Team Leader Experience in Improvement Teams

Transcription

Accepted Manuscript
Title: Team Leader Experience in Improvement Teams: A
Social Networks Perspective
Author: George S. Easton Eve D. Rosenzweig<ce:footnote
id="fn1"><ce:note-para id="npar0020">Tel.: +(404) 727
3326; fax: +404 727 2053.</ce:note-para></ce:footnote>
PII:
DOI:
Reference:
S0272-6963(15)00047-9
http://dx.doi.org/doi:10.1016/j.jom.2015.05.001
OPEMAN 909
To appear in:
OPEMAN
Received date:
Revised date:
Accepted date:
1-9-2014
29-4-2015
4-5-2015
Please cite this article as: Easton, G.S., Rosenzweig, E.D.,Team Leader Experience
in Improvement Teams: A Social Networks Perspective, Journal of Operations
Management (2015), http://dx.doi.org/10.1016/j.jom.2015.05.001
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Team Leader Experience in Improvement Teams:
cr
an
us
George S. Easton
The Goizueta Business School
Emory University
1300 Clifton Road NE
Atlanta, GA 30322
Email: george.easton@emory.edu
Phone: (404) 727-3326
Fax: (404) 727-2053
ip
t
A Social Networks Perspective
Ac
ce
p
*Corresponding Author
te
d
M
Eve D. Rosenzweig*
The Goizueta Business School
Emory University
1300 Clifton Road NE
Atlanta, GA 30322
Email: eve.rosenzweig@emory.edu
Phone: (404) 727-4912
Fax: (404) 727-2053
April 29, 2015
Page 1 of 55
Team Leader Experience in Improvement Teams: A Social Networks Perspective
Abstract
In this research, we disentangle the relationship between several key aspects of a team leader’s
ip
t
experience and the likelihood of improvement project success. Using the lens of socio-technical systems,
we argue that the effect of team leader experience derives from the social system as well as the technical
cr
system. The aspects of team leader experience we examine include team leader social capital (a part of
us
the social system) and team leader experience leading projects of the same type (a part of the technical
system).
an
We examine four different, yet related, dimensions of a team leader’s social capital, which we
motivate based on the social networks literature. One dimension, team leader familiarity, suggests that
M
social capital is created when team leaders have experience working with current team members on prior
improvement projects, and that such social capital increases the likelihood of improvement project
d
success. We develop three additional dimensions, using social network analysis (SNA), to capture the
te
idea that the improvement team leader’s social capital extends beyond the current team to include
everyone the leader has previously worked with on improvement projects. Contrasting our SNA-based
Ac
ce
p
dimensions with team leader familiarity enables us to better understand the impact of a team leader’s
social capital both inside and beyond the team. We also examine the effect of a team leader’s experience
leading prior projects of the same type, and consider the extent to which organizational experience may
moderate the impact of both team leader social capital and same-type project experience.
Based on analysis of archival data of six sigma projects spanning six years from a Fortune 500
consumer products manufacturer, we find that two of our SNA-based dimensions of team leader social
capital, as well as experience leading projects of the same type, increase the likelihood of project success.
In addition, we show that organizational experience moderates the relationship between team leader
same-type project experience and project success. However, this is not the case for the relationship
between the dimensions of team leader social capital and project success. These results provide insights
2
Page 2 of 55
regarding how dimensions of team leader experience and organizational experience collectively impact
the operational performance of improvement teams.
team leader experience; team leader social capital; social network analysis; learning;
quality improvement teams; six sigma
Ac
ce
p
te
d
M
an
us
cr
ip
t
Keywords:
3
Page 3 of 55
1. Introduction
Recent empirical research in the operations management literature that studies the effects of
experience on team performance has found a divergence in results between work teams and improvement
ip
t
teams. Specifically, while a team leader’s experience has been found to improve operational performance
in both work teams and improvement teams, the predominantly social dimension of team familiarity (the
cr
extent to which team members have prior experience working together) has not been found to matter in
improvement teams (Easton and Rosenzweig 2012; Huckman et al. 2009; KC and Staats 2012; Moore
us
and Lapre 2015; Reagans et al. 2005; Staats 2012). Because of this divergence in findings, and the
importance of team leader experience in both work teams and improvement teams, we explore the extent
an
to which there may be social dimensions of a team leader’s experience that affect performance in
improvement teams. This leads us to use a socio-technical systems lens to develop new hypotheses
M
concerning dimensions of team leader experience and their effects on improvement team performance.
From a socio-technical systems perspective, the effect of team leader experience on team
d
performance derives from both the social system and the technical system. The social system is
te
composed of people and their relationships, while the technical system pertains to processes, methods,
Ac
ce
p
materials, tools, and technology (Cummings 1978; Ketchum and Trist 1992; Trist 1981; Trist and
Bamforth 1951).
One aspect of the social system is social capital, which is a central theme in the management
literature on social networks (Borgatti and Foster 2003; Burt 1992; Granovetter 1985; Nahapiet and
Ghoshal 1998). Such research on social capital suggests that a team leader’s social capital should be
related to team performance. As a result, we develop dimensions of team leader experience that focus
both on team leader social capital internal to the team using a team familiarity-like measure, and team
leader social capital that extends beyond the current team using approaches based on social network
analysis (SNA) (Bonacich 1987; Wasserman and Faust 1994).
The dimension of team leader social capital internal to the team that we use in this paper is based on
recent work-team research by Staats (2012) and Moore and Lapre (2015), which highlights the
Page 4 of 55
importance of a team leader’s prior experience working with each of the other team members. We refer
to this kind of leader-focused familiarity as team leader familiarity. Perhaps it is just team leader
familiarity that has an effect in improvement teams rather than team familiarity more broadly as observed
ip
t
in work teams. Work teams and improvement teams differ in important ways, which could plausibly
explain the differences in the effects of experience between these two types of teams. Note that
cr
improvement teams are generally temporary and tend to focus on experimentation and learning, while
work teams tend to focus on delivering products and services and persist over time.
us
Team leader familiarity captures the social capital associated with the team leader working with the
same people over time. That is, the focus of team leader familiarity is on the relationships, developed
an
through shared experience, between the team leader and members of the current team. The concept of
social capital is broader, however, and we argue that it extends to the prior experience that a team leader
M
has working with a variety of people, whether or not they are members of the team leader’s current team.
That is, in lieu of working with the same people, what might matter more to improvement project success
d
is the team leader’s previous experience working with different people on prior improvement teams.
te
Such a perspective is consistent with research in the management literature on work teams that
Ac
ce
p
indicates that boundary-spanning social capital, in addition to social capital internal to the team, is
important for team performance (Chauvet et al. 2011; Edmondson and Nembhard 2009; Faraj and Yan
2009; Oh et al. 2004). This leads us to develop dimensions of team leader social capital that extend
beyond the team and capture the team leader’s “connectedness.” These dimensions are based on network
centrality methods from the SNA literature. Taking such a perspective is consistent with recent calls for
more applications of social network theory in operations management research (Borgatti and Li 2009;
Ketchen and Hult 2007; Kim 2014; Kim et al. 2011).
Returning to the technical aspect of the socio-technical systems lens, we also hypothesize that the
value of a team leader’s experience is influenced by proficiency with the technical system. This suggests
that characteristics of the team leader’s prior improvement projects may also influence the likelihood of
project success. Thus, experience leading the same type of project may be important. We consider
2
Page 5 of 55
projects to be of the same type if they require a similar problem-solving approach. We expect the
knowledge a team leader gains from leading one type of improvement project to be transferred to future
projects of the same type, therefore increasing the likelihood of project success.
ip
t
We also explore the potential for organizational experience, which includes both social and technical
aspects, to moderate the relationship between the different dimensions of team leader experience that we
cr
examine in this paper and the likelihood of improvement project success. In doing so, we address Argote
and Miron-Spektor’s (2011, 5) call to investigate when “different types of experience are complements or
us
substitutes for one another.”
While the bulk of the operations management literature on experience and teams focuses on work
an
teams, in contrast, in this research we study improvement teams. We test our hypotheses using six years
of data from improvement teams at a large Fortune 500 consumer products firm that has extensively
M
implemented six sigma. Based on logistic regression analysis using a sample of 152 team-based six
sigma projects, we find that the relationship between team leader experience and project success can be
d
explained by two of our SNA-based dimensions of team leader connectedness (a social aspect) as well as
te
by project type experience (a technical aspect). We go on to show that while organizational experience
Ac
ce
p
diminishes the positive impact of team leader project type experience, it does not alter the relationship
between team leader connectedness and project success. These results go beyond previous studies to
provide a more fine-grained analysis of the effects of a team leader’s experience on the performance of
improvement teams.
2. Conceptual Development
In this section, we provide the theoretical basis for our hypotheses concerning dimensions of team
leader experience and improvement team performance (see Figure 1). To provide context for the
development of these hypotheses, we begin by discussing improvement teams in six sigma systems, and
describe in more detail how they differ from work teams.
<< Insert Figure 1 about here >>
2.1
Improvement Teams in Six Sigma Systems
3
Page 6 of 55
In six sigma systems, operational improvement is primarily driven by team-based improvement
projects. These six sigma project teams are led and/or facilitated by improvement specialists (e.g., black
belts) who are rigorously trained in a structured problem-solving framework, and also in the use of
ip
t
statistical (e.g., control charts) and nonstatistical (e.g., flow charts) analysis tools (Choo et al. 2007b;
Linderman et al. 2003; Pyzdek and Keller 2009; Schroeder et al. 2008; Shafer and Moeller 2012; Swink
cr
and Jacobs 2012; Zu et al. 2008).
The majority of improvement projects in six sigma are led by black belts, who are extensively trained
us
process improvement specialists. Green belts typically receive basic six sigma training, and most often
serve as project team members. When green belts lead projects, they usually do so with the facilitation
an
and support of a black belt. Master black belts train and certify black belts, serve as coaches to green
belts and black belts, and work to implement six sigma throughout the organization.
M
In six sigma, improvement teams follow a formal, well-defined problem-solving process that is akin
to W. Edwards Deming’s Plan-Do-Study-Act cycle (Deming 1994). The most common form of the
d
problem-solving framework in six sigma is a five-stage process referred to as DMAIC (Define-Measure-
te
Analyze-Improve-Control) (Pyzdek and Keller 2009). This structured problem-solving framework guides
Ac
ce
p
the team in implementing the scientific method to diagnose problems and develop solutions in order to
achieve project goals (Linderman et al. 2003, 2006; Zu et al. 2008). The problem-solving process, along
with the set of analysis tools, represents standard team-based learning routines by which knowledge can
be created, acquired, and implemented (Argote and Miron-Spektor 2011; Benner and Tushman 2002,
2003; Choo et al. 2007a, b; Upton and Kim 1998).
In six sigma systems, the structured problem-solving framework generally serves not only to guide
deliberate team-based learning, but also to document learning activities (e.g., data analyses and findings).
Archiving this documentation to capture improvement team learning is integral to such six sigma systems
(Anand et al. 2010). Typical six sigma projects average six months in duration (Pyzdek and Keller 2009).
Improvement teams in six sigma differ from typical work teams in several important ways.
Improvement teams focus on experimentation and learning. Specifically, improvement teams engage in
4
Page 7 of 55
data collection, hypothesis generation, experimentation and analysis, development of evidence, and
deliberate capture of knowledge. This is not the primary focus of most work teams, which instead tend to
focus on delivering products or services (Cohen and Bailey 1997). In addition, improvement teams in six
ip
t
sigma follow a well-defined, structured problem-solving process as discussed above. All improvement
team members are typically trained in this process (Schroeder et al. 2008). While work teams might
cr
similarly employ some form of a clearly-defined performance strategy to guide the team’s work, such
strategies are much less likely to be as pervasive and standardized across all work teams as is the
us
structured improvement process in six sigma teams. Work team performance strategies are also far less
likely to focus on learning.
an
At the same time, a six sigma system is a parallel organizational structure that operates outside of an
organization’s normal structures and procedures (Schroeder et al. 2008). Thus, unlike work teams, six
M
sigma improvement teams are not part of the usual day-to-day work. In addition, improvement teams in
six sigma are formed to solve specific problems and disband upon completion, and are thus typically
d
relatively short-lived (Anand et al. 2010). While there is variation in work teams in this regard, work
Ac
ce
p
often of longer duration.
te
teams are usually formed to deliver specific products and services over a period of time, and are thus
For these reasons, different kinds of experience may very well affect the performance of work teams
and six sigma improvement teams in different ways. For example, because new improvement teams are
frequently being formed as problems arise, and are relatively short-lived, one might expect team
familiarity to be important in this context in order to reduce coordination losses in team formation and
functioning. That is, team members with high familiarity presumably already know how to work
together. However, the use of the structured problem-solving process, extensive training of team
members, and team leader facilitation mitigates, to some extent, the need for team familiarity since there
is already clarity about the improvement team performance strategy (Easton and Rosenzweig 2012;
Hackman 1987).
5
Page 8 of 55
In addition to facilitating the team’s use of the structured problem-solving process, team leaders play
other important roles in improvement projects. In many six sigma systems, team leaders propose the
problems to work on and recruit members to join their teams (Choo et al. 2007b). Leaders are critical to
ip
t
maintaining the team’s focus and ensuring that the improvement team remains within appropriate scope
and boundaries. Team leaders also manage the team’s schedule and ensure that milestones are met
cr
accordingly. Beyond facilitating the use of the problem-solving process, team leaders coach their
members in the use of analysis tools (Anand et al. 2010). As leaders, they assign and coordinate tasks
us
among team members, and deal with conflict as it arises. Improvement team leaders also help motivate
team members, and provide appropriate feedback and evaluation.
an
External to the team, leaders can be instrumental in obtaining access to information and data, and in
acquiring other necessary resources for the project (Schroeder et al. 2008). Improvement team leaders
M
