Week-14

o Academy of Management ¡oumal
1996. Vol. 39, No. 4, 1005-1023.

FROM CHIMNEYS TO CROSS-FUNCTIONAL
TEAMS: DEVELOPING AND VALIDATING A

DIAGNOSTIC MODEL

DANIEL R. DENISON
University of Michigan and Hitotsubashi University

STUART L. HART
JOEL A. KAHN

University of Michigan

This article develops a Tratnework for studying cross-funclional teams
in organizations that focuses on three domains: organizational context,
internal process, and outcome measures. The framework was developed
from qualitative data from over 200 individual and group interviews,
written descriptions, and team observations. We then operationally de-
fined this model through a set of questionnaire items and validated
it through quantitative analysis of data from 565 members of cross-
functional teams. The resulting framework provides a base for the future
study of cross-functional teams.

Cross-fimctional teams (CFTs) are spreading rapidly in organizations as
they attempt to improve coordination and integration (Ford & Randolph,
1992; Knight, 1976), span organizational boundaries (Ancona, 1990; Ohmae,
1990), and cut cycle time in new product development (Ancona & Caldweli,
1992a; Hitt, Hoskisson, & Nixon, 1993; Takeuchi & Nonaka, 1986). Cross-
functional teams take many forms, hut they are most often structured as
working groups, created to make decisions lower in an organization’s hierar-
chy, that have links to multiple subunits—or “chimneys”—and are designed
as an overlay to an existing functional organization (Galbraith, 1994). CFTs
share many characteristics with conventional teams, but they also differ in
important ways: First, they are usually representative groups in which each
member has a competing social identity and obligation to another subunit
of the organization (Alderfer. 1987; Alderfer & Smith, 1982; Brown, 1983).
Second, they are often temporary task teams experiencing abundant pressure
and conflict, so the early development of stable and effective group processes

We wish to Ihank Susan Ashford, Jane Dutton. Richard Hackman, Susan Schneider, James
Walsh, Karl Weick, and the members of the Michigan Interdisciplinary Organization Studies
Seminar for their comments on earlier drafts of this article, as well as the many managers and
executives who participated in and supported this study. In addition, our thanks to this journal’s
editors and anonymous reviewers for their comments and advice. Finally, we wish to thank
Karen Gromala for her many contributions to this article.

1005

1006 Aeademy of Management Joumal August

is critical to their success (Ford & Randolph, 1992; Gersick & Hackman,
1990). Third, cross-functional teams typically confront a different set of
performance expectations than conventional work teams and are often ex-
pected to reduce cycle time, create knowledge, and disseminate organiza-
tional learning (Nonaka, 1991; Pinto & Slevin, 1988a, 1988h).

Empirical research on cross-functional teams, however, has lagged well
behind their rate of adoption (Knight, 1976). Existing research on teams and
their organizational contexts provides a useful point of departure hut has left
many important questions unanswered (Ancona, 1990; Ancona & Caldwell,
1992a; Gladstein, 1984; Hackman, 1987, 1990). In this article, we sought to
develop a framework and a set of measures for examining cross-functional
teams by first collecting and analyzing qualitative data from 200 members
of CFTs to develop a conceptual framework and an item pool. We then
refined those measures through the analysis of survey data from three separate
samples totaling 565 team members. Tbus, the primary contribution of this
research is to examine a new and emerging organizational form and to de-
velop a framework and a set of measures that can be used in studying that
form. In addition, this article makes a contribution by contrasting CFTs witb
other types of teams and linking this researcb witb the existing literature on
teams and their contexts.

CROSS-FUNCTIONAL TEAMS: BUaOING ON THE LITERATURE

Although there is little existing literature directly on cross-functional
teams, two related streams of research have served as a starting point: The
first of these is the literature on teams and their organizational contexts
(Ancona, 1990; Ancona & Caldwell, 1992b; Gladstein, 1984; Hackman, 1987,
1990), and the second is the work on product development teams, which
has frequently addressed the process of combining varied sources of expertise
to create innovative outcomes (Gohen & Ledford, 1991; Donnellon, 1995;
Katz, 1982; Katz & Allen, 1985; Pelz & Andrews, 1976; Takeuchi & Non-
aka, 1986).

