ArticlePDF Available

The Convergence of Planning and Execution: Improvisation in New Product Development

July 1998
Journal of Marketing 62(3):1-20

DOI: 10.1177/002224299806200301

Authors:

Christine Moorman

Duke University

Anne S. Miner

University of Wisconsin–Madison

The field of marketing strategy often makes the important assumption that marketing strategy should occur by first composing a plan on the basis of a careful review of environmental and firm information and then executing that plan. However, there are cases when the composition and execution of an action converge in time so that, in the limit, they occur simultaneously. The authors define such a convergence as improvisation and develop hypotheses to investigate the conditions in which improvisation is likely to occur and be effective. The authors test these hypotheses in a longitudinal study of new product development activities. Results show that organizational improvisation occurs moderately in organizations and that organizational memory level decreases and environmental turbulence level increases the incidence of improvisation. Results support traditional concerns that improvisation can reduce new product effectiveness but also indicate that environmental and organizational factors can reduce negative effects and sometimes create a positive effect for improvisation. These results suggest that, in some contexts, improvisation may be not only what organizations actually practice but also what they should practice to flourish.

Content uploaded by Christine Moorman

Content may be subject to copyright.

The Convergence of Planning and Execution: Improvisation in New Product Development

Author(s): Christine Moorman and Anne S. Miner

Source:

Journal of Marketing,

Vol. 62, No. 3 (Jul., 1998), pp. 1-20

Published by: American Marketing Association

Stable URL: http://www.jstor.org/stable/1251740 .

Accessed: 23/09/2013 16:38

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .

http://www.jstor.org/page/info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of

content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms

of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to

Journal of Marketing.

http://www.jstor.org

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

Christine Moorman and Anne S. Miner

The

Convergence

Planning

and

Execution:

Improvisation

New

Product

Development

The field of marketing strategy often makes the important assumption that marketing strategy should occur by first

composing a plan on the basis of a careful review of environmental and firm information and then executing that

plan. However, there are cases when the composition and execution of an action converge in time so that, in the

limit,

they occur simultaneously. The authors define such a convergence as improvisation and develop hypotheses

to investigate the conditions in which improvisation is likely to occur and be effective. The authors test these hy-

potheses in a longitudinal study of new product development activities. Results show that organizational improvi-

sation occurs moderately in organizations and that organizational memory level decreases and environmental

turbulence level increases the incidence of improvisation. Results support traditional concerns that improvisation

can reduce new product effectiveness but also indicate that environmental and organizational factors can reduce

negative effects and sometimes create a positive effect for improvisation. These results suggest that, in some con-

texts, improvisation may be not only what organizations actually practice but also what they should practice to

flourish.

he fields of marketing strategy in general and new

product development in particular

appear

to assume

that

marketing

strategy

should occur by first compos-

ing a plan on the basis of a careful review of environmental

and firm information and

then

executing

that

plan.

In this ar-

ticle, we question this assumption

by suggesting that there

are cases when the composition and execution of an action

converge in time so that,

in the limit, they occur simultane-

ously. We define such a convergence

of composition

and

ex-

ecution as improvisation

and suggest that the narrower

the

time gap between composing and performing

(or planning

and implementation),

the more that act is improvisational.

Marketing

literature,

similar to many businesses fields,

has paid relatively little attention to extemporaneous

be-

havior and has taken a rather

strong stand in favor of mar-

keting planning, especially of a formal type (e.g.,

Armstrong 1982; Sinha 1990). Although there have been

important exceptions in marketing and related literature

(Holbrook 1995; Hutt, Reingen, and Ronchetto 1988;

March and Simon 1958; Mintzberg 1994), a rational

plan-

ning norm has become deeply entrenched. Table I

overviews a sample of marketing

and product

strategy

text-

books that indicates a fairly uniform

acceptance of the be-

Christine

Moorman

Associate

Professor

Marketing

and

Anne S.

Miner

is Associate

Professor of

Management,

Graduate

School

Business,

Uni-

versity

Wisconsin-Madison.

This research

benefited

from

the

support

Paula

Bassoff,

Aric

Rindfleisch,

Bob

Drane,

Gabor

Kemeney,

Stephanie

Dixon,

David

Robinson,

and

Rona

Velte and

the

comments

Peter

Dick-

son,

Kathy

Eisenhardt,

Mary

Hatch,

Jan

Heide,

and

Thekla

Rura

pre-

vious versions

of the article.

National

Science Foundation

Grant

SBR-9410419,

the

Marketing

Science

Institute,

and the

University

Wis-

consin School

of Business

Sabbatical

Fund

have

supported

this

research.

lief that

marketing

strategies

should, in general, first be for-

mulated and then implemented.

Much of the research

literature

in marketing

shares this

view. For example, a review of published

articles

cited in a

recent

special issue of the Journal

of Marketing

Research

innovation and new products reveals a tendency to view

planning as the accepted norm while acknowledging that

business environments

are becoming increasingly dynamic

(Wind

and Mahajan

1997). Studies there

and elsewhere sug-

gest that

product

development

cycle times are faster

(Griffin

1997), failure rates lower (Cooper and Kleinschmidt 1986;

Montoya-Weiss and Calatone 1994), financial returns

greater

(Ittner

and Larcker

1997; Song and Parry

1997), and

innovation

levels higher (Olson, Walker,

and Ruekert

1995)

when companies take certain

advanced

planning

steps.

Our

purpose

in this article is not to detract

from the val-

ue of planning.

We believe that

planning

is an important

as-

pect of effective marketing management and decision

making.

However,

prior

theory

also points to several

reasons

why improvisation

sometimes can be a valuable and effec-

tive approach

to marketing

action. First, improvisation

can

be an effective choice when a firm faces environmental

tur-

bulence that requires

action in a time frame that is shorter

than

a regular

planning

cycle. For example, Egge (1986) de-

scribes how a salesperson

might improvise

when immediate

action is required

in the face of changing client demands;

Dickson (1997) suggests that fast learning and adaption

without much advance planning are important

to firm sur-

vival; Moorman

and Miner (1995) describe how a team im-

provised a new product formula in response to a surprise

introduction

of a competitive product;

and Eisenhardt

and

Tabrizi (1995) find that an experiential strategy involving

improvisation

works better in the computer industry

when

product

development must operate within the high level of

Journal of Marketing

Vol.

62 (July 1998), 1-20 Improvisation

in New

Product

Development

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

TABLE 1

Overview of How Marketing Strategy Textbooks View the Timing of Planning and Implementation

Aaker

(1988)

Boyd and Walker

(1990)

Boyd, Walker and

Larreche

(1998)

Cravens and Lamb

(1993)

Dalrymple

and

Parsons (1990)

Day (1990)

Dickson

(1997)

Jain (1997)

Kerin and

Peterson (1995)

Kotler

(1994)

Lehmann and

Winer

(1994)

Peter and Donnelly

(1998)

Quelch, Dolan, and

Kosnik

(1993)

Figure

2-1 provides an overview of the strategic market management process. This figure shows

external and self-analysis occurring

first, followed by strategy identification

and selection, which

includes, in order: mission specification, strategic alternative identification, strategy selection,

implementation,

and evaluation (p. 22).

The authors define strategy as "a fundamental pattern of present and planned objectives, re-

source deployments, and interactions

of an organization with markets, competitors, and other

environmental factors"

(p. 43). This definition

emphasizes planning, or the setting of objectives,

as an integral

part of strategy formulation.

Exhibit

1-7 depicts the marketing

management process as moving from analysis to the formula-

tion of strategic marketing programs, and finally

to the implementation and control of such pro-

grams (p. 16).

