-474.
Bunderson, J. S. and K. M. Sutcliffe (2002). "Comparing alternative conceptualizations of functional diversity in
management teams: Process and performance effects." Acad. of Management J. 45(5): 875-893.
Using What We Know
- 29 -
Burns, T. and G. M. Stalker (1961). The Management of Innovation. London, Tavistock Publications.
Cabrera, A. and E. F. Cabrera (2002). "Knowledge-sharing dilemmas." Organ. Studies 23(5): 687-710.
Cameron, A. C. and P. K. Trivedi (1998). Regression Analysis of Count Data. New York, Cambridge Univ Press.
Cohen, S. G. and D. E. Bailey (1997). "What makes teams work." J. of Management 23(3): 239-290.
Cohen, W. M. and D. A. Levinthal (1990). "Absorptive capacity: A new perspective on learning and innovation."
Admin. Sci. Quart. 35(1): 128-152.
Cramton, C. D. (2001). "The mutual knowledge problem and its consequences for dispersed collaboration." Organ.
Sci. 12(3): 346.
Cummings, J. N. (2004). "Work groups, structural diversity, and knowledge sharing in a global organization."
Management Sci. 50(3): 352–364.
Cummings, J. N., J. A. Espinosa and C. K. Pickering (2009). "Crossing spatial and temporal boundaries in globally
distributed projects: A relational model of coordination delay." Info. Systems Res. 20(3): 420-439.
Davenport, T. H. and L. Prusak (1998). Working Knowledge: How Organizations Manage What They Know.
Boston, Harvard Business School Press.
De Dreu, C. K. W., B. A. Nijstad and D. van Knippenberg (2008). "Motivated information processing in group
judgment and decision making." Personality and Social Psychology Review 12(1): 22-49.
Edmondson, A. C. (2002). "The local and variegated nature of learning in organizations: A group-level perspective."
Organ. Sci. 13(2): 128.
Edmondson, A. C. and I. A. Nembhard (2009). "Product development and learning in project teams: The challenges
are the benefits." J. of Product Innovation Management 26(2): 123-138.
Espinosa, J. A., S. A. Slaughter, R. E. Kraut and J. D. Herbsleb (2007). "Familiarity, complexity, and team
performance in geographically distributed software development." Organ. Sci. 18(4): 613-630.
Ethiraj, S. K., P. Kale, M. S. Krishnan and J. V. Singh (2005). "Where do capabilities come from and how do they
matter? A study in the software services industry." Strategic Management J. 26(1): 25-45.
Faraj, S. and L. Sproull (2000). "Coordinating expertise in software development teams." Management Sci. 46(12):
1554-1568.
Fisher, M. L. and C. D. Ittner (1999). "The impact of product variety on automobile assembly operations: Empirical
evidence and simulation analysis." Management Sci. 45(6): 771-786.
Fleming, L. (2001). "Recombinant uncertainty in technological search." Management Sci. 47(1): 117-132.
Fleming, L. and O. Sorenson (2004). "Science as a map in technological search." Strategic Management J. 25(8-9):
909-928.
Frei, F. X. (2006). "Breaking the trade-off between efficiency and service." Harvard Business Rev. 84(11): 92-101.
Galbraith, J. R. (1974). "Organization design: An information processing view." Interfaces 4(3): 28-36.
Garud, R. (1997). On the distinction between know-how, know-what and know-why. Advances in Strategic
Management. A. Huff and J. Walsh. Greenwich, CT, JAI Press. 14: 81-101.
Using What We Know
- 30 -
Garud, R. and A. Kumaraswamy (2005). "Vicious and virtuous circles in the management of knowledge: The case
of Infosys Technologies." Mis Quarterly 29(1): 9-33.
Gibson, C. and F. Vermeulen (2003). "A healthy divide: Subgroups as a stimulus for team learning behavior."
Admin. Sci. Quart. 48(2): 202-239.