may also play a role in garnering senior management support, removing external barriers to successful
2.2.
te
parts of the organization.
d
project completion, and in facilitating implementation of solutions, especially those that impact other
Team Leader Experience and Social Capital
Ac
ce
p
Socio-technical systems support the idea that work is more than just a technical system, and that the
social system is important to the success of teams (Cummings 1978; Ketchum and Trist 1992; Trist 1981;
Trist and Bamforth 1951). At the same time, a central theme in the management literature is that
individuals are embedded in social networks, and that these networks have impact on key outcomes like
performance (Granovetter 1985; Burt 1992; Uzzi 1996). One important aspect of the social system that
relates to embeddedness in social networks is social capital, which captures the value of relationships
between people (Borgatti and Foster 2003; Burt 1992; Granovetter 1985; Nahapiet and Ghoshal 1998).
Specifically, Nahapiet and Ghoshal (1998, 243) define social capital as “the sum of the actual and
potential resources embedded within, available through, and derived from the network of relationships
possessed by an individual or social unit.”
6
Page 9 of 55
Management research on work teams suggests that team leaders play a critical role in project
performance (Chauvet et al. 2011; Cohen and Bailey 1997; Devine et al. 1999; Hackman 2002) and,
further, that a team leader’s social capital is important (Balkundi and Harrison 2006; Oh et al. 2004).
ip
t
Similarly, in improvement teams, team leaders are considered to be a critical resource (Anand et al. 2010;
Choo et al. 2007b). In addition, a team leader’s experience, as measured by counting the number of prior
cr
improvement projects a team leader has led, has been found to be a key driver of improvement project
success (Easton and Rosenzweig 2012). In this study, we posit that a critical part of a team leader’s
working with other people on prior improvement projects.
us
effectiveness involves social capital, and that social capital extends to the team leader’s experience
an
We thus examine both the team leader’s experience working with the current team members on prior
improvement projects, and, more broadly, the team leader’s experience working with a variety of people
M
beyond those on the current team. The first perspective, which focuses on current team members, leads
us to examine team leader familiarity, a leader-based familiarity measure that rewards experience working
d
with the same people over time (Moore and Lapre 2015; Staats 2012). The second perspective, which
te
focuses on variety, takes the view that what might be more important to project success is working with
Ac
ce
p
different people on prior improvement teams. This leads us to develop team leader experience measures
based on SNA concepts (Bonacich 1987; Wasserman and Faust 1994). We further develop these ideas
and the corresponding hypotheses in subsections 2.2.1 and 2.2.2 below.
2.2.1.
Team Leader Familiarity
Some research on work teams supports the idea that the social capital created when team members
have prior experience working together leads to better outcomes than when team members have no such
experience (e.g., Huckman et al. 2009; Reagans et al. 2005). For example, with high team familiarity,
there are likely to be fewer start-up issues associated with that work team’s activities (Banker et al. 2001;
Tuckman 1965, 1977). However, as we discussed earlier, this result has not been observed in the context
of quality improvement teams (Easton and Rosenzweig 2012).
7
Page 10 of 55
Instead of capturing the experience working together between all team members, we argue that what
may be more important is a leader-focused form of familiarity involving only the relationships between
the team leader and each team member individually. Staats (2012) and Moore and Lapre (2015) examine
ip
t
this kind of team leader familiarity to capture the benefits of team leader coordination with team members
in software development and surgical teams, respectively.
cr
In improvement teams, our research context, it is reasonable to expect that close coordination
between a team leader and each team member is important. Prior experience working together provides
us
team leaders with an opportunity to personally know team members, and to become familiar with what
matters to them and how they work (Staats 2012). As a result, the team leader is presumably in a better
an
position to understand individual motivations and skillsets. This allows the team leader to better allocate
tasks among team members, and to better maintain team member engagement.
M
In addition, having worked together previously increases the likelihood that the team leader and team
members have shared norms. This means that team members may be better positioned to understand and
d
respond to the team leader’s expectations. Similarly, the team leader and members are more likely to
te
trust one another and, correspondingly, to promptly raise issues as they arise over the course of the project
Ac
ce
p
(Choo et al. 2007a; Edmondson 1999). Finally, previously developed relationships are likely to involve
mutual obligation, and therefore reciprocity, which can enhance commitment to one another and thus the
team (Cropanzano and Mitchell 2005). For these reasons, we expect team leader familiarity to improve
the likelihood of project success:
Hypothesis 1a:
2.2.2
Higher team leader familiarity is associated with increases in the likelihood of six
sigma project success.
Team Leader Connectedness
Beyond team leader familiarity, the value of a team leader’s experience with improvement projects
may be related to all of the “connections” the team leader has developed to members from prior
improvement teams (Brass and Krackhardt 1999; Leavitt 1951; Mehra et al. 2006; Yukl 2002). That is,
we argue that the value of an improvement team leader’s social capital extends beyond the current team to
8
Page 11 of 55
include everyone the leader has previously worked with on improvement teams. The idea that the team
leader’s social capital outside of their current team is important is suggested by Barden and Mitchell
(2007) and Uzzi (1996, 1997), who examine leader ties external to the firm.
ip
t
In the context of our study, we expect well-connected team leaders to be better able to identify
suitable individuals to join their improvement teams, either through direct connections or via referral
cr
(Granovetter 1973; Uzzi 1996). They are also more likely to have the necessary political astuteness to
propose improvement project topics that can be effectively pursued in the organization, and whose
us
solutions can actually be implemented. In addition, well-connected team leaders should be better able to
locate and obtain data and other information that the project requires (Chauvet et al. 2011; Mehra et al.
an
2006; Shaw 1964). Such data and information may be obtained from the team leader’s direct connections
or, indirectly, through those people their direct connections have worked with on prior improvement
M
teams.
Moreover, well-connected team leaders have experience working on improvement projects with a
d
variety of people with presumably a range of differing personalities. As a result, such leaders may be
te
better able to recognize important individual team member characteristics, and to correspondingly adjust
Ac
ce
p
how they facilitate the use of the problem solving process, what approaches they take to motivate
individuals, and to whom they assign tasks. While high connectedness, in contrast to high familiarity,
does not necessarily mean the team leader already knows a particular team member, it does mean that the
leader is more likely to have worked with someone with similar characteristics in the past, and thus more
likely to know how to relate to the member in the role as the team leader.
At the same time, not all of the team leader’s connections are necessarily of the same value. When a
team leader is connected to someone who has extensive experience with improvement projects, the team
leader is more likely to be able to obtain relevant and insightful advice regarding how to conduct their
project and lead their own improvement team. Furthermore, such well-connected connections are more
likely to be able to help to remove barriers or jointly resolve issues external to the team that might prevent
team progress (Uzzi 1996). In addition, well-connected connections may be very useful in supporting
9
Page 12 of 55
implementation of solutions developed by the team, especially those that potentially affect areas of the
organization not represented on the team.
To examine these ideas relating to the value of a team leader’s connectedness, we turn to a network-
ip
t
based view. Such a view allows us to take into account the team leader’s social context. That is, this kind
of network-based approach, as indicated in Chauvet et al. (2011), allows us to go “behind the chart” to
cr
look at the informal social structure as a dimension of the team leader’s experience that affects the
performance of the team. Specifically, for each improvement project, we consider the network to be
us
composed of the project’s team leader and all of the individuals who have previously worked on an
improvement project within the team leader’s business group. In this network, two people are connected
an
if they have the experience of working together on a prior improvement team.
We then use the concept of network centrality to capture a team leader’s connectedness. The most
M
basic idea of network centrality is that some positions in the network are more important (i.e., more
central) than others (Borgatti and Everett 2006; Leavitt 1951; Wasserman and Faust 1994). In addition,
d
the importance of a position (node) in the network may not only be determined by its relationship to the
te
overall network, but also by the local network structure surrounding that position (Burt 1992; Uzzi 1996).
Ac
ce
p
As we have argued above, we do not expect all of a team leader’s connections to others who have
worked on improvement projects to have the same value. Specifically, we posit that connections to
individuals who are, in turn, more highly connected result in more social capital because these
connections are both a more valuable direct resource and a more valuable source of referrals to others.
This leads us to our first approach in examining team leader connectedness, which is a network centrality
measure based on Bonacich centrality (Bonacich 1972, 1987) because it allows one to weight connections
to others based on how well connected those other individuals are (see subsection 3.2.3 for details
regarding this measure). We call this measure Bonacich connectedness and posit:
Hypothesis 1b:
Higher values of team leader Bonacich connectedness are associated with
increases in the likelihood of six sigma project success.
10
Page 13 of 55
As discussed earlier, in most six sigma systems, team leaders have the primary responsibility for
recruiting members for their improvement teams. Furthermore, recruiting the “right” members for the
team is a very important factor influencing team success (Faraj and Sproull 2000; Hackman 1987). By
ip
t
the “right” team members, we mean those who, in addition to having the requisite domain-related
knowledge, have bought into the six sigma initiative from a cultural perspective. This suggests that an
cr
important aspect of a team leader’s connectedness is how readily he/she can identify and reach such
potential team members.
us
Ready access to others can aid the team leader in additional ways as well, for example, by making
sources of needed information more easily available. That is, team leaders that are relatively “close” to
an
other individuals can communicate and interact without having to go through too many intermediaries.
Furthermore, information passed over fewer connections is less likely to suffer from distortion, so team
M
leaders with such ready access should obtain more accurate information (Bellamy et al. 2014; Hansen
2002; Peng et al. 2013).
d
Note that this idea of “ready access” is related to the idea behind team leader Bonacich
te
connectedness—that it is important to be connected to others who are, in turn, highly connected.
Ac
ce
p
Nevertheless, ready access places a somewhat different emphasis on the pattern of connections.
From a network perspective, ready access to an individual means that a team leader is connected to
that individual either directly or via a short path through a small number of other individuals. This idea is
captured by a network centrality measure called closeness centrality. Closeness centrality for an
individual (node) in a network is defined as the reciprocal of the average of the minimum path length to
every other individual in the network (Bavelas 1950; Sabidussi 1966; Wasserman and Faust 1994). We
thus develop a closeness connectedness measure based on a team leader’s closeness centrality (see
subsection 3.2.3 for details) and hypothesize:
Hypothesis 1c:
Higher values of team leader closeness connectedness are associated with
11
Page 14 of 55
Many papers in the literature on social networks have examined the effects of a network’s local
structure (e.g., Burt 1992; Granovetter 1973; Oh et al. 2004). One idea that has been the subject of much
research is that an individual’s connections are more valuable when those connections are not directly
ip
t
connected to one another (Borgatti and Li 2009). Such connections likely have access to different
sources, and presumably different kinds, of information. In contrast, when an individual’s connections
cr
are directly connected, they are more likely to provide the individual with redundant information, similar
advice, and so on. This idea is readily expanded to subnetworks. That is, when an individual’s
us
connections provide a view into relatively distinct subnetworks, they will likely result in the individual
having access to less-redundant information, more diverse perspectives, etc. This is the basic idea of a
an
“structural hole” (Burt 1992).
Similar arguments can be made in the context of the connectedness of improvement team leaders.
M
Team leaders with a wider variety of less-redundant information are more likely to be able to find and
recruit team members with distinct perspectives, identify important sources of data and information across
d
functional boundaries, obtain support from or even collaborate with other groups within the organization,
Ac
ce
p
improvement team.
te
and attain the support and cooperation necessary to implement the solutions generated by the
One commonly used centrality measure that, to some extent, captures whether or not a network
position has access to a wide variety of information is betweenness centrality. Specifically, betweenness
centrality measures the extent to which a node (individual) falls on the shortest path between all pairs of
other nodes in the network (Carnovale and Yeniyurt 2014; Freeman 1977; Mardsen 2002; Wasserman and
Faust 1994). From an information flow perspective, nodes with high betweenness will have access to a
relatively large portion of the information flowing between other nodes, and are thus likely to have a
broader view of the diversity of information and perspectives represented in the network.
In practice, individuals with high betweenness are often located at bridges, or areas with few
connections, between relatively isolated subnetworks (Freeman 2011). High betweenness occurs at such
individuals because they are the only way, or one of the few ways, that pairs of individuals in the separate
12
Page 15 of 55
subnetworks can connect to one another. In addition, such individuals who connect relatively isolated
subnetworks have either direct or close connections to individuals in the separate subnetworks which, as
argued previously, is likely to be associated with access to diverse information and perspectives.
ip
t
The above discussion leads us to develop a betweenness connectedness measure based on a team
leader’s betweenness centrality (see subsection 3.2.3 for details). We thus hypothesize the following:
Higher values of team leader betweenness connectedness are associated with
cr
Hypothesis 1d:
us
We note here that the concept of betweenness is related to, but at the same time, somewhat different
from, the ideas of Bonacich and closeness centrality. A common theme that relates all three measures of
an
network centrality is access to information. Betweenness is motivated more by information flow and, to a
lesser extent, by local network structure that involves connections between otherwise relatively distinct
M
subnetworks. Bonacich and closeness centrality are motivated more by access to information, either
because of connections that are well-connected or because the sources of the information are close.
Team Leader Project Type Experience
d
2.3
te
As mentioned in the Introduction, the technical system is comprised of processes, methods, materials,
tools, and technology. Key aspects of six sigma that are a part of the technical system include the