"Strategic

planning is a continuing

cycle of making plans, launching them, tracking performance,

identifying

performance gaps, and then initiating problem-solving

actions" (p. 688). The authors

provide eight comprehensive steps in preparing

and implementing the strategic marketing plan

(Exhibit

1, p. 688).

"Once

marketing plans have been prepared, they are used to guide field marketing

activities for

the planning period." Marketing

performance should be monitored to compare results with the

goals of the marketing

plan (p. 822).

Although Day incorporates both bottom-up and top-down approaches to decision making, the

implicit

strategy model in his book is that strategic planning precedes implementation and execu-

tion of the plan. This is evidenced in Figure

2-1 and because implementation is not an important

part of the text; instead, the emphasis is on strategy development.

Defines improvisation

as "impromptu

action" and notes (p. 37), "An organization's survival de-

pends on its ability

to learn and adapt quickly;

in practice, this means that plans often must be

altered at the very time they are being implemented." He sums up his position on improvisation

when he notes (p. 399), "In

summary,

the ying and the yang of product development is the plan-

ning discipline

of up-front

continuous environmental

analysis, targeting/positioning,

product speci-

fication,

judicious stage-gate reviews

of the emerging product's fit, feasibility,

and estimated prof-

itability

combined with the creative improvisation

from a team's many iterations

of prototype de-

sign and testing."

Exhibit

2-4 depicts the process of strategic marketing, which involves a clear separation of strat-

egy development and strategy implementation

(p. 33).

"The

selection of a course of action must be followed by development of a plan for its implemen-

tation. Simply deciding what to do will not make it happen. The execution phase is critical,

and

planning for it forces [you]

to consider source allocation

and timing

questions" (pp. 35-36).

Figure 3-1 (p. 63) shows a linear process moving from planning to implementing to controlling.

The process has feedback loops from controlling

back to implementation

and planning.

"In

general, the planning process works as shown in Figure 2-4. Whereas the collection and

analysis of data and the development of product strategies takes place in a limited time frame,

there is no beginning

or ending to the planning process as a whole. The formal part of the proc-

ess is followed by implementation,

during which programs

such as distribution,

promotion, adver-

tising, and the like are executed. Monitoring

and evaluating both the performance of the plan

and changes in competition and customers in the external environment are also continuous

tasks. This information

feeds back to the formal planning part of the process" (p. 31).

"[T]he

organization gathers information

about the changing elements of its environment.... This

information

is useful in aiding the organization to adapt better to these changes through the

process of strategic planning.

The strategic plan(s) and supporting

plan are then implemented in

the environment"

(p. 10).

'The marketing

process can be divided in several different ways. (Our)

conceptualization of mar-

keting tasks is: (1) marketing research, (2) marketing strategic formulation, (3) marketing plan-

ning, programming,

and budgeting, and (4) marketing organization

and implementation"

(p. 9).

2 / Journal of Marketing, July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

uncertainty created by quickly changing markets and

technologies.

Second, improvisation

might be prompted

when plan-

ning has not provided

all the details or tactics of implemen-

tation. Quinn's (1980, 1986) investigation of ten large

organizations finds that these organizations refined their

general strategic course incrementally

as new information

emerged from the environment.

Likewise, researchers have

reported

that many tactical marketing

decisions are not in-

cluded in marketing plans (Cosse and Swan 1983; Sutton

1990).

These reasons for improvisation

are consistent with a

more behaviorally based view of strategy development,

which asserts that organizationall

action often occurs with-

out much advance planning (Cohen, March, and Olsen

1972; Cyert and March 1992; Pfeffer 1982; Weick 1979).

This research

tradition therefore tends to focus on the be-

havioral

dynamics that

produce

effective action as opposed

to documenting

whether

organizations

adhere to normative

models of action (Burgelman 1983; Miner 1987, 1990;

Mintzberg

and McHugh 1985).

Several marketing

scholars have followed in this tradi-

tion and suggest that there is a gap between normative

and

descriptive

accounts

of marketing strategy

(Anderson 1982;

Day and Wensley 1983;

Wind and Robertson

1983). For ex-

ample, Hutt, Reingen, and Ronchetto (1988) find that un-

planned, innovative new product

activities, or autonomous

strategic

behavior

(Burgelman

1983), occur in organizations

and that marketing plays a key role in such activities by

virtue of its boundary-spanning

and product-championing

behavior.

Therefore,

there is a precedent

both in and outside mar-

keting literature

for examining

the topic of unplanned,

inno-

vative behavior.

However,

there are few cases of systematic

empirical investigations

of improvisation

specifically. Out-

side marketing,

research has tended

to focus on depth

analy-

sis of single improvisations by persons, groups, or

organizations (e.g., Hutchins 1991; Preston 1991; Weick

1993a). In marketing,

research

has been qualitative

(Moor-

man and Miner 1995) or focused more broadly on innova-

tion (Hutt, Reingen, and Ronchetto 1988).

We build on these important

observations

by developing

and systematically

testing a framework

that

studies the inci-

dence of improvisation

in new product activities. Because

improvisation

is a type of innovative

behavior

that often in-

volves fast learning

(Eisenhardt

and Tabrizi

1995), we draw

on existing literature

on innovation

and

learning

to make

our

case for why certain

environmental

and organizational

fac-

tors are important

to improvisation.

However, we are also

careful to show that improvisation

is distinct from innova-

tion and that these factors provide special insight into im-

provisation,

given its extemporaneous

nature.

In addition to examining the incidence of improvisa-

tion, we also seek to understand

the conditions in which im-

'Although

we use the

term

organizational

throughout

this

arti-

cle, we do not

mean

to suggest

that the

focal

action

is necessarily

at the

overall

organizational

level. In

general,

measurement

of or-

ganizational

phenomena

occurs

within

a strategic

business

unit

(Eisenhardt

and

Tabrizi

1995).

provisation is effective. As we have reviewed previously,

one research

tradition

is based on the belief that a lack of

advance planning reduces the chances of a firm's success

(e.g., Cooper and Kleinschmit 1986, 1987). Other

research,

however, has found success stories when organizations

have improvised. For example, Mintzberg and McHugh

(1985) detail the effective improvisations

of the National

Film Board of Canada;

Preston

(1991) describes a group of

managers

that improvises an effective solution to a manu-

facturing plant strike; Eisenhardt and Tabrizi (1995) show

that an experiential approach involving improvisation ac-

celerates product development in the computer industry;

Weick (1993a) describes the effective improvisations

sev-

eral firefighters used to escape the firestorm at Mann

Gulch; and Pascale's (1984) portrait

of Honda's successful

introduction

of its 50cc bikes into the U.S. market involves

improvisation.

It is not our goal to suggest that improvisation inher-

ently is either helpful or harmful for organizations.

Instead,

we study the literature on improvisation

to uncover

the sys-

tematic influence of various factors that affect whether im-

provisation hurts or helps organizations. We seek to

supplement

current

knowledge regarding

the types of envi-

ronmental

conditions and organizational

competencies that

might determine

the effectiveness of improvisation.

There-

fore, our objectives are twofold: (1) to investigate the con-

ditions in which improvisation

is likely to occur and (2) to

examine the conditions in which improvisation

is likely to

be effective.

Improvisation

Definition and Discrimination from Related

Constructs

Our review of prior

research

suggests that an assessment of

whether improvisation

occurs requires looking not just at

what

happens,

but also at the temporal

order

in which it hap-

pens (see also Moorman

and Miner 1998). Observers

typi-

cally assume that

composition

or planning

occurs first and is

followed at a later time by implementation

or execution. In

improvisation,

the time gap between these events narrows

that, in the limit, composition converges with execution

(Weick 1993a). Therefore, the more proximate the design

and implementation

of an activity in time, the more that ac-

tivity is improvisational.