Gibson, C. B. and S. G. Cohen (2003). Virtual Teams That Work. San Francisco, Jossey-Bass.
Gibson, C. B. and J. L. Gibbs (2006). "Unpacking the concept of virtuality: The effects of geographic dispersion,
electronic dependence, dynamic structure, and national diversity on team innovation." Admin. Sci. Quart.
51(3): 451-495.
Gino, F. (2008). "Do we listen to advice just because we paid for it? The impact of advice cost on its use." Organ.
Behav. Human Decision Proc. 107(2): 234-245.
Greene, W. H. (2003). Econometric Analysis. Upper Saddle River, N.J., Prentice Hall.
Guzzo, R. A. and G. P. Shea (1992). Group performance and intergroup relations in organizations. Handbook of
Industrial and Organizational Psychology. M. D. Dunnete and L. M. Hough. Palo Alto, CA, Consulting
Psychologists Press: 269-313.
Haas, M. R. (2006a). "Acquiring and applying knowledge in transnational teams." Organ. Sci. 17(3): 367 - 384.
Haas, M. R. (2006b). "Knowledge gathering, team capabilities, and project performance in challenging work
environments." Management Sci. 52(8): 1170-1184.
Haas, M. R. and M. T. Hansen (2005). "When using knowledge can hurt performance: The value of organizational
capabilities in a management consulting company." Strategic Management J. 26(1): 1-24.
Haas, M. R. and M. T. Hansen (2007). "Different knowledge, different benefits: toward a productivity perspective
on knowledge sharing in organizations." Strategic Management J. 28(11): 1133-1153.
Hackman, J. R. (2002). Leading Teams: Setting the Stage for Great Performances. Boston, HBS Press.
Hansen, M. T. (1999). "The Search-Transfer Problem: The role of weak ties in sharing knowledge across
organization subunits." Admin. Sci. Quart. 44(1): 82-111.
Hansen, M. T. and M. R. Haas (2001). "Competing for attention in knowledge markets: Electronic document
dissemination in a management consulting company." Admin. Sci. Quart. 46(1): 1-28.
Hansen, M. T., N. Nohria and T. Tierney (1999). "What's your strategy for managing knowledge?" Harvard
Business Rev. 77(2): 106-116.
Harrison, D. A. and K. J. Klein (2007). "What's the difference? Diversity constructs as separation, variety, or
disparity in organizations." Acad. of Management Rev. 32(4): 1199-1228.
Heath, C. and N. Staudenmayer (2000). Coordination neglect: How lay theories of organizing complicate
coordination in organizations. Research in Organizational Behavior. B. M. Staw and R. I. Sutton.
Helfat, C. E. (2000). "Guest editor's introduction to the special issue: The evolution of firm capabilities." Strategic
Management J. 21(10/11): 955-959.
Henderson, R. M. and K. B. Clark (1990). "Architectural innovation: The reconfiguration of existing product
technologies and the failure of established firms." Admin. Sci. Quart. 35(1): 9-30.
Using What We Know
- 31 -
Hinds, P. and S. Kiesler (2002). Distributed Work. Boston, MIT Press.
Hinds, P. J. and D. E. Bailey (2003). "Out of sight, out of sync: Understanding conflict in distributed teams." Organ.
Sci. 14(6): 615-632.
Hinds, P. J. and M. Mortensen (2005). "Understanding conflict in geographically distributed teams: The moderating
effects of shared identity, shared context, and spontaneous communication." Organ. Sci. 16(3): 290-307.
Hinsz, V. B., R. S. Tindale and D. A. Vollrath (1997). "The emerging conceptualization of groups as information
processors." Psychological Bulletin 121(1): 43-64.
Huckman, R. S. and B. R. Staats (Forthcoming). "Fluid tasks and fluid teams: The impact of diversity in experience
and team familiarity on team performance." Manufacturing Service Oper. Management.