The Organizational Context of Teamwork

Many authors have addressed the topic of team-based organizational
design (Galbraith, 1973; Hackman & Walton, 1986; Leavitt, 1975; Manz, 1992;
Pasmore, 1988), but relatively few have discussed the organizationai context
of teams—the overarching structures and systems external to a team that
facilitate or inhibit its work. We begin by reviewing two distinct approaches
to organizational context that appear in the literature, the first focusing on
an organization’s impact on teams (Hackman, 1987, 1990) and the second
examining the teams’ attempts to influence the larger organization (Ancona,
1990; Ancona & Caldwell, 1992a, 1992b).

Tbese two approacbes sbare several characteristics. For example, both
perspectives examine context by looking at the interactions that occur in the
interface between individual teams and their organizational environment.
As such, they share a definition of context as something external to the team

1996 Denisan, Hart, and Kahn 1007

yet interna! to the organization. In both approaches, team effectiveness is
also a concern and is addressed through three-stage context-process-outcome
models. These models are linked to more traditional social psychological
input-process-outcome models (Hackman & Morris, 1975; Sundstrom, De-
Meuse, & Futrell, 1990) and incorporate many of the suggestions offered hy
Guzzo and Shea (1993), Mowday and Sutton (1993), and O’Reilly (1991) in
their discussions of the importance of developing a contextual perspective
in group and organizational research.

Hackman’s (1987,1990) model of team effectiveness conceptualizes the
impact of organizational context on teams in terms of three main influences:
(1} the design of a team and its task, (2) the transfer of information, resources,
and rewards to the team, and (3) the existence of process assistance that can
he provided to facilitate a team’s work. These three factors define the context
domain and incorporate it within a general model of the effectiveness of work
teams that also includes team processes and task outcomes. Our research was
also influenced hy Hackman’s discussion of the relationship between context,
team processes, and outcomes as well as by Gladstein’s (1984) study showing
that contextual factors were more potent determinants of team effectiveness
than internal team processes.

The second approach to studying teams and their organizational contexts
focuses on how teams manage their interface with the larger organization
(Tushman, 1977; Tushman & Katz, 1980). Ancona and Caldwell (1988), for
example, examined how project teams managed their organizational contexts
in high-technology companies. Ancona (1990) extended this research by
descrihing the role orientations of effective and ineffective consulting teams
in an educational organization. Teams that were isolated, passive, or overly
technical were far less successful than teams that proactively managed the
political dynamics of their client organization. Ancona and Caldwell (1992b)
reported similar findings from a study of product development teams.

Both of these perspectives have been applied to a wide range of teams,
and they help to explain many aspects of the functioning of cross-functional
teams. But several aspects of the multifaceted context of such representative
teams are not addressed, since the primary purpose of these models is to
conceptualize a team’s interaction with a primary context. Hackman’s model,
for example, conceptualizes context in terms of the transfer of information,
resources, and rewards from organization to team and pays only limited
attention to possihle conflicts and inconsistencies among the sources of these
inputs. In addition, little empirical research has been done on the factors
outlined in Hackman’s (1987, 1990) model of team effectiveness.

Ancona (1990) and Ancona and Caldwell (1988, 1992a, 1992h) also
presented a model that tends to conceptualize the influence of a team on a
relatively unitary (albeit complex) organizational context. They argued that
team memhers manage context through their strategic role orientations—
teams with “task coordinators” or “ambassadors,” for example, will be hetter
able to control their context than teams whose members play the more passive
role of “scout.” These authors paid less attention, however, to the targets of

1008 Academy of Management Journal August

a team’s scouting or diplomacy, or to the need to scout one context, coordinate
with a second context, and be diplomatic toward a third. CFTs must, however,
coordinate action among a diverse set of functions, interests, and areas of
expertise in order to he successful (Donnellon, 1995; Manz & Sims, 1987).