This definition is consistent with core conceptualiza-

tions of improvisation

in several

bodies of literature.

For ex-

ample, improvisation

is referred

to as "thinking

in the midst

of action"

in education

(Irby 1992, p. 630), occurring

when

"acts

of composing and performing

are inseparable"

in com-

munication (Bastien and Hostager 1992, p. 95), "reading

and reacting in parallel" in sports psychology (Bjurwill

1993, p. 1383), "real-time

composition"

(Pressing 1984, p.

142; Pressing 1988) and "making decisions affecting the

composition of music during its performance"

(Solomon

1986, p. 226) in music, and representing

"no split between

design and production"

in organizational

studies (Weick

1993a, p. 6). By focusing on the simultaneity

of events, this

research

also follows in the tradition

of organizational

theo-

Improvisation

in New

Product

Development/

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

ries of temporal

order

(Cohen,

March,

and Olsen 1972; Van

de Ven 1986, 1993).

Improvisation

also can be distinguished

from other, re-

lated concepts in literature

(Moorman and Miner 1998).

Most important,

we argue that improvisation

is a special

case of intraorganizational

innovation,

which is defined as

deviation from existing practices or knowledge (Rogers

1983; Zaltman,

Duncan, and Holbek 1973). All improvisa-

tion, by definition, involves some degree of innovation be-

cause improvisation

involves the creation of action outside

current

plans

and

routines.

There are

many

other

kinds of in-

novation

beyond improvisation,

however.

For

example, if an

organization

innovates a new way to distribute a product by

analyzing customer needs, gathering

facts, and planning a

new channel, the organization

has innovated but not impro-

vised. In addition,

though

all improvisation

has some degree

of innovation,

the degree of innovativeness

can vary enor-

mously. For example, some improvisations

are very innova-

tive and deviate far from existing routines, such as in the

case of "free

jazz" (Berliner 1994) or of NASA teams im-

provising

to rescue

Apollo XIII

by making

radical use of ob-

jects outside prior

routines or structures

(Lovell and Kluger

1995). Other improvisations

only involve minor deviations

from current

routines, such as when musicians add embell-

ishments to existing melodies (Bailey 1980) or product

de-

velopment teams add features to existing products

(Miner,

Moorman,

and Bassoff 1997).

Improvisation

is also distinct from other constructs

im-

portant

to strategic

firm behavior,

such as adaptation,

learn-

ing, and opportunism.

First, improvisation

is distinct from

adaptation, which involves the adjustment

of a system to

external

conditions (Campbell 1989). We argue

that

adapta-

tion does not necessarily invoke improvisational

action by

organizations.

Instead,

adaptation

can be achieved in a vari-

ety of ways, including both preplanning

and improvisation.

Second, if learning is a process that requires

the discovery,

retention, and exploitation of stored knowledge, including

information

or behavioral routines

(Huber 1991; Levitt and

March 1988), then not all learning

is improvisation.

For ex-

ample, learning

might involve a carefully

preplanned exper-

iment whose results are recorded and interpreted

and that

requires

no improvisation

whatsoever.

Third,

when organi-

zations are described as opportunistic,

they are likely to

seize attractive,

unexpected developments or opportunities

proactively (Aaker 1988; Miner 1987; Mintzberg and

McHugh 1985; Quinn 1980, 1986).2

As with the other dis-

crimination

variables,

opportunism

can be achieved

through

means other than improvisation,

such as developing a plan.

In addition, improvisation sometimes arises in disastrous

situations and involves overcoming obstacles more than it

does taking advantage

of unexpected opportunities.

Organizational Improvisation

Many observers have described

improvisation

by individu-

als. For example, people have described

ways in which in-

dividual actors, athletes, therapists,

musicians, or teachers

2This

view of opportunism

benign

(Miner

1987,

p. 334)

omits

the definition of opportunism

as "self-seeking

with guile"

that

Williamson

(1975)

proposes.

improvise in different settings. Weick (1993a) describes in

detail the improvisational

and other actions of individual

firefighters

in a disastrous

firetrap,

though he also explores

team aspects of the situation.

In addition to individual improvisation,

observers have

emphasized that collective improvisation also occurs

(Crossan

and Sorrenti

1997; Preston 1991; Weick 1993a, b,

c). One detailed report (Hutchins 1991) describes actions

taken by the crew of a ship whose navigational system had

broken to make their

way into a harbor. To avoid danger,

the

crew members called out estimates of coordinates,

calculat-

ed subparts

of the data needed to make navigational

choic-

es, and communicated

partial information to one another

repeatedly. Although a transcript

of their interactions indi-

cates that no one crew member understood

the complete

system they had improvised

or exactly why they were suc-

ceeding, the crew developed a set of routines that worked to

get the ship into the harbor.

We follow this approach

by focusing on organizational

improvisation,

which includes improvisation

by groups,

de-

partments,

or whole organizations.

Many observers assume

that such entities generate

and execute plans. Therefore,

attempt

to study them as carrying

out these activities either

in sequence or nearly simultaneously,

and thus improvising

or not. Although the nature of collective or organizational

features remains contentious (Argyris and Schon 1978;

Walsh 1995), we follow other work describing organiza-

tional features, such as culture (Deshpande, Farley, and

Webster

1993; Deshpande

and Webster

1989), organization-

al information

routines

(Jaworski

and Kohli 1993; Kohli and

Jaworski

1990;

Moorman

1995), and memory

(Cohen 1991;

Huber

1991;

Moorman and Miner 1997; Walsh 1995;

Walsh

and Ungson 1991). To highlight the nature of organization-

al improvisation,

Appendix A provides examples of firms

varying

in the degree of improvisation

in their new product

activities.

Prior

research

suggests that interactions

among persons

who are improvising

frequently

produces

collective improvi-

sation. In an improvisational

theatrical

group, for example,

one actor

might

make a comment,

to which a second will re-

spond

with an association

between the comment and another

topic, and then

a third actor

might link these issues to a third,

inclusive topic (Crossan

and Sorrenti

1997; Mangham

1986;

Spolin 1963). The theatrical

group did not plan the scene in

advance,

and the pattern

that arises is not a simple sum of in-

dependent

improvisational

actions but a collective system of

interaction

that

creates and

enacts the scene simultaneously.

The idea that a system of interaction can produce col-

lective improvisation also is supported by Hutchins's

(1991) case study of the crew improvising a solution to

their

ship's failed navigation

system and by Barley's (1986)

observation

that hospital technicians and radiologists

joint-

ly improvised

routines in response to new technology. In a

powerful set of case studies displaying improvisation,

Dougherty (1992) describes new product development

teams interacting

in ways that

do not follow established or-

ganizational

routines and notes, "Successful developers vi-

olated ... routines and created a new social order for their

collaborative

efforts. They developed mutually adaptive

in-

teractions in which knowledge of the work was developed

4 / Journal of Marketing, July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

as the work unfolded"

(p. 192). In one example, Dougher-

ty (1992) cites the case of SALECO's development team,

which broke routines by using products assembled from

off-the-shelf parts purchased externally rather

than those

manufactured

in-house. This team also opted to introduce

software

product

with some bugs in it instead of holding to

the routine

of perfect quality control, because they realized

that users cared

more about the number of applications

than

they did about perfect operation.

These examples and others suggest that collective im-

provisation

often builds on and incorporates

individual

im-

provisation.

However, individual

improvisation

alone is not

sufficient for collective improvisation.