Huckman, R. S., B. R. Staats and D. M. Upton (2009). "Team familiarity, role experience, and performance:
Evidence from Indian software services." Management Sci. 55(1): 85-100.
Jarvenpaa, S. L. and D. E. Leidner (1999). "Communication and trust in global virtual teams." Organ. Sci. 10(6):
791-815.
Kankanhalli, A., B. C. Y. Tan and K. K. Wei (2005). "Understanding seeking from electronic knowledge
repositories: An empirical study." J. Am. Soc. Inf. Sci. Technol. 56(11): 1156-1166.
Katz, D. and R. L. Kahn (1966). The Social Psychology of Organizations. New York, Wiley.
Kogut, B. and U. Zander (1992). "Knowledge of the firm, combinative capabilities, and the replication of
technology." Organ. Sci. 3(3): 383-397.
Leonardi, P. M. (2007). "Activating the informational capabilities of information technology for organizational
change." Organ. Sci. 18(5): 813-831.
MacCormack, A., R. Verganti and M. Iansiti (2001). "Developing products on "Internet time": The anatomy of a
flexible development process." Management Sci. 47(1): 133-150.
March, J. G. and H. A. Simon (1993). Organizations. Cambridge, MA, Blackwell.
Monteverde, K. (1995). "Technical dialog as an incentive for vertical integration in the semiconductor industry."
Management Sci. 41(10): 1624-1638.
Murphy, J. and J. Hackbush. (2007). "The knowledge management spending report, 2007–2008." AMR Research
Retrieved August 17, 2010.
Nickerson, J. A. and T. R. Zenger (2004). "A knowledge-based theory of the firm: The problem-solving
perspective." Organ. Sci. 15(6): 617-632.
Nonaka, I. and H. Takeuchi (1995). The Knowledge-Creating Company: How Japanese Companies Create the
Dynamics of Innovation. New York, Oxford University Press.
O'Leary, M. B. and J. N. Cummings (2007). "The spatial, temporal, and configurational characteristics of
geographic dispersion in teams." MIS Quart. 31(3): 433-452.
Pich, M. T., C. H. Loch and A. d. Meyer (2002). "On uncertainty, ambiguity, and complexity in project
management." Management Sci. 48(8): 1008-1023.
Rai, A., L. M. Maruping and V. Venkatesh (2009). "Offshore information systems project success: The role of social
Using What We Know
- 32 -
embeddedness and cultural characteristics." MIS Quart. 33(3): 617-641.
Reagans, R., L. Argote and D. Brooks (2005). "Individual experience and experience working together."
Management Sci. 51(6): 869-881.
Simon, H. A. (1962). "The architecture of complexity." Proceedings of the American Philosophical Society 106(6):
467-482.
Singh, J., M. T. Hansen and J. M. Podolny (2010). "The world Is not small for everyone: Inequity in searching for
knowledge in organizations." Management Sci. 56(9): 1415-1438.
Sole, D. and A. Edmondson (2002). "Situated knowledge and learning in dispersed teams." British Journal of
Management 13: S17-S34.
Sparrowe, R. T., R. C. Liden, S. J. Wayne and M. L. Kraimer (2001). "Social networks and the performance of
individuals and groups." Acad. of Management J. 44(2): 316-325.
Staats, B. R., D. J. Brunner and D. M. Upton (Forthcoming). "Lean principles, learning, and knowledge work:
Evidence from a software services provider." J. of Oper. Management.
Stasser, G. and W. Titus (1987). "Effects of information load and percentage of shared information on the
dissemination of unshared information during group discussion." J. of Pers. and Soc. Psych. 53(1): 81-93.
Szulanski, G. (1996). "Exploring internal stickiness: Impediments to the transfer of best practice within the firm."
Strategic Management J. 17: 27-43.
Teece, D. J., G. Pisano and A. Shuen (1997). "Dynamic capabilities and strategic management." Strategic
Management J. 18(7): 509-533.
Thompson, J. D. (1967). Organizations in Action. New Brunswick, NJ, Transaction Publishers.