The existing literature on teams provides a rich base for conceptualizing
the internal process of cross-functional teams, hut neither Hackman nor
Ancona included a particularly rich set of process dimensions in their studies
of context, process, and outcomes. Since the fragmented context in which
these teams operate may in fact place significant strains on team processes,
it seems important for new research to reexamine the role of such processes
in conjunction with a study of a multifaceted context. Finally, the relevant
range of outcomes for CFTs has received little attention in existing research
on context. Although such teams’ tasks are integrative and often creative,
existing studies of context have typically examined conventional outcomes,
such as task performance, cohesiveness. and memher satisfaction. Concern
with these outcome issues, in particular, led us to examine additional re-
search literature on product development teams.

Product Development Teams: Internal Processes and Outcomes

As noted in the introduction to this article, there are many different
varieties of CFTs, including planning teams, ad hoc project teams, quality
teams, process improvement teams, and product development teams. Al-
though there are important differences among these types of teams, they all
are typically task teams, with memhers representing multiple organizational
functions, that are formed to integrate expertise from those functions and
operate at a lower level in the hierarchy.

One research literature that helps contribute to a general understanding
of CFTs is that on product development teams, a specific type of cross-
functional team. Product development teams focus on comhining diverse
sources of expertise in order to develop innovative new products. Although
some product development is still done hy functional (for instance, engineer-
ing) teams, this approach is increasingly rare. Research on product develop-
ment teams in many cases helps to highlight the general problems facing
CFTs of all types.

Research on project management, new product development, and organ-
izing for innovation and creativity (Imai, Nonaka, & Takeuchi. 1985; Larson &
Gobeli, 1988: Pelz & Andrews. 1976; Tushman. 1978; Tushman & Nadler.
1986) has shown that effective new product development occurs when two
conditions are present: First, the requisite diversity of viewpoints, disci-
plines, and functional specialties is represented in a team, and second, the
team is able to span organizationai boundaries and integrate the functional
expertise represented hy team members (Fruin. 1996).

This literature addresses the issue of the relationship between CFTs and
their contexts but does so primarily in terms of the conflict created hy matrix
structures (Davis & Lawrence, 1977; Galbraith, 1973, 1994; Larson & Gobeli,
1987). Most product development teams are influenced by one line of author-

Denison, Hart, and Kahn 1009

ity for the functions—the source of the specific forms of expertise required
to develop a new product—and a second line of authority for the project
organization developing the product (Goheli & Larson, 1987).

Implicit in this literature is the notion that cross-functional teams can
perform systemic tasks such as product development better than functions
working in isolation (Ancona & Caldwell, 1992a; Clark & Fujimoto, 1991;
Nonaka, 1990; Takeuchi & Nonaka, 1986]. This idea, however, often neglects
the fact that team membership, identity, and loyalty in CFTs may be in
question (Donnellon, 1995). Donnellon argued that CFTs are less often teams
with a high degree of interdependence, a group task, and a strong group
identity than they are “co-acting” work groups composed of independent,
highly competitive individuals who pursue their own goals over those of
the teams (Hackman, 1987; Katzenbach & Smith, 1993). These issues seem
to suggest that internal team processes are an important mechanism to create
organizational coordination (Ancona & Caldwell, 1987, 1990; Barker, Tjos-
vold, & Andrew, 1988).

The literature on product development teams also suggests a more rele-
vant and complete set of outcome measures than traditional research on
groups and their contexts (Gobeli & Larson, 1987; Pinto & Slevin, 1988a,
1988b). This research adds to the traditional measures of task accomplish-
ment, cohesion, and member satisfaction, concerns such as time compres-
sion, innovation, the integration of diverse viewpoints, learning, and capabil-
ity development (Clark & Fujimoto, 1991; Takeuchi & Nonaka, 1986).

Although the team context and product development literatures have
generated important insights concerning the contexts, processes, and out-
comes of effective teams, they have produced less in the way of theory
development, comparative research, and development of measures. Thus,
our central concern, the development of a framework for understanding the
effectiveness of cross-functional teams and a complementary set of valid
measures, has received only limited attention in either literature.