Instead,

the joint ac-

tivities of individual

people create a collective system of im-

provisational action. In addition, there are occasions in

which a person's behavior, planned or improvisational,

sparks

collective activities that are improvisational

in nature

(Burgelman

1983; Hutt,

Reingen, and Ronchetto

1988). For

example, in the case of planned

or deliberate

individual

be-

havior

creating organizational

improvisation,

Eisenhardt

and

Tabrizi

(1995) find that leaders' deliberate

behaviors

played

an important

role in speeding the development

of highly it-

erative and experiential

new product

development

(see also

Miner 1987; Quinn 1986). Other times, collective improvi-

sation results from individual

behavior that is itself highly

extemporaneous

(Mintzberg

and McHugh 1985). In either

case, the individual

improvisation

must move to the organi-

zational

level for collective improvisation

to occur.

In short,

there must be an element of collective design and

execution.

Conceptual Framework

Having established the nature

of improvisation

and its po-

tential

importance

to contemporary

organizations,

we devel-

op arguments about the incidence and effectiveness of

improvisation

in organizations.

Factors Influencing the Incidence of

Improvisation

Theory and prior research

suggest many factors that might

enhance the chances that improvisation

will occur in orga-

nizational

activity.

First,

improvisation

might occur

because

of a lack of organizational

discipline, so that

an organization

makes up new plans as it goes along simply because it lacks

the rigor to follow prior plans (Cooper and Kleinschmidt

1986; Etzioni 1964). Second, an organization

deliberately

might

encourage

spontaneous

activities that

are inconsistent

with prior plans or activities, suggesting that

it has "learned

to improvise" (Burgelman 1983; Hutt, Reingen, and

Ronchetto 1988; March 1976; Moorman

and Miner 1998).

Third, improvisation

might occur within what we call the

logic of responsiveness.

This stream of thinking suggests

that organizations

sometimes face unexpected

jolts or sur-

prises that make prior

plans irrelevant

or incomplete

in im-

portant

ways. Such jolts often are coupled with a context in

which it is difficult to refrain

from taking

action

or complete

a new planning

cycle before taking action. Weick's (1993a)

work on improvisation

by firefighters,

Preston's

(1991) ex-

ample of improvised decision making during a strike, and

Eisenhardt

and Tabrizi's

(1995) findings about improvisa-

tional styles in product development

all focus on settings in

which unexpected

stimuli create the need for organizational

action but also weaken the effectiveness of prior planning.

The central

premise

of this line of thinking is that improvi-

sation might have special value in these circumstances

(Miner 1987). We formalize the logic of this stream of

thought in the next three hypotheses, which are shown in

Figure 1.

The impact

of environmental turbulence.

When the envi-

ronment

in which an organization

operates

experiences a lot

of change, the organization

has several choices. It can ignore

external

demands

or shocks that suggest the need to change

plans and continue

with previously

planned

activities; it can

attempt

to speed up its planning

and execution cycles so that

they remain distinct but happen more quickly (Eisenhardt

and Tabrizi 1995); or it can move toward

an improvisation-

al approach

that

merges planning

and execution processes.

In some cases, fast-changing

environments

can destroy

the value of existing competencies (Tushman

and Anderson

1986). In such circumstances, organizations might find it

necessary to improvise or compose new behaviors while

executing them. As Weick (1979, p. 102) states, "If there

exists a truly novel situation,

one in which there is no anal-

ogous experience in the past, then the only thing the person

can do is act." In other words, strategy implementation

ac-

tually can be "made

up as firms go along" (Weick 1993c, p.

2). Consistent with this view, organizational

scholars have

argued

that the increased

pace of competition

might require

organizations

to develop an improvisational

competency to

prosper (Brown and Eisenhardt 1995; Eisenhardt and

Tabrizi 1995; Mintzberg and McHugh 1985). The basic

logic here is that

exogenous shocks or demands

come along

more rapidly

than an organization

can anticipate,

and orga-

nizations often respond to such situations by improvising

rather

than not responding.

We hypothesize that

Hi:

The greater

the level of environmental

turbulence,

the

greater

the incidence

of organizational

improvisation

new

product

actions.

The impact of real-time organizational information

flows. The logic of responsiveness implies that awareness

of external or internal

surprises can trigger organizational

improvisation.

Therefore, the more an organization main-

tains access to information

flows, the more likely it is to be-

come aware of either external shocks or unexpected

internal

surprises.

Some literature

points

to a particular

type of information

flow-that which flows in real-time interaction among

group members-as an important

stimulus to group impro-

visation. Bastien and Hostager (1988), for example, docu-

ment the nonverbal

cues band members

give one another

jazz improvisations.

Likewise, Spolin (1963) points to the

criticality of real-time

cues flowing among improvisational

theater

players as their scenes unfold. In addition,

real-time

information

flows between the actors and the audience not

only inform but also stimulate specific improvisational

ac-

tivities. For example, a troupe might extend an improvisa-

tional skit, spurred

by real-time

audience

reactions.

We define real-time information

flows as those that oc-

cur during or immediately prior to an action (Eisenhardt

Improvisation

in New

Product

Development

/ 5

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

FIGURE 1

Factors Influencing the Incidence and Effectiveness of Organizational Improvisation

Main

Effects

- - - -- Moderating

Effects

1989). These flows contrast to information

processes that

occur well in advance of an action or that are used after an

action to evaluate its impact. Real-time information

flows

are likely to occur in face-to-face interactions and electron-

ic communications,

in which there are few time delays and

great opportunities

for feedback (Eisenhardt

1989; Sproull

and Keisler 1991). Therefore,

team meetings in which deci-

sions and interpretations

are made and behaviors

are

carried

out are often sources of real-time information

flows (Dick-

son 1997).

Using our logic of responsiveness,

we predict

that heav-

ier real-time information flows will create

more possibilities

for organizations

to be exposed to unexpected information

that invites improvised

action. We make this prediction

for

three reasons. First, real-time information

is, by definition,

timely.

Therefore,

unlike information that

may get to a deci-

sion maker too late for action to be taken, real-time infor-

mation is inherently more actionable. Second, because it

occurs during or immediately prior to an action, real-time

information has an urgency to it that is likely to evoke im-

mediate responses, which probably cannot be planned.

Third, real-time information

flows are more novel because

they evolve in a more random

manner

than

non-real-time

in-

formation. For example, real-time information flows during

a meeting would be more likely to involve unexpected in-

formation than those emanating from a series of written

memos. These qualities

provide greater potential

for impro-

visational

activity.

We predict

that

H2:

The

greater

the level of real-time

organizational

informa-

tion flows, the greater

the incidence of organizational

improvisation

in new

product

actions.

The impact

of organizational

memory.

We have suggest-

ed that real-time information about internal

or external sur-

prises might enhance the chances that improvisation

will

occur. In contrast,

prior

work suggests that stored informa-

tion, in the form of organizational

memory,

will reduce these

chances. As with organizational improvisation, there is

some disagreement regarding whether organizations

store

information

in memory as people do. However, there is a

growing sense across disciplines that organizations have

frames of reference,

routines,

and structures

that reflect the

presence

of stored

knowledge (for a review of this literature,

see Cohen 1991; Cohen and Bacdayan 1994; Cohen and

Levinthal 1990;

Walsh 1995;

Walsh and Ungson 1991;

Win-

ter 1987). We adopt

that

perspective

in this article

but focus

on the level of knowledge contained in an organization's

memory,

which we previously have defined as collectively

held beliefs, behaviors,

or physical artifacts

(Moorman

and

Miner 1997, p. 93). Therefore,

a high level of organization-

al memory would be present

when a project

or action phase

represents

familiar

territory,

a new product requires

only a

modest change in an old project, the technological or cus-

tomer basis for the new product

is part of the firm's long-

standing

repertoire,

there are well-established team routines

because the duration of the team members' service is high,

or a particular

action phase (e.g., prototype

development)

6 1 Journal of Marketing,

July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

an established

firm-level competency

(Moorman

and Miner

1997).