Ac
ce
p
structured problem-solving process and the supporting statistical and nonstatistical tools (Choo et al.
2007b; Linderman et al. 2003, 2006; Schroeder et al. 2008; Zu et al. 2008). However, not all six sigma
projects emphasize the same phases of the structured problem-solving process equally, or require the
same kind of data or analysis tools (Agrawal and Muthulingam 2015; Lapre et al. 2000; Pyzdek and
Keller 2009).
For example, one of the ways in which six sigma projects differ from one another is whether the
project focuses on reducing special cause or common cause variation. Projects that focus on elimination
of special cause variation often involve attribute variables, such as defect type, and emphasize use of
simple tools in order to generate hypotheses concerning possible root causes. In contrast, those focusing
on reduction of common cause variation are much more likely to use continuous variables and involve
13
Page 16 of 55
design of experiments (DOE). Correspondingly, projects focused on elimination of special cause
variation are likely to put more emphasis on early steps in the problem-solving process (e.g., the Measure
phase of DMAIC), while projects focused on reduction of common cause variation often involve DOE
ip
t
and generally put more emphasis on later phases of the problem-solving process (e.g., the Analyze phase
of DMAIC).
cr
From this point of view, six sigma projects are related, and thus of the same type, if they emphasize
similar steps in the problem-solving process, use similar kinds of data, and require similar analysis tools.
us
When we examine the six sigma projects from our research site, the projects naturally fall into five key
types as summarized in Table 1.
an
<< Insert Table 1 about here >>
In this study, we argue that a team leader’s experience leading a particular type of six sigma project is
M
applicable to future projects of that same type, and that such experience improves the likelihood of project
success. Research in the context of work teams supports this idea.
d
For example, in a laboratory setting, Schilling et al. (2003) found that the learning rate of work teams
te
was higher under conditions of related task variation than under conditions of unrelated task variation or
Ac
ce
p
even specialization. These authors attribute this effect to several factors. First, working on different, yet
structurally similar, kinds of problems may generate novel insights and synergies with respect to the logic
developed or the solutions applied. Second, over time, performing multiple related tasks is likely to result
in a deeper cognitive understanding of processes than would occur if only a single task were performed.
Several recent work-team studies involving archival data from organizations provide support for
Schilling et al.’s (2003) experiments, in that they find that completing multiple related tasks over time
accelerates learning and thus performance. Specifically, Muehlfeld et al. (2012) classify mergers and
acquisitions (M&As) into different types based on their structural similarity (e.g., within the same
country), and conclude that experience with a particular type of M&A improves the likelihood that
subsequent deals of that type will be completed. In addition, more in line with our study’s focus on team
leaders, KC and Staats (2012) analyze the impact of a surgeon’s experience with different types of cardiac
14
Page 17 of 55
procedures on the likelihood of patient mortality. These authors observe that increases in a surgeon’s
same-type procedure experience leads to improved surgical outcomes.
In the context of our study, improvement projects of the same type constitute “related task variation”
ip
t
in the sense of Shilling et al. (2003). Based on our arguments above, we expect a team leader’s prior
experience leading the same type of project to lead to a higher likelihood of project success. We call this
2.4
Increases in a team leader’s experience leading the same type of project (higher
team leader project type experience) is associated with increases in the likelihood
of six sigma project success.
us
Hypothesis 2:
cr
kind of experience team leader project type experience and posit:
Organizational Experience as a Moderating Variable
an
A growing body of empirical research suggests that organizational experience provides a distinct
positive contribution to learning outcomes over and above other dimensions of experience (Clark et al.
M
2013; Huckman and Pisano 2006; Reagans et al. 2005). In the context of our study, organizational
experience captures the scope and maturity of the six sigma system, and is based on the cumulative
d
number of improvement projects initiated (Agrawal and Muthulingam 2015; Choo 2014; Easton and
te
Rosenzweig 2012; Lapre and van Wassenhove 2001; Lapre et al. 2000). Conceptualizing organizational
Ac
ce
p
experience in terms of cumulative projects initiated is consistent with definitions of organizational
experience based on an organization’s cumulative production volume (Argote et al. 1990; Pisano et al.
2001; Reagans et al. 2005).
As an organization gains experience deploying a six sigma system, we expect that both the social and
technical systems will evolve in a manner that supports effective execution of improvement projects. For
example, changes in the technical system likely include “debugged” and improved training, and an
expanding database of archived improvement projects (Cummings 1978). Such archived projects
represent captured learning that can be drawn upon by individuals on subsequent projects (Anand et al.
2010; Faraj and Sproull 2000; Haas 2006; Kim 1993). Changes in the social system likely include
development of shared representations, interpretations, and systems of meaning resulting in a common
view of the scope and nature of the six sigma system (Nahapiet and Ghoshal 1998). In addition, an
15
Page 18 of 55
overall culture of trust in the six sigma system and buy-in to the improvement process is likely to develop.
Such cultural changes should increase willingness to participate on improvement project teams, and
enhance cooperation with the implementation of project solutions. Increasing organizational experience,
ip
t
therefore, is likely to increase the odds of project success (Easton and Rosenzweig 2012). Accordingly,
we add organizational experience to our analysis.
cr
As mentioned in the Introduction, organizational experience may moderate the effects of the different
dimensions of team leader experience that we examine. Specifically, we hypothesize that increases in
us
organizational experience diminish the expected positive effects of the team leader social capital-related
dimensions (team leader familiarity, Bonacich connectedness, closeness connectedness, and betweenness
an
connectedness) and of team leader project type experience.
The expected moderating effect of organizational experience could result from a variety of
M
mechanisms. For example, the team leader can search the six sigma project database for prior projects of
the same type, and examine the methods used and the solutions developed. The team leader can also use
d
this database to quickly identify individuals who have expertise with a particular project type—and to do
te
so without having worked directly with those individuals on prior six sigma projects or knowing about
Ac
ce
p
those individuals through his/her connections. In this way, organizational experience may, to some
extent, substitute for the project type experience and social capital aspects of a team leader’s prior
experience.
The culture of buy-in associated with high organizational experience likely reduces the need for a
team leader to be particularly skilled at selecting the “right” team members in terms of six sigma
receptivity, thus reducing the advantage provided by the team leader’s social capital. Similarly, the level
of commitment to the six sigma system associated with high organizational experience means that a team
leader may not have to rely a great deal on his/her connections to implement a project’s solution
successfully. That is, when organizational experience is high, a team leader is less likely to have to rely
upon personal influence through his/her connections to obtain cooperation with implementation of the six
sigma project results.
16
Page 19 of 55
Taken together, we expect increases in organizational experience to lessen the positive association
between six sigma project success and both the team leader’s social capital dimensions and project type
experience:
As organizational experience increases, the positive relationship between our
team leader social capital-related dimensions (team leader familiarity, Bonacich
connectedness, closeness connectedness, betweenness connectedness) and the
likelihood of six sigma project success diminishes.
Hypothesis 4:
As organizational experience increases, the positive relationship between team
leader project type experience and the likelihood of six sigma project success
diminishes.
us
cr
ip
t
Hypothesis 3:
3. Measures and Methods
an
In this section, we discuss the six sigma improvement project data at our research site and the
measures used in our analysis. Most of the measures were culled directly from the project documentation
M
(e.g., six sigma project success) while others involved coding by our research team (e.g., use of statistical
tools), which we discuss below. We note here that this paper is a part of a larger, ongoing research
te
3.1 Project Data and Coding
d
project (Easton and Rosenzweig 2012).
Our six years of improvement project data comes from a Fortune 500 firm that has successfully and
Ac
ce
p
extensively implemented six sigma throughout its business groups. This firm (hereafter “RS”) is a large,
market leading manufacturer and marketer of consumer products, with over 15,000 employees
worldwide. RS invested in the six sigma initiative with the intent to improve its overall performance
along dimensions such as quality, cost, and cycle time.
RS began its six sigma deployment in 2002 by training approximately twenty-five to thirty people,
with the first project initiated early in 2003. The widespread deployment of six sigma at RS occurred in
2005 and 2006. Over two hundred projects had been closed at the time of our data collection in
December 2008. According to the classification in Jacobs et al. (2015), RS would be considered a
relatively late adopter of six sigma since it began its deployment after 2001.
17
Page 20 of 55
Six sigma project teams at RS are led and/or facilitated by green belts, black belts, and master black
belts, all of whom have been trained in a structured problem-solving framework and in the use of analysis
tools. The six sigma project team improvement activities at RS follow a formal, well-defined problem-
ip
t
solving process consistent with six sigma’s DMAIC methodology (Pyzdek and Keller 2009).
The structured problem-solving process is central to, and provides the framework for, the structure
cr
and organization of project documentation at RS. That is, the standard project documentation that each
improvement team must complete incorporates the structured problem-solving framework. Specifically,
us
six sigma project team leaders are required to submit detailed initiation and final project reports. Teams
with projects of longer duration complete interim reports as well. All of these standard reports—together
an
with other relevant documentation, such as project notes, cause-and-effect diagrams, experimental
designs, etc.—are archived in a central database. In this way, the database serves as a repository of
M
documented learnings and as a resource and reference for future project teams.
RS granted us full access to their archival database of 212 six sigma projects closed prior to
d
December 1, 2008. We randomly selected a pilot sample of 12 of these projects to calibrate our coding.
te
Thirty-nine projects were also removed because they were conducted by a single person rather than a
Ac
ce
p
team, as were 2 projects with a headquarters designation and 7 that could not be clearly classified into one
of the five project types. Our final research sample is thus 152 projects.
With respect to the coding of the relevant study variables, we utilized a three-step process (Lapre et
al. 2000; Mukherjee et al. 1998). During the first step, each author conducted an individual, thorough
review of all documentation associated with each six sigma project and then entered his/her
corresponding coding assessments into separate databases.
The second step in our coding process involved reviewing the authors’ coding assessments for each
six sigma project, one-by-one. Naturally, there were some projects for which the authors’ coding
assessments differed for a particular study variable. In the event this occurred, the authors jointly
reviewed the six sigma project documentation associated with the project until complete agreement on the
proper coding of the variable of interest was reached. The results of this consensus coding were recorded
18
Page 21 of 55
into another database, which we utilized for subsequent analysis. We should note that our interrater
agreement of 87.3 percent for the study variables—calculated using the percentage agreement method—
indicates excellent overall agreement (Boyer and Verma 2000).
ip
t
The third and final step in our coding process provides additional confirmation of the content validity
of the study variables. This step involved conducting several interviews with the RS Director of Six
cr
Sigma, as well as presenting, during a site visit, preliminary results to the RS Director of Six Sigma and
master black belts from multiple business groups and functional areas. Throughout these interviews and
us
the site visit, all involved readily agreed with our coding scheme and the resulting descriptive statistics
associated with the study variables. Such concurrence offers additional evidence of the quality of our
an
coding process and the validity of our data.
Measurement Development
3.2.1
The Dependent Variable: Six Sigma Project Success
M
3.2
Our dependent variable is whether or not a six sigma project was determined to be successful by RS
d
during the project closure process. According to the Director of Six Sigma, even at the outset, RS
te
intended to conduct projects that could potentially result in significant knowledge gains, despite the fact
Ac
ce
p
that such projects might be difficult and quite large in scope.
RS designates a project as successful if it achieves a measureable, positive impact in line with the
project’s objectives. We note that some six sigma projects at RS have outcomes that are challenging to
quantify in financial terms, such as improving employee morale, but which are nonetheless measurable.
Such projects are commonly found in six sigma implementations in other organizations (see, e.g., Choo
2014; Choo et al. 2007a; Schroeder et al. 2008; Zu et al. 2008).
During the project closure process, RS relies on input and expertise from a variety of different sources
to rigorously determine whether or not a six sigma project is indeed successful. As a first step, the team
leader reviews the project’s outcome(s) in order to provide a recommendation regarding its closure status
in the draft of the final project report.
19
Page 22 of 55
The team leader’s manager and the master black belt associated with the project then review this draft
report, and suggest revisions on an as-needed basis. At the same time, every draft report is reviewed by a
member of the RS finance group to identify whether or not it contains cost and benefit estimates, and if
ip
t
so, to confirm their accuracy. Importantly, to ensure an independent evaluation, the finance
representative conducting the review is not a member of the six sigma project team.
cr
Finally, the Director of Six Sigma at RS confirms the accuracy of all final project reports. According
to the Director, the project closure process described above is often quite iterative, and involves much
us
thought, discussion, and rigor regarding whether a project is ultimately deemed to be successful.
To provide criterion validity for the dependent variable, we independently assessed whether or not the
an
project documentation, which was often quite extensive, provided evidence of the project’s positive
impact. Every project determined to be successful by RS in the final project report was similarly coded
M
by us as having a positive impact, thus further validating RS’ designation of project success.
Our dependent variable, Six Sigma Project Success, is coded as a ‘1’ if the project was recorded as
d
successful in the final project report and ‘0’ otherwise. RS’ willingness to tackle difficult problems via
Ac
ce
p
unsuccessful.
te
six sigma projects is apparent in Table 2, as approximately half of the projects were deemed to be
3.2.2
Control Variables
At the time of our study, RS was organized into three main business groups, each of which uses
diverse technologies, offers a variety of products, and serves different markets. We thus control for the
possibility that the three business groups at RS have different likelihoods of six sigma project success