METHODS

This study was conducted in two separate stages; First, we gathered
qualitative data from stories, interviews, written descriptions, and observa-
tions. A content analysis of these data suggested a set of issues that cross-
functional team members thought were important to the effectiveness of their
teams. Next, we used these issues and the qualitative data from which they
were drawn as a basis for generating a pool of questionnaire items that began
to operationally define influences on team effectiveness (Pinto & Prescott,
1988). Drawing on previous research that has defined context, process, and
outcomes as important domains in team research (Ancona, 1990; Ancona &
Caldwell, 1992b; Gladstein, 1984; Hackman, 1987, 1990), we grouped the
issues and items into three domains and began to identify the most salient
items and dimensions for each domain. Analysis of data from two samples
of members from a diverse collection of CFTs was used to refine a set of
measures for the dimensions within each domain. All data were collected

1010 Academy of Management Journal August

from multiple divisions and locations of one multinational organization,
a large American automobile manufacturer, and included data from team
members in 15 different countries. This article reports the results in detail
for the final sample of 364 members of 43 product development teams.

We began collecting data by conducting 50 informal interviews with
members of CFTs, including planning teams, quality teams, process improve-
ment teams, and product development teams. Each interview focused on the
effectiveness of a team that the interviewee had been a member of, and the
data were captured as a set of stories about the important characteristics of
CFTs. Next, we collected written descriptions of teams from 174 middle
managers who had been members of CFTs and were participants in a manage-
ment development program. They were asked to describe, in writing, an
effective cross-functional team and an ineffective cross-functional team and
to list the characteristics that they thought distinguished effective and ineffec-
tive teams.

Data from both of these sources were content-analyzed and then summa-
rized in terms of a set of concepts and themes that captured the range of
issues represented in the qualitative data. We grouped these concepts and
themes into the three domains of context, process, and outcomes as suggested
by the literature. From this set of concepts and the coded qualitative data
from which they were drawn, we then developed an initial pool of question-
naire items to be used as the basis for a set of measures.

The second stage of the research focused on refining a valid and reliable
set of measures for the context, process, and outcome domains suggested by
the literature. In several cases in which the context was similar to existing
scales in the literature, we borrowed items to help measure dimensions
(Hackman, 1987; Hart, 1985). Next, we administered a questionnaire witb this
set of 123 items to 200 middle managers in a company executive development
program. Respondents were asked to identify a team that they had been a
member of and to respond to the items with respect to that team. On the
basis of resuhs, we added, discarded, or reworded items. This refinement
process also involved frequent reference back to our qualitative data (Glaser &
Strauss, 1967). This process was then repeated a second time on a second
sample of 200 middle managers.

The resulting set of items was then administered to the members of 43
intact CFTs who were involved in three large product development programs,
each of which was dedicated to one vehicle. Each of tbese programs was led
by a program manager and was made up of 15-30 cross-functional teams
that held responsibility for separate modules or subunits of the vehicle.
Each of these teams had a core of 5-20 members who came from functional
organizations such as design, body or powertrain engineering, manufactur-
ing, marketing, purchasing, and so on. Since the official list of team members
often included nominal members who did not attend meetings or contribute
much, we asked the leaders of each team to identify core members of the
team, those who attended meetings regularly, spent at least 20 percent of
their time working for the team, and were central to the team’s activities.

1996 Denison. Hart, and Kahn 1011

Using this procedure, we targeted 565 individuals who were members of 43
CFTs as questionnaire respondents. Surveys were distributed to core team
members by their team leader at one of their regular meetings, and respon-
dents were given two weeks to complete the survey and return it to us
using an enclosed envelope. After three weeks, a second set of surveys was
distributed to those who had not yet responded. Overall, 364 respondents
representing 43 CFTs responded for an overall response rate of 65 percent.
The respondents were primarily managers and engineers between the ages
of 25 and 55, 85 percent of whom were men.