By definition,

organizational

memory

represents

learned

ways of thinking and behaving.

As such, memory often is

activated automatically

in certain situations. For example,

firms tend to develop line extensions of existing products

rather

than create

completely

new ones (Andrews

and Smith

1996). Firms also tend to use well-developed routines and

processes for developing and introducing new products

(Day 1994; Leonard-Barton

1992; March 1991; Moorman

1995; Moorman and Miner 1997) and therefore learn fewer

new routines

(Sinkula 1994).

Preston (1991, p. 89) discusses the negative effect that

prior memory

is likely to have on the incidence of improvi-

sation, noting:

"In

the case of these familiar situations ... the

scope for improvisation is more constrained."

Likewise,

jazz musicians have commented

on the paradox

of needing

to learn

great

artists' works to improvise

well but then find-

ing themselves trapped by this learning

(Weick 1993c). As

Berliner

(1994, p. 206) notes, "In

one of the greatest

ironies

associated with improvisation,

as soon as artists complete

the rigorous

practice

required

to place a vocabulary pattern

into their

larger

store, they must guard

against

its habituated

and uninspired

use."

The tendency for existing knowledge to restrict the

range

of options is a common

challenge for innovation

of all

types but is an especially strong impediment

to improvisa-

tional action. In improvisation,

the time between composing

and executing is small and/or nonexistent.

We suggest that

the pressure

of fast action enhances the possibility that an

organization will rely on existing routines, regardless of

whether a learned

response is warranted.

Therefore,

we hy-

pothesize that

H3:

The

greater

the

organizational

memory,

the

lower the

inci-

dence of organizational improvisation

in new product

actions.

Factors Influencing the Effectiveness of

Improvisation

In the prior

section, we highlighted

three factors

that we pre-

dict will influence the incidence of improvisation.

The val-

ue or effectiveness of that improvisation,

once engaged in

by a firm, is another matter.

There has been a tendency to

think of improvisation

as either

helpful or hurtful to organi-

zations, as we noted previously.

We address

these equivocal

perspectives by identifying selected factors that moderate

the impact

of improvisation

and determine

whether impro-

visation benefits or hurts organizations.

We begin our dis-

cussion of these factors by identifying the new product

development

outcomes we believe might be potentially

as-

sociated with improvisation.

We then

consider

how the same

important

informational

factors

that influence the incidence

of improvisation

also might moderate its impact on these

outcomes. This dual role creates several potential

trade-offs

for organizations

to manage.

Focal new product development outcomes. There are

several outcomes typically associated with improvisation

(Brown and Eisenhardt

1995; Eisenhardt

and Tabrizi 1995;

Preston 1991; Weick 1993a, c, 1996). For example, when

improvisation works well, it can produce aesthetically

pleasing outcomes in a theatrical or musical setting (Hatch

1997). Likewise, it can provide instrumental

value for orga-

nizations or groups by solving problems or capitalizing on

unique

opportunities

(Weick 1993b, 1996).

Here, we investigate several types of outcomes that fo-

cus on organizational

effectiveness, which we define as the

degree to which an action achieves instrumental outcomes

of value for a firm (Walker

and Ruekert

1987). We focus on

two types of effectiveness outcomes: product

and process.

For product

outcomes, we investigate design effective-

ness, defined as the degree to which new product

features

are high quality and high performance,

and market effec-

tiveness, defined as the degree to which the new product

meets the demands of target customers. An organizational

example of these outcomes might be a product

development

team's improvisation

of a new casing unit for a previously

unprotected part on a commercial research

instrument.

Ef-

fective improvisation

of the unit would require

that the ma-

terials and size of the casing match each other and reflect a

quality level (design effectiveness), and that the unit fit the

customer's needs so that the product

works in the settings

for which it was designed (market

effectiveness).

Improvisation

not only influences the effectiveness of a

new product,

but, we reasoned, it also could affect the ef-

fectiveness of the new product

development process. Two

indicators

of term

process effectiveness seemed particularly

important:

teamfunctioning,

which refers to the team's com-

mitment

level to the project,

and team learning, the level of

knowledge the team gains in performing

a new product

ac-

tion. Finally, the effectiveness of new product

development

processes is also, to some degree, revealed in the cost effi-

ciency (financial

investment

level) and time efficiency (time

investment

level).3

The next three sections relate

improvisa-

tion to these product

and process outcomes.

The

moderating

impact of environmental

turbulence.

predict

that the rate of environmental

turbulence

will mod-

erate the improvisation-effectiveness

relationship,

because

it shifts the advantages

and disadvantages

of formal plan-

ning versus improvisation.

If, for example, an organization's

environment

is stable and continuous, planning in advance

of action offers that organization

many possible advantages

(e.g., Armstrong

1982; Miller and Cardinal 1994). The or-

ganization

can take the time to do complete planning,

know-

ing that the assumptions and facts guiding the plans

probably will still hold at the end of the planning cycle.

Thus, the organization

can harvest the tremendous

coordi-

nation and control benefits of good planning, including

avoiding inconsistent

and wasteful action, coordinating

ac-

tivities of multiple actors who may not communicate

easily

3Although

it is more

typical

to separate

efficiency

and

effective-

ness,

product

development

efficiency

can

influence

overall

organi-

zational

effectiveness.

Therefore,

to maintain

a common,

unifying

focus

on new

product

effectiveness,

we place

efficiency

under

the

rubric

of effectiveness

in our

mix

of dependent

variables.

Improvisation

in New

Product

Development

/ 7

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

with one another,

and assuring

that most actions are focused

on a single goal. Improvisation

offers no particular

advan-

tage in this setting. Therefore,

in the presence of low envi-

ronmental

change, improvisation will be more disruptive

than

helpful.

In contrast,

rapid

environmental turbulence

increases

the

odds that improvisation

will provide value. Extensive for-

mal plans in such conditions can have negative conse-

quences because they consume time and resources and

provide counterproductive guides to action

when the context

changes faster than the planning cycle (Eisenhardt and

Tabrizi 1995). These circumstances,

therefore,

actually im-

prove the chances that improvisation

will be effective. We

propose

the following:

H4:

The greater

the level of environmental

turbulence,

the

greater

the likelihood that improvisation

will generate

effective

(a) products

and (b) processes

in new product

development.

The moderating impact of real-time organizational in-

formation

flows. Real-time

information

flows also can facil-

itate a positive effect for improvisation by playing a

powerful coordinating

role. One of the crucial functions of

plans is to coordinate the action of multiple actors (Gal-

braith

1973). In the absence of plans, coordination

must oc-

cur through other mechanisms. Immediate information

about

the context in which the action is occurring

and the ac-

tions of other participants enables such coordination

(Bastien and Hostager 1988, 1992; Menzel 1981). In the-

atrical

improvisation,

for example, actors

continually

attend

to and process instant information

on audience reaction to

guide their

subsequent

actions (Spolin 1963). This feedback

replaces the coordinating

function of a plan because the ac-

tors respond to the same audience cues (Huber and Mc-

Daniel 1986).