using three dummy variables, BG1, BG2, and BG3 (Adler 1990; Ittner and Larcker 1997; Lapre and van
Wassenhove 2001; Mukherjee et al. 1998; Pisano et al. 2001). Business group affiliation is given in RS’
project reports.
Our analysis also controls for the size of the six sigma project team. Communication and
coordination across multiple members may be difficult at times in large project teams (Cohen and Bailey
20
Page 23 of 55
1997), which could reduce the likelihood of six sigma project success. Six sigma projects with numerous
team members may also signify that a project is large in scope, and thus fairly complex, which is similarly
likely to result in a lower project success rate (Huckman and Staats 2011). We operationalize Team Size
ip
t
by counting the number of unique team members identified in each project’s initiation report, interim
status reports (when available), and final report.
cr
Finally, we control for the effect of the use of advanced statistical tools—analysis of variance
(ANOVA), regression, reliability/failure analysis, and DOE—on the likelihood of project success. Using
us
such advanced analysis tools within the six sigma methodology should greatly facilitate root cause
analysis and problem-solving, and ultimately the project team’s ability to meet its goals (Choo et al.
an
2007b; Easton and Rosenzweig 2012; Huber 1991; Linderman et al. 2003, 2006; Mukherjee et al. 1998;
Zu et al. 2008).
M
We reviewed each project’s full set of documentation to determine whether or not advanced statistical
tools were used. Specifically, we coded the UsesStatTools dummy variable as a ‘1’ if the documentation
d
indicated that the six sigma project team utilized one or more of the advanced statistical tools and ‘0’
te
otherwise (see Table 2 for descriptive statistics associated with the control variables). Note that all
3.2.3
Ac
ce
p
projects used one or more basic statistical tools.
Independent Variables
The business groups at RS are largely autonomous as discussed above. Thus, our organizational
experience variable focuses on experience within each business group. Specifically, it is based on the
number of six sigma projects within the business group that are initiated prior to the focal project’s
initiation date (Agrawal and Muthulingam 2015; Choo 2014; Easton and Rosenzweig 2012; Lapre and
van Wassenhove 2001; Lapre et al. 2000). Both the project initiation date and business group affiliation
are given in RS’ project reports.
At RS, projects typically take months to be completed, during which time the business group can gain
considerable experience. Thus, basing the organizational experience measure on prior completed projects
would severely underrepresent the organization’s experience at the time of a project’s initiation. It is for
21
Page 24 of 55
this reason that our organizational experience measure is based on the number of prior six sigma projects
initiated.
The business groups at RS vary substantially in size, which also has implications for a suitable
ip
t
organizational experience measure. For example, if a large business group and a small business group
have both implemented the same number of projects, the scope of their six sigma deployments, and thus
cr
the maturity of the two six sigma systems, will be quite different. This means that the organizational
experience measure should be adjusted for business group size.
us
At RS, the number of employees in each business group was relatively constant over the duration of
the study. This allows us to use the number of employees in each business group at the end of 2008—the
an
time of our data collection—to adjust for size differences in the business groups. Thus, the organizational
M
BG ( k )
/ NEBG ( k ) where BG ( k ) indicates
experience measure corresponding to project k is OEk  NPk
which business group initiated project k , NPkBG ( k ) is number of prior projects initiated in that business
d
group, and NEBG ( k ) is the business group’s number of employees (in 1,000s) at the end of 2008.
te
Team leader familiarity is the first of our four measures relating to team leader social capital. Team
Ac
ce
p
leader familiarity for a project is the average number of times that the team leader has worked with each
of the team members on prior improvement projects (Moore and Lapre 2015; Staats 2012). Thus, team
leader familiarity for the leader of project k is TLFk 
F
kj
/ (nk  1) , where Fkj is the number of times
j
that the leader of project k has previously worked with team member j ( j  1, , nk  1 ), and nk is the
number of people on team k . We computed Fkj from the list of team members for each project, which
we culled from each project’s initiation report, interim status reports (when available), and final report.
As discussed in subsection 2.2.2, we take a network-based view to capture the idea of team leader
connectedness. We begin by conceptualizing a network of connections between a project’s team leader
and everyone within his/her business group that has served on a six sigma team prior to the initiation of
22
Page 25 of 55
the focal project. These individuals correspond to the nodes of the network. Two nodes in this network
are connected if the corresponding two individuals have previously worked together on the same six
sigma team within the business group. As with team leader familiarity, we computed whether or not two
ip
t
individuals have previously worked together from the list of team members for each project.
We note that we limit the network to the business group because, as indicated above, the business
cr
groups at RS are largely autonomous. Because of this autonomy, it is the individuals within the business
group that are in a position to provide information, resources, and other forms of social capital most likely
us
to be important for the team.
Since the network of team leader connections corresponding to a project is determined at the initiation
an
of the project, the networks corresponding to each project differ. There are two reasons for the
differences between the networks. First, the projects may be in different business groups, so the network
M
nodes differ because they involve a different set of individuals. Second, the number of people that have
worked on prior projects increases as more projects are undertaken within a business group, so the size of
d
the network within a business group grows over time.
te
Based on the network of team leader connections at the initiation of a project, we capture the concept
Ac
ce
p
of a team leader’s connectedness using the idea of network centrality. We now provide an overview of
our network centrality-based measures. We leave the technical details of the development of each of
these connectedness measures to an Appendix.
Our first team leader connectedness measure, Bonacich Connectedness, is based on Bonacich
centrality (Bonacich 1987; Wasserman and Faust 1994). Bonacich centrality captures the idea that
connections to individuals who are, in turn, more highly connected are more important than connections
to individuals who are isolated or have few connections. Bonacich centrality is a family of measures
specified by a parameter that controls how much the connectedness of direct connections is weighted.
For positive values of this parameter, Bonacich centrality ranges in its weighting of connections from
degree centrality, which gives equal weight to each direct connection, to eigenvector, which gives
maximum weight to the “connectedness” of the direct connections centrality (Bonacich 1987; Wasserman
23
Page 26 of 55
and Faust 1994). We choose the value of the parameter so that the weighting given to “connectedness” is
“mid-way” between the equal weighting given by degree centrality and the maximal weighting given by
eigenvector centrality (see the Appendix for details).
ip
t
Our second team leader connectedness measure, Closeness Connectedness, is based on closeness
centrality. Closeness centrality measures the centrality of a node by how nearby the other nodes of the
cr
network are (Bavelas 1950; Sabidussi 1966; Wasserman and Faust 1994). Specifically, the closeness
centrality measure for a particular network node is calculated by first finding the average of the minimum
us
path lengths between that node and every other node in the network. The reciprocal of this average is
then taken so that a small average path length corresponds to a high value of the closeness centrality
an
measure.
Our final team leader connectedness measure, Betweenness Connectedness, is based on betweenness
M
centrality. Betweenness centrality is best motivated by thinking about information flows between all
other pairs of nodes in the network. From this point of view, nodes that are centrally located (and
d
therefore have high betweenness centrality) are ones that are on the path of information flowing between
te
many of the pairs of other nodes. To compute the betweenness centrality for a node, the shortest paths
Ac
ce
p
between all pairs of other nodes are first determined. Betweenness centrality, then, is the number of these
shortest paths that go through the node (Carnovale and Yeniyurt 2014; Freeman 1977; Mardsen 2002;
Wasserman and Faust 1994).
The fact that the network of team leader connections is different for each project raises issues with
respect to the comparability of the team leaders’ centrality measures across networks of different sizes
and connections. We address this issue of comparability by focusing on the rank, and corresponding
percentile, of the team leader’s centrality in the distribution of centrality measures for all nodes in the
corresponding network. For example, a statement like “team leaders A and B are both in the top 5% of
their networks in terms of centrality” is meaningful for networks of different sizes and/or with entirely
different members. The exact procedure we use to address this issue is described in the Appendix.
24
Page 27 of 55
In summary, our SNA-based team leader connectedness measures (Bonacich Connectedness,
Closeness Connectedness, and Betweenness Connectedness) are based on the team leader’s relative
position in the distribution of the values of the centrality measures for the nodes of the network defined by
ip
t
all individuals that have served on six sigma teams prior to the initiation of the current project. Because
these measures are derived from percentiles, they are comparable across networks of different sizes, and
cr
thus comparable across projects, and they capture the team leader’s comparative degree of
“connectedness” for each kind of network centrality measure.
us
Our final independent variable is Team Leader Project Type Experience. As we discussed in
subsection 2.3, RS’ six sigma projects naturally fall into five key types. The type of each project was
an
coded by reviewing the project reports and supporting documentation. The Team Leader Project Type
Experience variable is based on the number of prior six sigma projects the team leader has previously led
3.2.4
d
independent variables are shown in Table 2.
M
of the same type as the current project. The descriptive statistics for this variable and the other
Competing Models
te
The dimensions of team leader social capital (i.e., team leader familiarity, Bonacich connectedness,
Ac
ce
p
closeness connectedness, and betweenness connectedness) that we examine represent different, yet
related, views of the importance of relationships to the ability of the team leader to drive improvement
project success. For example, as we discuss above, in the case of team leader familiarity, the team
leader’s social capital results from the team leader having personally worked with team members on a
previous team. Because of the team leader’s firsthand knowledge of his/her team members, the leader is
better able to perform the roles of team leadership such as facilitation of the use of the problem solving
process, assignment of tasks, etc.
In the case of the team leader connectedness measures, the mechanisms for impact are likely similar
to those for familiarity (better facilitation of the problem solving process, better assignment of tasks, etc.).
However, in the case of connectedness, the ability of the team leader to perform these same roles derives
not necessarily from prior experience working with his/her current team members, but rather from the
25
Page 28 of 55
knowledge that the team leader’s experience and connections provide about working with people like
his/her current team members. Thus, team leader familiarity and team leader connectedness represent
related, but different, concepts.
ip
t
Furthermore, with regards to the team leader connectedness dimensions, prior research has shown that
Bonacich centrality, closeness centrality, and betweenness centrality are typically correlated (Valente et
cr
al. 2008). Again, these three measures represent different, but related, approaches to the same concept:
measurement of the centrality of an individual in a network. As discussed in Valente et al. (2008), since
us
these measures are addressing the same general idea (centrality), they should be correlated. However, at
the same time, they should not be too highly correlated or they would not represent useful and somewhat
an
different perspectives on centrality. The empirical evidence developed in Valente et al. (2008) shows that
these measures are indeed correlated, but not too correlated in real-world networks.
M
Because all of the dimensions we use to capture aspects of team leader social capital are related, we
examine models that include only one of these dimensions at a time. This methodological approach
d
allows us to understand and compare the relationship between each of these dimensions and the
te
likelihood of improvement project success, while avoiding problems, such as multicollinearity, due to
Ac
ce
p
their relatedness. To develop evidence concerning whether or not there are statistically significant
differences between these dimensions in terms of their relationship to improvement project success, we
analyze hypotheses 1a to 1d as competing hypotheses and use statistical hypothesis tests developed for
that context as described in subsection 4.1 below.
4. Results and Discussion
The correlation matrix of the study variables is shown in Table 3. Based on prior related studies, we
expect the relationship between several of our independent variables (Team Size, Team Leader
Familiarity, Team Leader Project Type Experience, and Organizational Experience) and Six Sigma
Project Success to have slopes that diminish in magnitude as the variable increases (Banker et al. 2001;
Epple et al. 1991; Hackman 2002; Huckman et al. 2009; Reagans et al. 2005). We thus utilize a square
26
Page 29 of 55
root transformation of these variables, which also improves their symmetry and normality (Anscombe
1948; Box and Cox 1964; Mosteller and Tukey 1977).
Competing Model Results
ip
t
4.1
We test our hypotheses using logistic regression analysis within the R statistical software (R
cr
Development Core Team 2014). Panels A through D in Table 4 provide the results associated with our
four competing models. These models differ depending on which of the four team leader social capital
us
measures is included in the model.
an
Within each of the four panels in Table 4, we provide the logistic regression results for three different
models. These models are hierarchical in nature, and the base model, Model 1, simply examines the
M
effects of the control variables on Six Sigma Project Success. Model 2 adds the main effects to the base
effects as well.
d
model. In the final model (Model 3) in each panel in Table 4, we include the hypothesized interaction
te
We use likelihood ratio tests to compare the fit of Model 2 to Model 1 and Model 3 to Model 2 at the
Ac
ce
p
bottom of each panel in Table 4. The p-values associated with these likelihood ratio tests are all highly
significant (p = .000), indicating that Model 3 gives the best model fit in all four panels. We thus focus
our description of the results and ensuing discussion on the Model 3 logistic regressions in Panels A
through D.
We examine the four Model 3 logistic regression results to determine which competing model has the
best overall fit with the data. In doing so, we first compare the corrected Akaike Information Criterion
(AICc) goodness-of-fit values for each panel. We note here that smaller values of the AICc indicate
better model fit (Akaike 1974).
The AICc value for the model using the Bonacich Connectedness variable to operationalize team
leader social capital (AICc = 177.97; see Panel B) indicates the best fit, although this value is close to the
AICc for the model using Closeness Connectedness (AICc = 178.24; see Panel C). The Model 3 AICc
27
Page 30 of 55
values for both of these models are superior to the AICc values for the Team Leader Familiarity (AICc =
184.31) and Betweenness Connectedness (AICc = 186.26) models in Panels A and D, respectively. We
defer further discussion of which of the four competing models is best-fitting to later in this subsection.
ip
t