RESULTS

We begin by presenting a brief story for each domain, illustrating some
of tbe dimensions identified by our qualitative researcb; we then present a
factor analysis of the item pool for the domain. Finally, we present a second-
order confirmatory factor analysis conducted to determine if the derived
factors fit into the three domains suggested by the literature and by our
qualitative analysis.

Organizational Context

The context of cross-functional teams is complex and differs from that
of more conventional teams in that it includes hierarchical, lateral, and
interteam dependencies that require continuous negotiation. For teams to
exist within an organizational environment, they must define their role in
relation to upper management and resolve the inherent conflicts between
the functions tbat tbey represent. An example drawn from an interview in
one product development team shows how difficult tbis can be;

The total amount of electrical power in a vehicle is determined
by the capacity ofthe alternator. This power must serve over 20
subsystems, such as the stereo, the engine, the instrument panel,
and so on. These subsystems are developed and controlled hy
separate “chimney” organizations, and power allocations must
be made for each system. The problem was, in this vehicle pro-
gram, when the requirements of all the chimneys and teams
were added up, they equaled 125 percent of the capacity of the
alternator. Keith, who had recently taken over as head of this
vehicle program, which had made many changes in direction
and was behind schedule to begin with, called a meeting ofthe
Program Steering Team designed to resolve this conflict and reach
a compromise. However, many of the chimney representatives
who were the members of the team came to this meeting with
instructions from their bosses (who, incidentally, did their per-
formance appraisals . . . ) not to make any compromises, but to
make certain that their chimney “got what it needed” and “didn’t
lose out.” After Keith presented the group with the problem and
the need to reach a compromise solution, their response surprised
him: “It’s not our problem,” they replied, “it’s your problem.”
Keith soon changed jobs again.

Tbis story illustrates several of tbe critical dimensions of organizational
context and tbeir relation to team effectiveness. Power dynamics between

1012 Academy of Management Journal August

teams and functional organizations can greatly limit the autonomy of a team
and must be managed proactively. Furthermore, when a team’s mission is a
“moving target,” its power is diminished. In addition, coordination with
other cross-functional teams is often required. Finally, many needed re-
sources and rewards, as this example illustrates, are controlled hy the func-
tional organizations rather than the team itself.

Table 1 presents the exploratory factor analysis (principal components
with varimax rotation) for the context domain. Factor loadings above .50,
our cutoff point for including an item in a factor, are presented in hold type.

The results show six dimensions with eigenvalues greater than 1.0: coor-
dination with other teams; autonomy and power; linkage to fimctions; re-
sources; mission and direction; and reward for team performance.

Team Processes

The conflicting demands inherent in the organizational context place a
premium on a team’s internal processes. Team members need to he both
representatives of their functional authority structures and creative prohlem-
solvers. Meeting both of these demands requires a broad and flexible team
process. The following story, taken ft-om a process improvement team, illus-
trates some of these dynamics:

Our formalized approach to managing cross-functional teams can
really get in the way. [For example,] . . . our organization has
several common practices that make an effective team process
difficult. [First,] . . . we always meet in large horseshoe shaped
conference rooms and take turns making presentations to each
other. Very few of these meetings actually lead to creative proh-
lem solving. Sometimes, a representative from one of the chim-
neys will come in and “drop the homh,” (deliver unexpected
bad news) and then leave. [Second.] . . . the team is really a
collection of subteams that come and go during a meeting de-
pending on what’s on the agenda. The “heavy hreathers” (bosses)
hack home in the ftmctions will sometimes even send a delegate
to our mootings with instructions to just watch and “make sure
nothing happens.” The result is thai the group doesn’t take collec-
tive responsibility, and that can be very demotivating. [Third,]
. . . many teams are assigned “content” leaders and “process”
leaders. The content leader is the technical expert—the “real”
boss—and the process leader helps run meetings and manage
the team. The problem is that an effective team has to have
flexihle leadership—leadership and expertise need to change
according to the issue and get passed aronnd as needed.

This example illustrates the demands for hoth functional representation
and creativity and the need to develop team identity and normative expecta-
tions if a team is to take collective responsihility for resolving a diverse
set of demands. Leaders who facilitated flexihle problem solving and team
development seemed to have an advantage.