In a product development context, Robins (1991) de-

scribes a thriving

company that introduces

new products

least once a month but has no formal

strategic

planning.

claims that coordination

was achieved because teams have

"an insatiable appetite for market intelligence," and each

team member

"continuously

gathers

and disseminates

anec-

dotal data from the marketplace"

(p. 336). Imai, Nonaka,

and Takeuchi

(1985, p. 358) also describe

product

develop-

ment projects marked

by a high degree of experimentation

as effective when team members

are encouraged

"to extract

as much information from the marketplace and ... to bounce

[ideas] off other members." Following this research, we

suggest that real-time information

flows can not only bring

the "news" that

prompts

improvisation

(H2),

but

also replace

the coordinating

role of a plan when actors

improvise

to the

same incoming information.

In addition to providing coordination,

real-time infor-

mation flows enable actors to learn the consequences of

their

actions as they improvise.

This immediate

information,

in turn,

enhances the chance that improvisation

will be ef-

fective because it creates learning about relevant ongoing

events (Gioia 1988; Granovetter 1973, 1985). Consistent

with this line of reasoning, Eisenhardt

and Tabrizi (1995)

find that repeated

iterations

in the product

development

cy-

cle-which, they argue, provide real-time feedback-are

important

to the success of the new product

outcomes. We

predict

that

H5:

The greater

the level of organizational

real-time

informa-

tion flows, the greater

the likelihood that

improvisation

will generate

effective

(a) products

and (b) processes

new

product

development.

The moderating impact of organizational memory.

Al-

though H3 predicts

that organizational

memory will reduce

the incidence of improvisation,

there is considerable prior

research supporting the idea that a strong organizational

memory will enhance the effectiveness of improvisational

action. First, much improvisation

appears

to arise from the

recombination of previously successful subroutines of

knowledge and action (Borko and Livingston 1989; Levi-

Strauss 1967; Nonaka 1990). For example, a firm with well-

developed marketing

research

competencies and significant

consumer insight could recombine

existing knowledge and

skills to improvise

new strategies

to respond

to unanticipat-

ed changes in consumer behavior. In support of this idea,

Weick's (1993a) analysis of the Mann Gulch disaster em-

phasizes that the ability of key members

of the fire team to

use their

existing skills in a novel way was crucial to effec-

tive improvisation. Music observers note that musicians

with strong preexisting

repertoires

of melodies, chords, and

rhythms

and familiarity

with other

players produce

the most

powerful improvisations

(Berliner 1994). Weick's (1993b,

p. 353) characterization

of organization

design as involving

improvisation

reflects the importance

of memory.

He states,

"If we think of designers

as people who improvise, then the

materials

they have available

to work

with are the residue

their past experience and the past experience of people in

their design group,

the meanings attached

to this past expe-

rience, observational

skills, and their

own willingness to re-

ly on imaginative

recombination

of these materials."

The now-famous

account

of Honda's

introduction

of the

U.S. Supercub

in the United States provides

a final example

of the importance of organizational memory. Honda's

planned

introduction

of large motorcycles

experienced

tech-

nical difficulties because nontraditional

motorcycle cus-

tomers tried to buy the smaller Supercubs

being ridden by

Honda's representatives.

The Honda

team responded

to this

demand by improvising a new strategy to sell Supercubs

through

sports stores, in contrast

to following their original

plan to focus on large motorcycles. Clearly, the effective-

ness of this improvisation

depended

on the Honda

team hav-

ing a rich repertoire of marketing, sales, financial

management,

and technical routines that could be recom-

bined into an internally

consistent strategy that linked suc-

cessfully to the changing environment (Mintzberg 1996;

Mintzberg

et al. 1996; Pascale 1996).

Although this argument

applies to many forms of inno-

vation, we believe it carries

special strength

with respect to

improvisation. In planned innovation, organizations can

gather in advance the tools needed to implement change.

They can acquire physical resources, such as machines, as

well as advice and ideas from sources outside the organiza-

tion. The extemporaneous

nature

of improvisation,

howev-

er, dictates

that

there

is little or no space between conceiving

of and executing an innovation.

Thus, the improvisation-ef-

8 / Journal

of Marketing,

July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

fectiveness relationship

will be even more dependent

on ex-

isting organizational memory.

We predict

the following:

H6: The greater

the organizational

memory,

the greater

the

likelihood that

improvisation

will generate

effective

(a)

products

and

(b) processes

in new

product

development.

Method

Setting

Data were collected from two midsized firms: FastTrack,

developer

of electronic instruments,

and SeeFoods, a manu-

facturer of food products.

These companies

offer several ad-

vantages for testing our hypotheses. First, both companies

are well established

and have formalized structures but vary

in size ($2.4 million and $2.6 billion in annual sales, re-

spectively). In addition, both companies have formalized

product development processes. Each has detailed steps

through

which the development

process

must go and

various

hurdles to meet before its products

move from step to step.

For example, SeeFoods has developed its product

planning

process to the degree that it considers the process a distinct

competency and treats it as a trade secret. FastTrack

suc-

cessfully has achieved IS09000 certification of its product

development processes, which indicates

some degree of for-

malization. These properties

make the firms good settings

for examining improvisation.

The firms also provide rich

settings for testing our hypotheses because each company

takes product

development seriously and links it to overall

organizational success. They therefore engage in various

product development activities, which provided us with

many opportunities

to observe the development process.

The two firms do contrast

on two dimensions, therefore

providing

some variation

in our study conditions. First,

one

is a consumer packaged-goods

firm, and the other a tech-

nology-oriented industrial firm. Second, in one firm we

studied, the project

was in the concept and prototype

devel-

opment stages, whereas in the other

firm, the project

was in

the market

development and product

introduction

stages of

the new product

development process. Both aspects of dif-

ferentiation improve the generalizability

of our results. In

both firms, employees understood

we were conducting a

study of product

development,

but no mention was made of

improvisation

to avoid demand effects about

its incidence

impact. In each firm, we focused on one development pro-

ject, which was selected because it was representative

other

projects

in each firm.

The teams

working

on these pro-

jects both were cross-functional,

with ten active members

SeeFoods and seven at FastTrack.

Data Collection Procedures

Investigators

attended,

recorded,

and transcribed

the meet-

ings of the product

development

teams during

a nine-month

period. Meetings generally

were held once a week. Howev-

er, holidays, vacations, and schedule conflicts meant some

meetings were missed, which resulted in approximately

25-30 meetings attended

in each of the firms. To generate

fine-grained

but systematic

data

at the level of specific team

actions, we identified

action events from these meetings and

asked key informants

to evaluate

them

in terms

of a wide va-

riety of variables,

including activities that occurred

prior

or at the same time as the event and outcomes associated

with the action event.

Determining

the scope and nature of these action events

was, therefore,

an important part

of the research

effort. New

product

development projects are composed of a series of

ongoing activities that are, in some sense, seamless. There-

fore, to understand

and evaluate improvisation,

we had to

both appreciate

this larger

unfolding

of activities and decide

how to divide it into a series of events that could be evalu-

ated and

judged independently.

Action events are defined as discrete activities under-

taken by a new product development unit. Following from

our conceptualization, we restricted

our sample to organi-

zational-level actions, defined as those undertaken

by the

new product team. Therefore, actions taken outside of the

group or without the group's approval were not consid-

ered organizational for our purposes. An action event

could involve a team engaging in any of the following ac-

tivities: making changes to a new product, calling a sup-

plier to change the size of a part,

making a decision to use

a new distributor,

releasing test procedures, finding prod-

uct problems, preparing documents for regulation filing,

doing a store walk, creating concept boards, making tar-

geting decisions, generating brand names, deciding to de-

lay a project, engaging in a focus group briefing, or

participating

in a focus group or a creativity session. One

additional condition was placed on the action identifica-

tion process: The action had to have some possibility

(even remote) of influencing new product outcomes. This

condition did not bias our sample toward actions that were

likely to influence outcomes. Instead, it eliminated minor

activities of the team, such as decisions about when and

where to meet and other conversation that was related to

social activities among team members and not the prod-

uct.