The results in Table 4 provide statistical evidence that several of the control variables have an impact.
Specifically, business group affiliation appears to be related, to some extent, to the likelihood of six sigma
cr
project success. Recall that each business group at RS deals with different technologies and markets.
This leads the business groups to have relatively different emphases on functions such as engineering and
us
marketing. For example, relative to BG1 and BG2, BG3 has more of an engineering-based focus given
the nature of its products, while BG1 and BG2 are more marketing-focused. Six sigma tends to be more
an
readily adopted in engineering-focused organizations because of its emphasis on disciplined and datadriven problem-solving. As a result, this may explain why BG3 appears to have a higher likelihood of
M
project success.
In addition, the control variable UsesStatTools has a positive and statistically significant relationship
d
with Six Sigma Project Success in all of the Model 3 results in Panels A through D (p ≤ .01). The use of
te
advanced statistical tools is likely associated with more systematic experimentation and more effective
success.
Ac
ce
p
root cause analysis, which in turn results in improved learning and an increased likelihood of project
Recall that hypotheses 1a through 1d relate increases in team leader social capital to increases in the
likelihood of Six Sigma Project Success. The Team Leader Familiarity coefficient in Model 3 of Table 4,
Panel A is not statistically significant (β = 0.348; p ≥ .10). We therefore find no support for hypothesis
1a. We do, however, find strong support for hypothesis 1b, as the coefficient of Bonacich Connectedness
in Model 3 of Panel B is significant in the expected direction (β = 0.631; p ≤ .01). Panel C provides the
Model 3 results using Closeness Connectedness to measure team leader social capital. Given the sign and
significance of the Closeness Connectedness coefficient (β = 0.650; p ≤ .01), hypothesis 1c is likewise
strongly supported. The observed relationship between Betweenness Connectedness and Six Sigma
Project Success (β = 0.112; p ≥ .10) in Model 3 of Panel D does not provide support for hypothesis 1d.
28
Page 31 of 55
Hypothesis 2 posits that a six sigma project is likely to be more successful when the team leader has
prior experience leading the same type of project. The direction of the sign and the significance of the
Team Leader Project Type Experience coefficients in all four Model 3 results indicate that this hypothesis
ip
t
is supported. In three of the four competing models we observe strong significance for this coefficient,
with significance at the 0.10 level for the remaining model (the model based on Closeness Connectedness
cr
shown in Panel C).
We found no support for hypothesis 3, irrespective of which measure of team leader social capital is
us
included in the model. This suggests that the positive effect of team leader social capital on the likelihood
of Six Sigma Project Success is not influenced by the overall experience of the organization.
an
Our results do, however, indicate a consistently significant negative interaction between Team Leader
Project Type Experience and Organizational Experience (p ≤ .01) in all fours panels in Table 4, which
M
provides support for hypothesis 4. Specifically, when organizational experience is high, a team leader’s
prior experience leading the same type of six sigma project does not necessarily increase the likelihood of
d
project success. However, when a team leader has minimal prior experience leading the same type of
te
project, greater experience at the organizational level certainly improves the odds of project success.
Ac
ce
p
We provide a summary of the degree of support for our hypotheses in Table 5. When we examine the
strength of the support for hypotheses 1a to 1d in Table 5 together with the corresponding AICc values
from Table 4, it appears that the Bonacich Connectedness and Closeness Connectedness models are
substantially better than the Team Leader Familiarity and Betweenness Connectedness models. To
examine this more formally, we test the competing models using the Clark (2007) procedure for
comparing the fit of non-nested models. Specifically, we test whether or not there is statistical evidence
in favor of one of the models for all six pairs that can be formed from the four competing models we
consider. In each of these tests, the null hypothesis is that the true model is equally distant (as measured
by the Kullbach-Leibler distance) from both of the models in the pair of competing models considered.
The alternative hypothesis is that one of the two models is closer to the true model.
29
Page 32 of 55
Table 6 gives the p-values for the Clark (2007) test for each pair, with the model favored by the test
indicated in italics and underlined. Table 6 shows that there is strong statistical evidence that the model
based on the Bonacich Connectedness measure is superior to the models based on Team Leader
ip
t
Familiarity (p ≤ .01) and Betweenness Connectedness (p ≤ .01). There is also strong statistical evidence
that the model based on Closeness Connectedness is superior to the model based on Betweenness
cr
Connectedness (p ≤ .01). Finally, there is no evidence that there is a difference between the Bonacich
Connectedness and Closeness Connectedness models. In summary, there is strong statistical evidence
these models are essentially indistinguishable from one another.
us
that the best models are the ones based on Bonacich Connectedness and Closeness Connectedness, but
4.2
an
Discussion
M
Our findings do not support the idea that team leader familiarity increases the likelihood of six sigma
project success. Measures of familiarity reward experience working with the same people. Given team
d
leader familiarity does not seem to matter in our research context, it is natural to consider broader social
te
capital-related measures, such as those associated with team leader connectedness that instead capture a
Ac
ce
p
team leader’s experience working with different people.
Using an approach based on social networks provides an intuitive way to capture a team leader’s
boundary-spanning social capital. We began this broader framing of team leader social capital by
considering Bonacich connectedness. Bonacich connectedness is consistent with our theoretical
development in that it captures the idea that a team leader’s connections may vary in importance. Team
leaders with high Bonacich connectedness are connected to people who are, in turn, highly connected.
Such well-connected connections have contacts and experience that may aid a team leader in recruiting
suitable team members, and help the leader to facilitate the structured problem-solving process. Our
Panel B results in Table 4 are consistent with these ideas.
We also examined how closeness and betweenness connectedness might be associated with the
likelihood of six sigma project success. Closeness and betweenness provide related, but different,
30
Page 33 of 55
perspectives on the concept of network centrality that we use to capture a team leader’s social capital.
The logistic regression and Clark (2007) test results in Tables 4 and 6, respectively, provide evidence that
both Bonacich and closeness connectedness have a stronger relationship with project success than does
ip
t
betweenness connectedness.
Our closeness connectedness results suggest that the number of steps that it takes team leaders to
cr
reach other individuals may be just as important as the number and connectedness of a leader’s direct
connections as captured by Bonacich connectedness. Such a result may be due to information distortion,
us
which is less likely to occur along paths with fewer intervening individuals (Bellamy et al. 2014; Hansen
2002; Peng et al. 2013). We suspect this might be the case because six sigma team leaders with a high
an
closeness value are well-positioned to know “what is out there” with respect to “good” team members,
“good” data, and other relevant and perhaps undocumented information and expertise.
M
What is striking about our results is that no matter the level of organizational experience, a team
leader’s network, as captured by Bonacich and closeness connectedness, matters. Even if an organization
d
is fairly far along in its six sigma deployment, these two social capital-related dimensions of team leader
te
experience continue to have an important effect on the likelihood of six sigma project success.
Ac
ce
p
In contrast, the positive impact of team leader project type experience on six sigma project success is
indeed significantly smaller when organizational experience is high. This result suggests that, in a mature
six sigma deployment, the cumulative body of documented learnings in the database of archived projects
may well substitute for a team leader’s own prior experience leading a particular project type. For
example, a team leader can search the database to gain a better understanding of the metrics, tools, and
tasks that are characteristic of a particular type of project, and then effectively apply that knowledge in the
context of his/her current project of the same type. This suggests that the team leader’s technical
expertise derived from project type experience is perhaps substituted by the technical knowledge captured
by the database of archived projects.
Our findings certainly support the idea that it is useful to disentangle various facets of experience, and
to examine how such facets are interrelated (Argote and Miron-Spektor 2011; Lapre and Nembhard
31
Page 34 of 55
2011). For example, Easton and Rosenzweig (2012) found an interaction between organizational
experience and a measure of team leader experience, with only moderate support for this interaction in
some of the models they examined. However, their measure of team leader experience simply tallies the
ip
t
total number of prior projects led by the team leader, and does not account for the distinct effects of team
leader social capital (a social aspect) or project type experience (a technical aspect). Because this prior
cr
study does not separate team leader experience into more fine-grained dimensions as we do in the present
study, this might explain the varying results.
us
Taken together, the interaction results in the present study indicate that the benefits we associate with
certain aspects of team leader experience, such as experience leading a particular project type, can instead
an
be obtained by way of extensive organizational experience, while others, such as team leader social
capital as measured by Bonacich and closeness connectedness, cannot. Perhaps these results are due, in
M
part, to the relative ease with which project type specific learnings can be codified and transferred to
future projects via the database of archived projects (Anand et al. 2010). Alternatively, some project
d
learnings may be less explicit and thus undocumented. In this case, we suspect that a team leader’s access
te
to others either by direct connections to well-connected people or via short paths is instrumental to
4.3
Ac
ce
p
facilitating the transfer of such knowledge (Hansen 2002).
Limitations and Future Research
Our study is subject to several limitations. One limitation is that we derive our results from data
collected from a single firm. We should note that our firm-specific data is rich with detail, and at the
same time, provides a large enough sample size to allow for statistical testing of our hypotheses
(DeHoratius and Raman 2008; Huckman et al. 2009; van Donselaar et al. 2010). Nonetheless, analyzing
data from a single firm impacts the extent to which we can generalize our findings beyond RS. In
addition, despite an extensive, six-year deployment of six sigma at RS throughout its business groups, it is
possible our results might differ for firms with even more mature six sigma systems (Nair et al. 2011).
Our data also does not capture the specific characteristics of a team leader’s experience working with
others on six sigma projects (e.g., time spent collaborating, nature of collaboration, etc.), nor does it
32
Page 35 of 55
account for his/her experience working with others above and beyond the six sigma system. Furthermore,
team leader familiarity, as currently defined, does not take into account variation in the number of times
that the team leader has previously worked with each team member. A high level of team leader
ip
t
familiarity due to the team leader working with one or two people many times may be different from a
similar level of familiarity due to the team leader working with all team members one or a few times. In
cr
addition, our team leader connectedness measures are based on networks that capture whether or not a
team leader has worked with others, and does not take into account the number of times they have worked
us
together. Such factors may influence the relationship between a team leader’s social capital and project
success, and are thus subjects for future study. Finally, it would be interesting for future research to
an
consider whether prior improvement project success or failure has a differential effect in terms of learning
M
and impact on experience.
5. Conclusion
Our study contributes to the growing body of empirical research that examines the ways in which
d
various dimensions of experience influence the operational performance of teams. We use six years of
te
rich, highly-detailed, improvement project data to disentangle the relationship between several aspects of
Ac
ce
p
team leader experience and six sigma project success.
The vast majority of the operations management literature that relates multiple dimensions of
experience to the operational performance of teams focuses on work teams. Our research context
provides a useful setting for advancing this literature because it involves improvement teams rather than
work teams. Prior research suggests that quality improvement teams and work teams differ in important
ways. As discussed previously, one of these differences involves the expected relationship between team
familiarity and performance.
Rather than a six sigma project benefiting from a team leader’s experience working with the same
individuals on prior projects (team leader familiarity), what seems to matter in our research on
improvement teams is working with a variety of people. That is, we find that a team leader’s social
capital resulting from experience working with different individuals on prior six sigma projects—which
33
Page 36 of 55
we capture by way of our SNA-based team leader connectedness measures—has an important impact on
the odds of project success. In particular, we find that the number and connectedness of a team leader’s
direct connections (Bonacich connectedness) as well as the number of steps it takes to reach other
ip
t
individuals (closeness connectedness) are important.
More broadly, by incorporating a social network viewpoint into our research, we were able to gain
cr
deep insights into aspects of team leader experience. This perspective led to our development of three
network-based dimensions capturing novel social capital-related aspects of team leader experience. To
us
the best of our knowledge, this is the first study that takes a network-based approach in examining the
relationship between various dimensions of experience and operational performance in improvement
an
teams.
In terms of managerial implications, our results imply that, for a team leader, “what you know”
M
(project type experience) and “who you know” (connectedness) matter to project success. As discussed
previously, team leaders often determine the focus of a six sigma project. Thus, with regards to “what
d
you know,” one way management can likely increase an organization’s overall odds of project success is
te
to simply ask each team leader to consider focusing their improvement efforts on problems of a particular
Ac
ce
p
type, especially during the early stages of the six sigma deployment. However, as the six sigma system
matures and organizational experience accumulates, our results suggest that management can essentially
ease up on this request of team leaders without severely impacting the likelihood of project success.
Our results also indicate that, regardless of the level of organizational experience, management should
expect improvement projects led by highly connected team leaders in terms of Bonacich or closeness
connectedness to have a higher likelihood of success. As a result, in the event it becomes necessary to
tackle an especially difficult project early on during the six sigma deployment, when team leaders and the
overall organization are relatively inexperienced, management should attempt to steer that improvement
project to a more connected team leader.
Importantly, in our quest to understand how dimensions of team leader experience and organizational
experience interact to affect the likelihood of six sigma project success, we contribute to the nascent
34
Page 37 of 55
literature addressing Argote and Miron-Spektor’s (2011) call to investigate when different types of