Tushman, M. L. and R. Katz (1980). "External communication and project performance - An investigation into the
role of gatekeepers." Management Sci. 26(11): 1071-1085.
Tushman, M. L. and D. A. Nadler (1978). "Information processing as an integrating concept in organizational
design." Acad. of Management Rev. 3(3): 613-624.
Wallace, L., M. Keil and A. Rai (2004). "How software project risk affects project performance: An investigation of
the dimensions of risk and an exploratory model." Decision Sciences 35(2): 289-321.
Weick, K. E. and K. H. Roberts (1993). "Collective mind in organizations: Heedful interrelating on flight decks."
Admin. Sci. Quart. 38(3): 357-381.
Wernerfelt, B. (1984). "A resource-based view of the firm." Strategic Management J. 5(2): 171-180.
Wipro. (2010). "Wipro Technologies: Winner of the Global MAKE Award 2009." Retrieved June 14, 2010, from
http://www.wipro.com/corporate/media/newsdetail.aspx?id=1295.
Zellmer-Bruhn, M. and C. Gibson (2006). "Multinational organization context: Implications for team learning and
performance." Acad. of Management J. 49(3): 501-518.
Using What We Know
- 33 -
8. Figures and Tables
Figure 1. Research model examining teams’ knowledge repository use and team performance.
a. Knowledge repository use and task change.
b. Concentration of knowledge repository use and geographic
dispersion.
Figure 2. Plots of the interaction effects on effort deviation.
-2.50
-2.00
-1.50
-1.00
-0.50
0.00
0.50
1.00
1.50
2.00
2.50
3.00
Low Knowledge Repository Use High Knowledge Repository Use
Ch
ange
in E
ffo
rt D
evi
atio
n
No Task Change
High Task Change
-6.00
-5.00
-4.00
-3.00
-2.00
-1.00
0.00
1.00
2.00
3.00
4.00
Distributed Repository Use Concentrated Repository Use
Ch
ange
in E
ffo
rt D
evi
atio
n
Low Geographic Dispersion
High Geographic Dispersion
Using What We Know
- 34 -
a. Knowledge repository use and geographic dispersion. b. Knowledge repository use and task change.
Concentration of knowledge repostiory use and geographic dispersion
Figure 3. Plots of the interaction effects on defects.
-50.0%
-40.0%
-30.0%
-20.0%
-10.0%
0.0%
10.0%
20.0%
30.0%
40.0%
Low Knowledge Repository Use High Knowledge Repository Use
Ch
ange
in D
efe
cts
Low Geographic Dispersion
High Geographic Dispersion
-40.0%
-30.0%
-20.0%
-10.0%
0.0%
10.0%
20.0%
30.0%
40.0%
50.0%
Low Knowledge Repository Use High Knowledge Repository Use
Ch
ange
in D
efe
cts
No Task Change
High Task Change
-60.0%
-40.0%
-20.0%
0.0%
20.0%
40.0%
60.0%
Distributed Repository Use Concentrated Repository Use
Ch
ange
in D
efe
cts
Low Geographic Dispersion
High Geographic Dispersion
Using What We Know
- 35 -
Table 1. Summary statistics and correlation table of variables of interest
(n= 330, except for post-delivery defects, where n = 255).