To ensure that our identification of the actions was

complete and unbiased, we used several safeguards. First,

four transcripts

were selected randomly

from each site, and

action events were coded independently

by two investiga-

tors familiar with the site. This approach

yielded high in-

terjudge reliability (92% agreement). Second, product

development team meetings evolved as primary data

sources from which we derived action events because they

represented

a consistent vehicle that brought

all the project

members together and because the meetings were the pri-

mary means by which members exchanged information

about

actions that influenced the project.

To reduce

the pos-

sible bias in team meetings, project teams were asked on

three

different

occasions to list all the things they had done

on behalf of a project during the preceding week. These

lists were compared

with meeting transcripts,

and coverage

was adequate

(90% coverage).

Each of the team meetings we observed

could potential-

ly produce

dozens of actions, making

their

evaluation

by re-

spondents difficult and burdensome.

Therefore, to reduce

the set of actions that would constitute

our sample, we ran-

domly selected 2 organizational

actions from those coded in

Improvisation

in New

Product

Development

/ 9

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

a given meeting.4

This resulted in the selection of 107 action

events for our sample. When an action event was selected

for informant

evaluation, we set in motion a series of data

collection events. To begin, informants

completed

an action

assessment

form, within a week of the action

occurring,

that

asked them to rate the action and various environmental and

organizational

activities that occurred

prior to and during

the time the action was unfolding. This included an infor-

mant rating of organizational

improvisation,

memory, and

real-time information flows and an assessment of the level

of environmental turbulence associated with the action

event. After approximately

four weeks, the same informants

evaluated the short-term impact of each action event in

terms of product

and process outcomes. The key informants

in completing both forms were the team leaders because

they were in attendance at every meeting, were aware of

team actions, and had a broad

view of the project.

Measures

In this section, we report

the properties

of our

measures,

ap-

proximately

half of which were multi-item and half single-

item measures. Multi-item measures were a mix of

formative and reflective indicators. This mix was adopted

safeguard against the hazards of key informant burnout.

Each of our key informants

completed approximately

100

40ur initial

goal was to sample systematically high

and

low im-

provisation

actions to ensure

that

our

sample

would contain

vari-

ance.

However,

after

attempting

to do this for several

weeks,

abandoned the

approach

for two reasons.

First,

from our

observa-

tions of team meetings

and interviews with key informants,

seemed that much of what occurred

during meetings

was,

at least

in part,

improvisation.

Therefore,

variance

would be

easier than

we had

expected

to capture.

Second,

we determined

it was more

appropriate

for our

informants

to assess

the

degree

to which

they

improvised

than

for us to make such

judgments

without

a clear un-

derstanding,

a priori,

of how

much or how little

planning

had

pre-

ceded an action. This realization

shifted our design, and we

allowed

informants

to rate the

degree

of improvisation

in an action

following

its enactment.

questionnaires:

50 action assessments immediately

follow-

ing the action and 50 short-term

impact forms four weeks

after the action. Therefore, they completed two question-

naires each week for the study. Given this workload, we

tried to make each questionnaire

no more than a page long.

This meant sacrificing some depth on individual measures.

We focused our efforts on developing multi-item

measures

for those constructs

that

were considered a priori

difficult to

measure,

such as improvisation.

Because the composition of formative measures is dri-

ven by conceptual criteria-which is coverage of the con-

struct domain-and not by predictions of correlation

between items in that space, formative measures were not

subject to reliability

or factor analytic approaches (Bagozzi

1994). The remaining reflective measures were examined

for unidimensionality

and reliability.

In Table

2, we present

an overview of the psychometrics associated with each

measure.

Appendix B contains a complete listing of all the mea-

sures used in this study. The organizational

improvisation

level of each action event was evaluated on the action

assessment form and was measured on three semantic

differential

seven-point scales with the following anchors:

(1) figured out action as we went along/action followed

a strict plan as it was taken, (2) improvised

in carrying

out

this action/strictly

followed our plan in carrying out this

action, and (3) ad-libbed action/not an ad-libbed action.

The mean improvisation

level is M = 4.252 (s.d. = 1.985)

and the coefficient alpha exceeded acceptable standards

(a = .79).

Because of the centrality

of improvisation

to our work,

two additional

safeguards

were taken to measure it. First, it

was rated

by two investigators

involved in the site, and there

was 70% agreement regarding

whether improvisation

was

low (1), moderate (4), or high (7). The mean investigator

rating

(M = 4.014, s.d. = 1.539) also compared

well with the

mean informant

rating.

Second, because we claim that im-

provisation is distinct from innovation, we examined the

discriminant

validity of the level of improvisation

and the

TABLE 2

Measurement Information

Standard

Mean Deviation (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

(1) Organizational

improvisation 4.25 1.98 .83

(2) Environmental turbulence 2.12 1.22 .18 -

(3) Organizational

real-time

information flows 5.86 1.85 .03 -.14 -b

(4) Organizational

memory 4.36 1.78 -.60 -.07 .04 .84

(5) Design effectiveness 5.03 1.06 -.08 -.23 .30 .07 a

(6) Market effectiveness 4.87 .81 -.22 -.12 .30 .29 .65 .70

(7) Cost efficiency 4.05 .76 -.08 -.20 .06 .08 .19 .11 b

(8) Time efficiency 4.08 .89 .01 -.14 .20 .03 .05 .03 .35 .89

(9) Team

functioning 4.61 .80 -.34 -.12 .20 .17 .43 .60 .11 .16 .86

(10) Team learning 5.12 .93 -.30 -.24 .29 .25 .53 .64 .17 .13 .60 .73

aFormative

scale,

therefore

alpha

is not

reported.

bSingle-item

measure,

therefore

no alpha

or correlation

is reported.

Notes:

The coefficient

alpha

for

each

measure

is on the

diagonal

and the intercorrelations

among

the measures are

on the

off-diagonal.

Corre-

lation coefficients

greater

than

.20 are

significant,

p < .05.

101 Journal of Marketing,

July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

level of innovation of the focal action. Informants

rated

in-

novation and improvisation

on the same form. Innovative-

ness was measured by asking key informants to rate the

level of innovation

on a two-item semantic differential scale

with (1) innovative

action/ordinary

action and (2) novel ac-

tion/standard

action as the anchors. The two innovativeness

items correlated well (p = .70).

The test of discriminant

validity

between innovation and

improvisation

required

constraining

and freeing the phi co-

efficient between the two measures using LISREL 8. The

model with the free coefficient was found to be superior

the fixed coefficient (AX2i)

= 5.31), exceeding the standard

necessary to show discriminant validity (AX2(l)

= 3.84).

Therefore, innovation and improvisation are empirically

distinct.5

The three

explanatory

variables

were measured on sev-

en-point

Likert

scales in which the anchors

depended

on the

variable. The degree of environmental

turbulence

occurring

around

the action event was measured using a three-item,

formative

measure, where I was "none" and 7 was "a lot"

for the following items: When the action was taken, how

would you rate

the level of change [defined

as any deviation

from the status

quo] within (1) your team, (2) your

firm, and

(3) external

sources [customers,

suppliers,

distributors].

The

mean level of external

change was 2.136 (s.d. = 1.243). Be-

cause the measure is a formative indicator,

rather than re-

flective, we do not report

a reliability

coefficient.