experience serve as complements or substitutes to one another. In this way, our research not only offers a
new network-based perspective of experience, but also provides insights regarding how dimensions of
ip
t
team leader experience and organizational experience collectively impact the operational performance of
cr
improvement teams.
Acknowledgements
us
We appreciate the time, effort, and access the RS Director of Six Sigma provided us in facilitating
an
this research project. Seb Heese also provided helpful feedback on the paper.
Appendix
M
In this Appendix, we provide technical details concerning the three SNA-based measures of team
leader connectedness. These three measures are based on the idea of network centrality. As discussed in
d
subsection 2.2.2, for each project, we begin by considering the network defined by the project’s team
te
leader and everyone within his/her business group that has previously participated on a six sigma team
prior to the initiation of the focal project. These individuals represent the nodes of the network, and two
Ac
ce
p
nodes are connected if the corresponding individuals have worked together on a prior improvement team.
The network corresponding to project k can be specified using an adjacency matrix. The adjacency
(k )
matrix A
for project k is a square symmetric matrix with each row (and column) corresponding to an
individual. Two individuals are “connected” if they have worked together on a prior improvement
project, and this is captured in the adjacency matrix by using a 1 when there is a connection and a 0
otherwise. More precisely, Aij( k )  A(jik )  1 , for i  j , if individuals i and j have worked together on a
six sigma team prior to the initiation of project k , and 0 otherwise. As discussed in subsection 3.2.3,
these networks differ for each project.
The network centrality metrics that underlie our three SNA-based team leader connectedness
measures are computed from the corresponding adjacency matrices using the igraph package (Csardi and
35
Page 38 of 55
Nepusz 2006) of the R statistical computing environment (R Development Core Team 2014). Our first
SNA-based measure, Bonacich Connectedness, is based on Bonacich centrality (Bonacich 1987;
Wasserman and Faust 1994), which takes into account the “connectedness” of an individual’s
ip
t
connections. Bonacich centrality, as developed in Bonacich (1987), depends on a parameter  that
controls the weighting of the “connectedness” of a node’s direct connections. Positive values of  give
cr
increased weight to connections that are more connected, negative values of  give increased weight to
us
connections that are more isolated, and a 0 value of  corresponds to degree centrality, which simply
counts the number of connections a node has and thus weights all connections equally. In our research
an
context, we are only interested in   0 since we are only interested in giving increased weight to
connections that are more highly connected.
M
The upper bound for valid positive values of  for the network corresponding to project k depends
(k )
(Bonacich 1987). As 
d
(k )
on the reciprocal of the largest eigenvalue max
of the adjacency matrix A
te
(k )
approaches max
, the Bonacich centrality measure approaches a limit giving the highest possible weight
to the connectedness of the direct connections. This limit corresponds to a centrality measure called
Ac
ce
p
eigenvector centrality (Bonacich 1972; Wasserman and Faust 1994). Thus, over the range of allowable
(k )
positive values for  (i.e., for 0    max
), Bonacich centrality ranges from degree centrality, which
gives equal weight to each direct connection, to eigenvector centrality, which gives maximum weight to
the direct connections’ connectedness.
Because the networks corresponding to each project are different, have different adjacency matrices,
(k )
), it is natural to re-parameterize
and, therefore, have different allowable ranges for  (i.e., 0    max
(k )
 in terms of the fraction of the way  is between 0 and max
. Thus, we parameterize the family of
(k )
Bonacich centrality measures using    / max
, so  takes on values in the range [0, 1). We discuss
36
Page 39 of 55
how we choose the value of  that we use in this study below after we introduce our other centrality
measures.
Our second team leader connectedness measure, Closeness Connectedness, is based on closeness
ip
t
centrality (Wasserman and Faust 1994). To calculate closeness centrality for a node, first, the minimum
path length from that node to every other node of the network is determined. Next, these minimum path
cr
lengths are averaged. Finally, because small average path lengths to the other nodes correspond to high
us
centrality, the reciprocal is taken. Thus, closeness centrality for node i of the network corresponding to
project k is
CCi( k )  ( n ( k )  1) /  Dij( k ) ,
an
j i
project k , and n
(k )
M
where Dij( k ) is the minimum path length from node i to node j in the network corresponding to
is the number of nodes in that network.
d
The third SNA-based team leader connectedness measure, Betweenness Connectedness, is based on
te
betweenness centrality (Wasserman and Faust 1994). Calculating the betweenness centrality for a node
begins by determining the shortest paths between all pairs of other nodes. The basic idea behind a node’s
Ac
ce
p
betweenness centrality is that it is the number of these shortest paths that go through the node (Freeman
1977; Wasserman and Faust 1994). As a technical detail, however, it is possible that there are multiple
tied minimum length paths between two nodes, with only some of these paths going through the focal
node. In that case, only the fraction of the minimum length paths going through the focal node is counted.
Thus, the betweenness centrality for node i in the network for project k is given by
BTWCi( k ) 
 NP
l ,m i
l m
(k )
lm
(i ) / NPlm( k ) ,
where the denominator NPlm( k ) is the number of shortest paths between nodes l and m in network k ,
(k )
(i ) is the number of those shortest paths that go through node i .
and the numerator N lm
37
Page 40 of 55
In this study, we focus on the relationship between project success and the connectedness of the
project’s team leader. Thus, for each kind of centrality measure we use (Bonacich, closeness, and
betweenness centrality), we must extract, for each project k , only the team leader’s network centrality
ip
t
measure from the centrality measures for all of the nodes of that project’s network. However, these team
leader centrality measures, which we refer to as CL( k ) , are not directly comparable across projects because
cr
the projects’ networks vary in both connections and size. To correct for this issue, we simply use the
us
team leader’s relative centrality within the project’s network. That is, we use the rank of the team
leader’s centrality measure in the network corresponding to project k in order to obtain the team leader’s
Specifically, suppose that C
(k )
an
relative connectedness.
 (Ci( k ) ,, Cn((kk)) ) is the vector of centrality measures for all of the
M
nodes in the network corresponding to project k . One of these centrality measures will correspond to the
team leader’s centrality CL( k ) , and the others will be the centrality measures of all of the other people who
te
d
have previously worked on improvement teams prior to the initiation of project k . To determine the
team leader’s relative centrality, we first determine the rank of the team leader’s centrality CL( k ) among the
(k )
Ac
ce
p
elements of the corresponding vector C
. Next, we compute the corresponding percentile. We then
convert the percentile to a normal score (i.e., the value for a standard normal distribution that corresponds
to the percentile) and use these normal scores as the measure of team leader connectedness. We use these
normal scores rather than the ranks or percentiles because they have better statistical properties when used
as independent variables in regression. Approaches based on normal scores are commonly used in nonparametric statistical methods (Lehmann and D’Abrera 2006).
Taken together, determining the connectedness of a specific project’s team leader, for each of the
three centrality measures we consider, involves the following steps. First, the network consisting of the
team leader together with all of the individuals within his/her business group who have worked on six
sigma projects prior to the initiation of the focal project is constructed. Next, the three centrality metrics
38
Page 41 of 55
are computed for each node of the network. Each centrality metric is then (separately) converted to ranks
and then normal scores. Thus, we will have three normal scores (one corresponding to each centrality
metric) for each node. Finally, we extract the normal scores for the node that corresponds to the team
betweenness connectedness) based on the three network centrality measures.
ip
t
leader. These normal scores are the values of team leader connectedness (Bonacich, closeness, and
cr
We now return to the issue of choosing the value of  in the computation of the Bonacich centrality
us
measure we use in our study. We seek a value of  that is a compromise between the extremes of equal
weighting of connections (degree centrality,   0 ) and the weighting that gives the maximum possible
an
weight to more connected connections (eigenvector centrality,   1 ). If we could find a value of 
which is, in a sense, halfway between these two extremes, this would represent an intuitively appealing
M
compromise. We operationalize this idea of being halfway between degree centrality and eigenvector
centrality by choosing  so that the correlations between the team leader connectedness measure based
te
d
on Bonacich centrality and those based on degree centrality and eigenvector centrality are equal.
References
Ac
ce
p
Adler, P. 1990. Shared learning. Management Science 36(8) 938-957.
Agrawal, A., S. Muthulingam. 2015. Does organizational forgetting affect vendor quality performance?
An empirical investigation. Manufacturing & Service Operations Management, forthcoming.
Akaike, H. 1974. A new look at the statistical model identification. IEEE Transactions on Automatic
Control 19(6) 716-723.
Anand, G., P. Ward, M. Tatikonda. 2010. Role of explicit and tacit knowledge in Six Sigma projects: An
empirical examination of differential project success. Journal of Operations Management 28(4) 303315.
Anscombe, F. 1948. The transformation of Poisson, binomial, and negative-binomial data. Biometrika
35(3-4) 246-254.
Argote, L., E. Miron-Spektor. 2011. Organizational learning: From experience to knowledge.
Organization Science 22(5) 1123-1137.
Argote, L., S. Beckman, D. Epple. 1990. The persistence and transfer of learning in industrial settings.
Management Science 36(2) 140-154.
Balkundi, P., D. Harrison. 2006. Ties, leaders, and time in teams: Strong inference about network
structure's effects on team viability and performance. Academy of Management Journal 49(1) 49-68.
Banker, R., J. Field, K.K. Sinha. 2001. Work-team implementation and trajectories of manufacturing
quality: A longitudinal field study. Manufacturing & Service Operations Management 3(1) 25-42.
Barden, J., W. Mitchell. 2007. Disentangling the influences of leaders’ relational embeddedness on
interorganizational exchange. Academy of Management Journal 50(6) 1440-1461.
39
Page 42 of 55
Ac
ce
p
te
d
M
an
us
cr
ip
t
Bavelas, A. 1950. Communication patterns in task-oriented groups. Journal of the Acoustical Society of
America 22 271-282.
Bellamy, M., S. Ghosh, M. Hora. 2014. The influence of supply network structure on firm innovation.
Journal of Operations Management 32(6) 357-373.
Benner, M., M. Tushman. 2002. Process management and technological innovation: A longitudinal study
of the photography and paint industries. Administrative Science Quarterly 47(4) 676-706.
Benner, M., M. Tushman. 2003. Exploitation, exploration, and process management: The productivity
dilemma revisited. Academy of Management Review 28(2) 238-256.
Bonacich, P. 1972. Factoring and weighting approaches to status scores and clique identification.
Journal of Mathematical Sociology 2(1) 113-120.
Bonacich, P. 1987. Power and centrality: A family of measures. American Journal of Sociology 92(5)
1170-1182.
Borgatti, S., M. Everett, M. 2006. A graph-theoretic perspective on centrality. Social
Networks 28(4) 466–484.
Borgatti, S., P. Foster. 2003. The network paradigm in organizational research: A review and typology.
Journal of Management 29(6) 991-1013.
Borgatti, S., X. Li. 2009. On network analysis in a supply chain context. Journal of Supply Chain
Management 45(2) 5-22.
Box, G., D. Cox. 1964. An analysis of transformations (with discussion). Journal of the Royal Statistical
Society 26(2) 211-252.
Boyer, K., R. Verma. 2000. Multiple raters in survey-based operations management research: A review
and tutorial. Production and Operations Management 9(2) 128-140.
Brass, D., D. Krackhardt. 1999. The social capital of 21st century leaders. In Out-of-the-Box Leadership:
Transforming the 21st Century Army and Other Top Performing Organizations, J. Hunt and R.
Phillips (Eds.), JAI Press, Greenwich, CT, 179-194.
Burt, R. 1992. Structural Holes: The Social Structure of Competition. Harvard University Press,
Cambridge, MA.
Carnovale, S., S. Yeniyurt. 2014. The role of ego networks in manufacturing joint venture formations.
Journal of Supply Chain Management 50(2) 1-17.
Chauvet, V., B. Chollet, G. Soda, I. Hault. 2011. The contribution of network research to managerial
culture and practice. European Journal of Management 29(5) 321-334.
Choo, A. 2014. Defining problems fast and slow: The u-shaped effect of problem definition time on
project duration. Production and Operations Management 23(8) 1462-1479.
Choo, A., K. Linderman, R. G. Schroeder. 2007a. Method and psychological effects on learning
behaviors and knowledge creation in quality improvement teams. Management Science 53(3) 437450.
Choo, A., K. Linderman, R. G. Schroeder. 2007b. Method and context perspectives on learning and
knowledge creation in quality management. Journal of Operations Management 25(4) 918-931.
Clark, J., R. Huckman, B. Staats. 2013. Learning from customers: Individual and organizational effects
in outsourced radiological services. Organization Science 24(5) 1539-1557.
Clark, K. 2007. A simple distribution-free test for nonnested model selection. Political Analysis 15 347363.
Cohen, S., D. Bailey. 1997. What makes teams work: Group effectiveness research from the shop floor to
the executive suite. Journal of Management 23(3) 239-290.
Cropanzano, R., M. Mitchell. 2005. Social exchange theory: An interdisciplinary review. Journal of
Management 31(6) 874-900.
Csardi G, T. Nepusz. 2006. The igraph software package for complex network research. InterJournal,
Complex Systems 1695. http://igraph.org.
Cummings, T. 1978. Self-regulating work groups: A socio-technical synthesis. Academy of Management
3(3) 625-634.
DeHoratius, N., A. Raman. 2008. Inventory record inaccuracy. Management Science 54(4) 627-641.
40
Page 43 of 55
Ac
ce
p
te
d
M
an
us
cr
ip
t
Deming, W. 1994. The New Economics for Industry, Government, Education, 2nd Edition. MIT Center
for Advanced Engineering Study, Cambridge, MA.
Devine, D., L. Clayton, J. Philips, B. Dunford, S. Melner. 1999. Teams in organizations: Prevalence,
characteristics, and effectiveness. Small Group Research 30 678 –711.
Easton, G., E. D. Rosenzweig. 2012. The role of experience in six sigma project success: An empirical
analysis of improvement projects. Journal of Operations Management 30(7-8) 481-493.
Edmondson, A. 1999. Psychological safety and learning behavior in work teams. Administrative Science
Quarterly 44(2) 350-383.
Edmondson, A., I. Nembhard. 2009. Product development and learning in project teams: The challenges
are the benefits. Journal of Product Innovation Management 26(2) 123-138.
Epple, D. L. Argote, R. Devadas. 1991. Organizational learning curves: A method for investigating intraplant transfer of knowledge acquired through learning by doing. Organization Science 2(1) 58-70.
Faraj, S., L. Sproull. 2000. Coordinating expertise in software development teams. Management Science
46(12) 1554-1568.
Faraj, S., A. Yan. 2009. Boundary work in knowledge teams. Journal of Applied Psychology 94(3) 604–
617.
Freeman, L. 1977. A set of measures of centrality based on betweenness. Sociometry 40(1) 35-41.
Freeman, L. 2011. The development of social network analysis – with an emphasis on recent events. In
The SAGE Handbook of Social Network Analysis, Chapter 3, J. Scott and P. Carrington (Eds.), Sage
Publications Ltd., Thousand Oaks, CA, 26-39.
Granovetter, M. 1973. The strength of weak ties. American Journal of Sociology 78(6) 1360-1380.
Granovetter, M. 1985. Economic action and social structure: The problem of embeddedness. American
Journal of Sociology 91(3) 481-510.
Haas, M. 2006. Knowledge gathering, team capabilities, and project performance in challenging work
environments. Management Science 52(8) 1170-1184.
Hackman, J. R. 1987. The design of work teams. In Handbook of Organizational Behavior, J. Lorsch
(Ed.), Prentice Hall, Englewood Cliffs, NJ.