Table 2 presents the analysis for the team process domain. It reveals six
factors: norms; the importance of the team’s work; effort; efficiency; creative
strategy; and breadth.

1996 Denison, Hart, and Kahn 1013

CQ

S
o
Q

£f

o O í r – r ” l O r ” 0 ) CT
eo ifl r- M l D N r – C

1 1 1

1014 Academy of Management Journal August

r – fv •!•

>- * rv r ” m o

q CO

IM

0

CO

ig96 Denison, Hart, and Kahn 1015

Outcomes

The criteria of effectiveness for cross-functional teams are many and
varied. As tbe example below illustrates, success can often require that all
criteria be met, since failure on any criterion may mean failure for the en-
tire project.

For one “high profile” new product, a program manager and
core team were selected and given the responsibility for hitting
specific “targets” for cost, weight, and timing. Since program
management represented a new way of working, the team mem-
bers necessarily “invented it as they went along.” to a great
extent. The excitement surrounding this new effort generated a
great deal of learning and innovation, both about the product
and how to work in a cross-functional environment. The team
members greatly expanded their abilities and understandings
over the course of the program. Despite several false starts and
blind alleys, the product ultimately won “Car of the Year” honors
upon its introduction. However, because the program missed the
cost and weight targets established by upper management, the
program managers never survived to apply their skills to a sec-
ond program.

This story points to the wide range of outcomes expected from cross-
functional teams: Not only are they expected to produce innovation, learning,
and new capabilities, but they are also called upon to compress time and
hit stringent task targets. Member satisfaction is also a key outcome if the
team approach is to remain a viable way to organize future product develop-
ment efforts.

Table 3 presents the analysis for the outcome domain. It reveals seven
factors: information creation, time compression, image expansion, learning,
growth satisfaction, overall effectiveness, and capability development.

Table 3 also shows that information creation and time compression items
loaded on the same factor, as did learning and growth satisfaction items. A
more detailed analysis of each of these sets of items sbowed that they sepa-
rated into two factors when analyzed alone. Since some potentially important
theoretical issues distinguish these constructs, they are reported separately
in Table 3.

Figure 1 presents the three-domain model and the factors included in
each domain. We examined tbe validity of these three domains by performing
a second-order confirmatory factor analysis. In this analysis, we took the
indexes from the exploratory analyses reported in Tables 1-3 as observed
variables and the three domains as latent variables. Table 4 presents a correla-
tion matrix for these indexes.

Table 5 presents the results of this confirmatory factor analysis. These
analyses were performed with LISREL VI (Jöreskog & Sörbom, 1984). Factor
loadings (ksi) are presented for each index. All measures have high loadings
in the expected domain, and all are statistically significant. The fit statistics
(GFI = .829, x^ = 343.02, df = 149) indicate that tbe data fit the model
reasonably well, although there is some room for improvement. In addition,

1016 Academy of Management Journal August

.5’3
S
o
Q
aBau
O
03

u o

5 “5
H »Hcd

u
cd

b .

m 05 o 05
f» o I^ M
q •-• r- q

v^ iD <0 Ö O O í - í r - O O O r - i - (O 01 m ĉ i c^ r r c J- if le a rE • < in O U I m op l (V CL, al u C! W he n It em 1 m C -aG m a o 'S fíi M U c o •a •v a• a o u ai 2 E >

c S

1996 Denison, Hart, and Kahn

FIGURE 1
A Model of Cross-Functional Team Effectiveness

1017

Context

Coordination with Other
Teams

Autonomy and Power

Linkage to Functions

Resources

Mission and Direction

Reward for Team
Performance

Process

Norms

Importance of Work

Effort

Efficiency

Creative Strategy

Breadth

Outcomes

information Creation

Time Compression

Image Expansion

Learning

Growth Satisfaction

Capability Development

Overall Effectiveness

when the fit of the domain model was compared to the fit of the nested null
model, the normed-fit index (Bentler & Bonett, 1980) was .891. The highest
modification index value was 22.14 for the satisfaction index and the …