Variable Mean σ 1 2 3 4 5 6 7 8 9 10 11 12
1. Effort Deviation -5.27 11.02
2. Defects 13.25 73.26 0.06
3. Knowledge Repository Usea 5.64 3.54 0.04 0.05
4. Concentration of Knowledge Repository Usea 0.25 0.26 -0.21 -0.06 -0.15
5. Geographic Dispersiona 0.14 0.14 -0.18 0.03 -0.09 -0.08
6. Task Changea 4.30 16.58 -0.01 -0.03 0.00 0.00 0.10
7. Team Familiarity 0.38 0.77 -0.10 -0.07 0.04 -0.05 0.12 0.11
8. Role Experience 1.64 0.80 -0.01 0.04 0.02 -0.07 0.04 -0.07 0.17
9. Project Complexity 3.15 1.87 0.03 0.00 0.18 -0.14 -0.05 0.13 -0.04 0.01
10. Estimated Effort 8.87 1.00 0.08 0.21 0.18 -0.39 0.12 -0.02 -0.19 -0.04 0.27
11. Team Size 3.05 0.80 0.11 0.17 0.04 -0.48 0.06 -0.01 -0.01 0.03 0.15 0.54
12. Estimated Duration 5.45 0.61 0.17 0.13 0.27 -0.24 0.07 0.04 -0.21 -0.19 0.20 0.62 0.33
13. Contract Type 0.46 0.50 -0.24 -0.06 -0.12 0.08 0.04 0.04 -0.10 0.05 0.00 0.05 -0.13 -0.04
Note. Bold denotes significance of less than 5%.
a In models this variable is standardized by subtracting the mean and dividing by the standard deviation. Values here are before standardization.
Using What We Know
- 36 -
Table 2. Summary results of the regression of effort deviation and post-delivery defects on knowledge
repository use (n = 330 and 255, respectively).
(1) (2) (3) (4)
-1.398** -1.724** -0.044 -0.196
(0.650) (0.694) (0.100) (0.120)
-2.091*** -2.086*** 0.351*** 0.339***
(0.748) (0.749) (0.118) (0.120)
-0.009 -0.228**
(0.720) (0.116)
-2.074** -0.588**
(1.038) (0.293)
-1.054* 0.218**
(0.586) (0.100)
0.182 -0.174
(0.925) (0.191)
-2.053*** -1.771** -0.008 -0.137
(0.775) (0.781) (0.108) (0.116)
0.160 0.459 0.195 0.057
(0.669) (0.697) (0.127) (0.146)
-1.977** -1.937** -0.401* -0.348
(0.968) (0.971) (0.217) (0.213)
0.467 0.767 0.063 0.017
(0.815) (0.818) (0.123) (0.124)
-0.079 -0.088 0.054 0.053
(0.357) (0.359) (0.064) (0.067)
-0.271 -0.404 0.149 0.180
(0.953) (0.952) (0.147) (0.158)
-2.011* -1.834* 0.037 -0.127
(1.094) (1.110) (0.168) (0.176)
4.151*** 4.388*** -0.048 0.011
(1.523) (1.514) (0.243) (0.247)
-2.824* -2.857* 0.188 0.130
(1.699) (1.707) (0.240) (0.244)
-17.022** -18.304** -3.010** -3.081**
(8.180) (8.135) (1.304) (1.277)
330 330 255 255
0.1699 0.1980 -371.8677 -364.1428
F Statistic / Wald chi-squared 2.4054*** 2.3767*** 26.9660*** 47.0853***
a Variable is standardized by subtracting the mean and dividing by the standard deviation.
Notes. *, ** and *** denote signficance at the 10%, 5% and 1% levels, respectively. Effort deviation models are GLS fixed-effects
models with heteroskedasticity robust standard errors, clustered on the customer. Defect models are conditional fixed effects negative
binomial regression models that condition on the customer. All models include, but results are not shown for the following variables:
number of languages, start year, software language, and number of technologies.
Dep Variable: Defects
Knowledge Repository Usea
Knowledge Repository Use ×
Geographic Dispersion
Role Experience
Overall R2
/ Log-likelihood
Observations
Constant
Contract Type
Concentration of Knowledge Repository
Use × Task Change
Task Changea
Concentration of Knowledge Repository
Usea
Knowledge Repository Use ×
Task Change
Estimated Effort
Estimated Duration
Team Familiarity
Geographic Dispersiona
Dep Variable: Effort
Deviation
Team Size
Concentration of Knowledge Repository
Use × Geographic Dispersion
Project Complexity