Hackman, J. R. 2002. Leading Teams: Setting the Stage for Great Performances. HBS Press, Boston,
MA.
Hansen, M. 2002. Knowledge networks: Explaining effective knowledge sharing in multiunit companies.
Organization Science 13(3) 232-248.
Huber, G. P. 1991. Organizational learning: The contributing processes and the literatures. Organization
Science 2(1) 88-115.
Huckman, R. S., G. P. Pisano. 2006. The firm specificity of individual performance: Evidence from
cardiac surgery. Management Science 52(4) 473-488.
Huckman, R., B. Staats. 2011. Fluid tasks and fluid teams: The impact of diversity in experience and
team familiarity on team performance. Manufacturing & Service Operations Management, 13(3)
310-328.
Huckman, R. S., B. R. Staats, D. M. Upton. 2009. Team familiarity, role experience, and performance:
Evidence from Indian software services. Management Science 55(1) 85-100.
Ittner, C. D., D. F. Larcker. 1997. The performance effects of process management techniques.
Jacobs, B., M. Swink, K. Linderman. 2015. Performance effects of early and late six sigma adoptions.
Journal of Operations Management, http://dx.doi.org/10.1016/j.jom.2015.01.002.
KC, D., B. Staats. 2012. Accumulating a portfolio of experience: The effect of focal and related
experience on surgeon performance. Manufacturing & Service Operations Management 14(4) 618633.
Ketchen, D., T. Hult. 2007. Bridging organization theory and supply chain management: The case of best
value supply chains. Journal of Operations Management 25(2) 573-580.
Ketchum, L., E. Trist. 1992. All Teams are Not Created Equal: How Employee Empowerment Really
Works. Sage, Newbury Park, CA.
41
Page 44 of 55
Ac
ce
p
te
d
M
an
us
cr
ip
t
Kim, D. H. 1993. The link between individual and organizational learning. Sloan Management Review
35(1) 37-50.
Kim, D. Y. 2014. Understanding supplier structural embeddedness: A social network perspective.
Kim, Y., T. Choi, T. Yan, K. Dooley. 2011. Structural investigation of supply networks: A social
network analysis approach. Journal of Operations Management 29 194-211.
Lapre, M., I. Nembhard. 2011. Inside the organizational learning curve: Understanding the organizational
learning process. Foundations and Trends in Technology, Information and Operations Management
4(1) 1-103.
Lapre, M., L. van Wassenhove. 2001. Creating and transferring knowledge for productivity improvement
in factories. Management Science 47(10) 1311-1325.
Lapre, M., A. S. Mukherjee, L. van Wassenhove. 2000. Behind the learning curve: Linking learning
activities to waste reduction. Management Science 46(5) 597-611.
Leavitt, H. 1951. Some effects of certain communication patterns on group performance. Journal of
Abnormal and Social Psychology 46(1) 38-50.
Lehmann, E., H. D’Abrera. 2006. Nonparametrics: Statistical Methods Based on Ranks. Springer, New
York, NY.
Linderman, K., R. G. Schroeder, A. Choo. 2006. Six sigma: The role of goals in improvement teams.
Linderman, K., R. G. Schroeder, S. Zaheer, A. Choo. 2003. Six sigma: A goal theoretic perspective.
Marsden, P. 2002. Egocentric and sociocentric measures of network centrality. Social Networks 24(4)
407–422.
Mehra, A., A. Dixon, D. Brass, B. Robertson. 2006. The social network ties of group leaders:
Implications for group performance and leader reputation. Organization Science 17(1) 64-79.
Moore, D., M. Lapre. 2015. The simultaneous impact of learning curve-heterogeneity, tightly-coupled
team familiarity, and workload on orthopedic procedure times. Stanford School of Medicine working
paper.
Mosteller F., J. W. Tukey. 1977. Data Analysis and Regression: A Second Course in Statistics. AddisonWesley, Reading, MA.
Muehlfeld, K., P. Sahib, A. Witteloostuijn. 2012. A contextual theory of organizational learning from
failures and successes: A study of acquisition completion in the global newspaper industry, 19812008. Strategic Management Journal 33(8) 938-964.
Mukherjee, A. S., M. Lapre, L. van Wassenhove. 1998. Knowledge driven quality improvement.
Management Science 44(11-2) S35-S49.
Nahapiet, J., S. Ghoshal. 1998. Social capital, intellectual capital, and the organizational advantage.
Academy of Management Review 23(2) 242-266.
Nair, A., M. Malhotra, S. Ahire. 2011. Toward a theory of managing context in Six Sigma processimprovement projects: An action research investigation. Journal of Operations Management 29(5)
529-548.
Oh, H., M. Chung, G. Labianca. 2004. Group social capital and group effectiveness: The role of informal
socializing ties. Academy of Management Journal 47(6) 860–875.
Peng, G., J. Mu, C. Benedetto. 2013. Learning and open source software license choice. Decision
Sciences 44(4) 619-643.
Pisano, G. P., R. Bohmer, A. Edmondson. 2001. Organizational differences in rates of learning: Evidence
from the adoption of minimally invasive cardiac surgery. Management Science 47(6) 752-768.
Pyzdek, T., P. Keller. 2009. The Six Sigma Handbook: A Complete Guide for Green Belts, Black Belts,
and Managers at All Levels. 3rd Edition. McGraw Hill, New York, NY.
R Development Core Team. 2014. R: A Language and Environment for Statistical Computing. R
Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, http://www.Rproject.org.
42
Page 45 of 55
Ac
ce
p
te
d
M
an
us
cr
ip
t
Reagans, R., L. Argote, D. Brooks. 2005. Individual experience and experience working together:
Predicting learning rates from knowing who knows what and knowing how to work together.
Sabidussi, G. 1966. The centrality index of a graph. Psychometrika 31 581-603.
Schilling, M., P. Vidal, R. Ployhart, A. Marangoni. 2003. Learning by doing something else: Variation,
relatedness, and the learning curve. Management Science 49(1) 39-56.
Schroeder, R. G., K. Linderman, C. Liedtke, A. Choo. 2008. Six sigma: Definition and underlying
theory. Journal of Operations Management 26(4) 536-554.
Shafer, S., S. Moeller. 2012. The effects of Six Sigma on corporate performance: An empirical
investigation. Journal of Operations Management 30(7-8) 521-532.
Shafer, S., D. Nembhard, M. Uzumeri. 2001. The effects of worker learning, forgetting, and
heterogeneity on assembly line productivity. Management Science 47(12) 1639-1653.
Shaw, M. 1964. Communication networks. In Advances in Experimental Social Psychology, Volume 1,
L. Berkowitz (Ed.), Academic, New York, 111-147.
Staats, B. 2012. Unpacking team familiarity: The effect of geographic location and hierarchical role.
Production and Operations Management 21(3) 619-635.
Swink, M., B. Jacobs. 2012. Six Sigma adoption: Operating performance impacts and contextual drivers
of success. Journal of Operations Management 30(6) 437-453.
Trist, E. 1981. The sociotechnical perspective: the evolution of sociotechnical systems as a conceptual
framework and as an action research program. In Perspectives on Organization Design and
Behavior, A. Van de Van, W. Joyce (Eds.), Wiley, New York, NY.
Trist, E., K. Bamforth. 1951. Some social and psychological consequences of the longwall method of
coal-getting. Human Relations 4(1) 3-38.
Tuckman, B. W. 1965. Developmental sequence in small groups. Psychological Bulletin 63(6) 384-399.
Tuckman, B. W. 1977. Stages of small-group development revisited. Group & Organizational Studies
2(4) 419-427.
Upton, D., B. Kim. 1998. Alternative methods of learning and process improvement in manufacturing.
Uzzi, B. 1996. The sources and consequences of embeddedness for the economic performance of
organizations: The network effect. American Sociological Review 61(4) 674-698.
Uzzi, B. 1997. Social structure and competition in interfirm networks: The paradox of embeddedness.
Administrative Science Quarterly 42(1) 35-67.
Valente, T., K. Coronges, C. Lakon, E. Costenbader. 2008. How correlated are network centrality
measures? Connections 28(1) 16–26.
van Donselaar, K. H., V. Gaur, T. van Woensel, R. Broekmeulen, J. Fransoo. 2010. Ordering behavior in
retail stores and implications for automated replenishment. Management Science 56(5) 766-784.
Wasserman, S., K. Faust. 1994. Social Network Analysis: Methods and Applications. Cambridge
University Press, New York, NY.
Yukl, G. 2002. Leadership in Organizations, 5th Edition. Prentice Hall, Upper Saddle River, NJ.
Zu, X., L. D. Fredenhall, T. J. Douglas. 2008. The evolving theory of quality management: The role of
six sigma. Journal of Operations Management 26(5) 630-650.
43
Page 46 of 55
Ac
ce
p
te
d
M
an
us
cr
ip
t
Figure 1: Conceptual Model
44
Page 47 of 55
ip
t
New product
development
(NPD) process
improvement
New product
development
(NPD)
us
Examples*
 Reduce process defects and downtime associated with pumps
delivering a liquid raw material to the process.
 Reduce the incidence of a contaminant affecting product aesthetics,
which is the primary cause of scrap.
M
an
ed
Characterize
product
ce
pt
Characterize
process
Description
“Classic” quality improvement projects focused on
reducing defects, waste, or cycle time. Generally, the
primary focus is on identifying special cause variation
and bringing the process into control, identifying and
removing waste, or reducing cycle time using
techniques such as process simplification.
Projects focused on identifying and understanding the
effects of process control parameters, generally via
conducting experiments. Such projects focus on
processes that are already in a reasonable state of
control. These projects usually attempt to optimize
the parameter settings once the effects are understood.
Projects focused on identifying and understanding the
effects of design parameters on product performance,
generally via conducting experiments. Similar in
concept to characterize process projects, except the
focus is on the product rather than the process.
Projects focused on improving the new product
development process, including prevention of design
errors, reduction of design and testing cycle time,
reduction of design costs, etc.
Ac
Project Type
Improve
existing
process
cr
Table 1: Six Sigma Project Type Descriptions
 Conduct designed experiments to understand how process
parameters, such as production line speed, temperature, etc., affect
dimensional conformance and thus fit among assembled
components.
 Experiment with product materials to determine if a less expensive
combination can be used.
 Experiment with the materials and geometry of parts to improve a
seal and thus increase product reliability and durability.
 Standardize the options available to designers for a specific product
component to reduce time to market. Developing such a portfolio of
possible choices saves design engineers’ time and reduces the need
to qualify this component for each new product developed.
 Reduce the cycle time of testing required for proof of new product
concept. Previously, several different tests were run at different test
settings. It was determined that fewer tests could be used with
settings optimized to gain maximum testing coverage.
 The current product offering requires two-handed use. Experiment
with alternative designs to determine if a version of the product can
be developed that can be used with one hand. For this particular
product, the geometry of the product case necessary for one-handed
use potentially affects its performance and durability.
Projects focused on developing new product features
or new product offerings (in contrast to optimizing an
existing feature or product). Such projects often
include designed experiments to assess feasibility of
new product concepts or to solve some specific
technical problem preventing the feasibility of a
product concept.
* These examples are derived from specific closed improvement projects at our research site. However, they have been described in a general way
without specifics in order to protect the identity of our research site and its confidential information.
Page 48 of 55
Table 2: Descriptive Statistics
PANEL A
50.7%
49.3%
100%
Business Group
BG1
BG2
BG3
Total
75
50
27
152
49.3%
32.9%
17.8%
100%
UsesStatTools
Yes
No
Total
108
44
152
71.1%
28.9%
100%
M
Variable Namea
Team Size
Team Leader Familiarity
Bonacich Connectedness
Closeness Connectedness
Betweenness Connectedness
Team Leader Project Type Experience
Organizational Experienceb
Min Max
2
19
0
3
-1.82 2.09
–2.25 2.09
–0.38 2.28
0
3
0 31.75
Ac
ce
p
te
d
PANEL B
a
cr
77
75
152
an
Six Sigma Project Success
Yes
No
Total
ip
t
Percentage
us
N
Median
4
0
-0.65
–0.71
–0.13
0
5.27
Mean
4.74
0.35
-0.33
–0.56
0.12
0.21
9.19
Std Dev
2.61
0.68
0.85
1.08
0.64
0.51
8.90
The descriptive statistics we report in Panel B are for the untransformed variables.
Organizational experience is the number of prior projects initiated in the business group per one
thousand employees.
b
46
Page 49 of 55
-0.33***
0.14*
-0.36***
-0.24***
-0.08
-0.11
-0.08
0.18**
0.27***
ip
t
(4)
(5)
(6)
(7)
(8)
(9)
(10)
us
-0.26***
0.11
-0.16**
0.30***
0.20**
0.24***
0.32***
0.10
(3)
M
an
(2)
-0.05
0.03
0.06
-0.04
-0.11
0.06
-0.11
-0.18**
0.12
0.03
0.01
0.09
-0.03
0.04
-0.01
0.41***
0.43*** 0.77***
0.27*** 0.59***
0.54***
-0.15*
-0.06
-0.10
0.06
0.43*** 0.27***
0.37***
0.24***
-0.46***
-0.29***
-0.15*
0.18**
0.38*** 0.20***
0.33***
0.10
ed
(1) Six Sigma Project Success
(2) BG2
(3) BG3
(4) Team Size
(5) UsesStatTools
(6) Team Leader Familiarity
(7) Bonacich Connectedness
(8) Closeness Connectedness
(9) Betweenness Connectedness
(10) Team Leader Project Type
Experience
(11) Organizational Experience
(1)
cr
Table 3: Pearson Correlation Matrix of Model Variablesa
p ≤ .10; ** p ≤ .05; *** p ≤ .01
a
The correlation matrix shown is for the untransformed variables.
Ac
ce
pt
*
0.35***
47
Page 50 of 55
Table 4: Logistic Regression Results for the Four Competing Models
Panel A: Team Leader Familiarity
0.318
(0.533)
1.230**
(0.599)
-0.291
(0.238)
1.342***
(0.439)
0.203
(0.216)
0.610**
(0.263)
0.316
(0.565)
1.374**
(0.648)
-0.394
(0.253)
1.680***
(0.483)
0.733***
(0.293)
0.580**
(0.289)
-0.762***
(0.246)
0.348
(0.234)
-0.075
(0.203)
UsesStatTools ("no" reference group)
M
Team Leader Project Type Experience (TLPT)
Organizational Experience (OE)
d
TLPT * OE
an
Team Size
0.197
(0.210)
Ac
ce
p
te
Team Leader Familiarity (TLF)
TLF * OE
-0.675*
(0.416)
0.336
(0.485)
-0.370
(0.230)
1.158***
(0.420)
us
BG3
Model 3
-1.128**
(0.527)
ip
t
Business Group (BG1 reference group)
BG2
Model 2
-1.164**
(0.492)
cr
Intercept
Model 1
-0.559
(0.437)
Model 2 statistic
d.f.
AICc
Likelihood Ratio test p-value
(Models compared)
21.69***
4
199.58
---------
34.48***
7
193.47
0.000
48.27***
9
184.31
0.000
(Model 2 vs. 1) (Model 3 vs. 2)
Notes:
 The basis for comparison or “reference groups” that correspond to zero values of the business group
and UsesStatTools dummy variables are BG1 and non-use of advanced statistical tools, respectively.
 The regression coefficients reported in this table are standardized coefficients; *p ≤ .10, ** p ≤ .05;
*** p ≤ .01; the standard errors are in parentheses.
 The corrected Akaike Information Criterion (AICc) is a measure of the goodness-of-fit of a statistical
model that allows comparison across models with different numbers of independent variables (Akaike
1974). Smaller values of the AICc indicate better model fit.
48
Page 51 of 55
Table 4: Logistic Regression Results for the Four Competing Models (continued)
Panel B: Bonacich Connectedness
BG2
-0.675*
(0.416)
0.336
(0.485)
-0.370
(0.230)
1.158***
(0.420)
0.290
(0.541)
1.322**
(0.617)
-0.459*
(0.260)
1.445***
(0.457)
0.108
(0.216)
0.625**
(0.265)
Team Size
M
an
us
BG3
TLPT * OE
Ac
ce
p
Bonacich Connectedness (BC)
te
d
0.568***
(0.211)
BC * OE
d.f.
AICc
(Models compared)
21.69***
4
199.58
---------
Model 3
-1.157**
(0.543)
ip
t
Intercept
Model 2
-1.194**
(0.507)
cr
Model 1
-0.559
(0.437)
0.257
(0.575)
1.515**
(0.678)
-0.604**
(0.283)
1.761***
(0.502)
0.624**
(0.294)
0.630**
(0.288)
-0.693***
(0.244)
0.631***
(0.240)
-0.200
(0.245)
41.42***
7
186.53
0.000
54.61***
9
177.97
0.000
(Model 2 vs. 1)
(Model 3 vs. 2)
Notes:
49
Page 52 of 55
Panel C: Closeness Connectedness
BG2
-0.675*
(0.416)
0.336
(0.485)
-0.370
(0.230)
1.158***
(0.420)
0.224
(0.550)
1.346**
(0.618)
-0.417
(0.261)
1.391***
(0.453)
0.044
(0.219)
0.529**
(0.268)
Team Size
M
an
us
BG3
TLPT * OE
Ac
ce
p
Closeness Connectedness (CC)
te
d
0.653***
(0.215)
CC * OE
d.f.
AICc
(Models compared)
21.69***
4
199.58
---------
Model 3
-1.105**
(0.535)
ip
t
Intercept
Model 2
-1.166**
(0.510)
cr
Model 1
-0.559
(0.437)
43.46***
7
184.5
0.000
0.151
(0.574)
1.448**
(0.660)
-0.487*
(0.275)
1.677***
(0.490)
0.563*
(0.300)
0.528*
(0.287)
-0.714***
(0.247)
0.650***
(0.232)
0.031
(0.268)
54.34***
9
178.24
0.000
Notes:
50
Page 53 of 55
Panel D: Betweenness Connectedness
BG2
-0.675*
(0.416)
0.336
(0.485)
-0.370
(0.230)
1.158***
(0.420)
0.306
(0.533)
1.255**
(0.599)
-0.310
(0.240)
1.332***
(0.437)
0.218
(0.215)
0.670***
(0.257)
Team Size
M
an
us
BG3
d
te
TLPT * OE
0.132
(0.197)
Ac
ce
p
Betweenness Connectedness (BTWC)
BTWC * OE
d.f.
AICc
(Models compared)
21.69***
4
199.58
---------
Model 3
-1.063**
(0.518)
ip
t
Intercept
Model 2
-1.130**
(0.489)
cr
Model 1
-0.559
(0.437)
34.05***
7
193.91
0.000
0.270
(0.560)
1.398**
(0.645)
-0.395
(0.251)
1.601***
(0.473)
0.757***
(0.297)
0.667**
(0.281)
-0.738***
(0.231)
0.112
(0.222)
0.016
(0.231)
46.32***
9
186.26
0.000
Notes:
51
Page 54 of 55

Table 5: Summary of Hypothesis Test Results
Degree of
Support
Hypothesis
ip
t
No support
cr
Strong support
Strong support
us
H1a: Higher team leader familiarity is associated with increases in the
likelihood of six sigma project success.
H1b: Higher values of team leader Bonacich connectedness are
associated with increases in the likelihood of six sigma project
success.
H1c: Higher values of team leader closeness connectedness are
success.
H1d: Higher values of team leader betweenness connectedness are
success.
H2: Increases in a team leader’s experience leading the same type of
project (higher team leader project type experience) is associated with
H3: As organizational experience increases, the positive relationship
between our team leader social capital-related dimensions and the
likelihood of six sigma project success diminishes.
H4: As organizational experience increases, the positive relationship
between team leader project type experience and the likelihood of six
sigma project success diminishes.
Supported
No support
Strong support
te
d
M
an
No support
Ac
ce
p
Table 6: P-Values for the Competing Model Comparison Tests
Model pair
Two-sided pvalue
Team Leader Familiarity vs. Bonacich Connectedness
0.000
Team Leader Familiarity vs. Closeness Connectedness
0.168
Team Leader Familiarity vs. Betweenness Connectedness
0.062
Bonacich Connectedness vs. Closeness Connectedness
0.570
Bonacich Connectedness vs. Betweenness Connectedness
0.001
Closeness Connectedness vs. Betweenness Connectedness
0.007
Note: The smaller the p-value, the more evidence for the italicized and underlined model over
the other model listed in the pair.
52
Page 55 of 55

Team Leader Experience in Improvement Teams

Transcription

Similar documents

Mark Bove, CPCU, ARe Sigma Chapter Ice Skating Extravaganza

Six Sigma Green Belt -‐ ASQ (10 Days)

MSME –DI, Chennai Six Sigma Green Belt Program

Is There a Science to Law Firm Management and Does Six

Tokuyama - M&S Dental Supply

“Jay” Rumbley Vice President

Sandy Wibisono

( )

Cashmaster Sigma 105

Lean Six Sigma For Law Firms Why are we here?