Informants

also evaluated the level of real-time

organi-

zational information flows that occurred

using a single-item

measure

that asked for a rating

of the level of face-to-face,

telephone, or e-mail information transferred

among team

members just before the focal action (M = 4.078, s.d. =

1.058), using the same anchors.

Organizational memory level was measured

by asking

informants to evaluate the memory

level regarding

an action

using a four-item scale adapted

from our previous (1997)

study. Informants were asked to rate the extent of their

agreement with the following items: For this action, my

team has (1) well-defined procedures, (2) a standard

ap-

proach,

(3) a great deal of knowledge, and (4) strong

skills.

The measure had a mean of 4.392 (s.d. = 1.784) and ade-

quate reliability

(a = .79).

Of all the independent

variables, organizational

impro-

visation and memory are the only measures that are multi-

item and reflective in nature.

Therefore, we examined the

discriminant

validity of organizational

memory

and impro-

visation. As previously, the test of discriminant

validity in-

volved constraining and freeing the phi coefficient

between the two measures using LISREL 8. The model

with the free coefficient again was found to be far superi-

5We

ideally

would

have

performed

discriminant

validity

checks

between all the variables

from

which

we previously

conceptually

distinguished

improvisation

(including

adaptation,

learning,

and

opportunism).

However,

because

of the severe

constraints

imposed

on our

informants,

we chose

to focus

our

empirical

efforts on in-

novation.

We believe

it is closest

conceptually

to improvisation

and,

therefore,

perhaps

the

competing

variable

most

crucial

to our

perspective.

or to the fixed coefficient (AX2(I)

= 52.00), which indicat-

ed discriminant validity between organizational memory

and improvisation.

Dependent variables. These variables were assessed by

our informants four weeks after the action event occurred.

Informants were given a one-page survey with the action

event described in detail at the top of the page. In addition,

because of the time lag, a portion

of the transcript

relevant

to the action generally

was attached to the page to jar the in-

formant's

memory

of the event. Informants did not have ac-

cess to their

original

evaluations of the event or their

ratings

of the informational actions occurring

around it. All depen-

dent variables were evaluated

on a seven-point scale, where

7 was a positive and I was a negative effect of the action on

the particular

outcome. Informants

were asked to rate how,

"on balance,"

the action event has or is likely to have influ-

enced each dependent

variable.

They appeared

to have no

problems

in making such assessments.

Following our conceptualization,

two product-effective-

ness dependent

measures

were used in this research.

Design

effectiveness was measured

with a two-item formative scale

that assessed the impact of the action on the performance

and design of the product.

The two items have an acceptable

correlation (p = .48). Influenced by Griffin and Page's

(1993, 1996) work, we measured

market effectiveness with

a three-item scale that described the impact

of the action on

the sales, customer

acceptance,

and success of the new prod-

uct. The items have an alpha of .70.

Four process effectiveness dependent variables also

were evaluated. Cost efficiency was measured

by a single-

item measure of the estimated cost structure

of the new

product

(Griffin

and Page 1993, 1996). Time efficiency was

measured

by a three-item

measure that asked respondents

rate the (1) length of the product

development process, (2)

speed of the product

development process, and (3) project

timeliness. These items are reliable (a = .89).

Team

functioning

refers

to the impact

of the action event

on the degree to which the team works well together.

This

was evaluated using a three-item

measure

that asked infor-

mants to assess the impact of the action on (1) team com-

mitment level, (2) team functioning, and (3) team

enthusiasm.

These items are reliable

(a = .86). Finally,

team

learning was measured

by asking informants

to assess the

impact

of the action

on (1) the way the team thinks

about

the

project,

(2) the team's certainty

level, (3) the team's under-

standing level, and (4) how much the team learned.

These

four items are reliable (a = .73).

Three of the process effectiveness outcomes (time effi-

ciency, team functioning,

and team learning)

are multi-item

and reflective in nature.

Therefore, measure development

required

examining their discriminant

validity. As we did

previously,

we established

a base model that did not reflect

the correlation

between measures.

Then we examined how

model fit changed when we constrained

the phi coefficient

between different pairs of the three measures

to equal one.

Results indicate that the model with the free phi coefficient

was a better

fit in all three cases, which indicated

discrimi-

nant validity in time efficiency and team learning (AX2(I)

16.53), time efficiency and team functioning (AX2(I)

Improvisation

in New

Product

Development

/11

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

28.17), and team learning and team functioning (AX2() =

8.78).

Model-Testing Approaches

We used three

distinct model testing approaches

to examine

the proposed hypotheses. First, we used simple descriptive

statistics to examine the existence of improvisation

in the

new product

actions we sampled. Second, we used a multi-

variate linear regression model to examine the impact of

environmental

turbulence,

organizational

real-time informa-

tion flows, and organizational

memory on the incidence of

improvisation

in new product

actions. Third,

we performed

a split group analysis (Arnold 1982; Cohen and Cohen

1983) to examine the impact

of the three moderators-envi-

ronmental

turbulence,

organizational

real-time information

flows, and organizational memory-on the organizational

improvisation-effectiveness

relationships.

This approach

in-

volved creating high and low levels of each moderator vari-

able by performing

a median split. We then examined the

relationship between improvisation and the various out-

comes in the high and low moderator variable conditions

and compared

the regression

results from these two condi-

tions using a t-test, to determine if the regression coeffi-

cients were different across the two moderator conditions

(Pedhazur

1982). If the t-test of differences in the beta coef-

ficients was significant, we had found evidence of modera-

tion and inspected the direction of moderation.

We chose

this approach over moderator

regression analysis (MRA)

because MRA demands

that all main and interaction

effects

associated with the proposed

moderating

influences be en-

tered into the model (Pedhazur

1982).6

In this study,

MRA

would have involved seven predictors

(i.e., improvisation,

memory, real-time information,

environmental

turbulence,

improvisation

x memory, improvisation

x real-time infor-

mation,

and improvisation

x environmental

turbulence).

Us-

ing MRA with the number

of variables and the sample size

of this study likely would have resulted in underpowered

tests of the hypotheses.

Results

The Incidence of Improvisation

Simple descriptive statistics suggest that

organizational

im-

provisation

occurred

in our sample of new product

actions.

The mean level of improvisation,

as rated by informants,

was 4.252 on a seven-point Likert scale (s.d. = 1.985),

where 7 represents

greater

improvisation.

The scale exhib-

ited considerable

range,

running

from I to 7 with a mode of

5 and a median of 4.667. The distribution

is fairly even

across all levels of improvisation

but is skewed slightly to-

ward

higher levels of improvisation.

For example, if we de-

fined as "primarily improvisational"

those actions that were

rated higher than five, 47.5% of the actions would qualify;

if we used a cutoff of higher than six, 24.1% would quali-

6Despite

differences

at the multivariate

level,

a univariate

MRA

test is identical

in structure

to the split-group

analysis

used here

(for

proofs,

see Arnold

1982;

Cohen

and Cohen

1983).

fy. The central tendency then was toward improvisation.

However, with a standard

deviation of 1.98, there was also

quite a bit of variance in improvised behavior. In Table 3,

Part

A, we depict the frequency distribution of improvisa-

tion in our sample.

Factors Influencing the Incidence of

Improvisation

Considering the factors that could influence the incidence

of improvisation, we tested the first three hypotheses in a

single multivariate regression model. The results, which

appear

in Table 3, Part

B, suggest that the overall model is

significant (Adjusted R2 = .39, F(388)

= 20.97, p < .001).

Results also indicate that environmental turbulence is a

marginally significant, positive predictor of the level of

improvisation in new product actions (b = .252, t = 1.88,

p < .10, two-tailed test), in support of H1. Organizational

real-time information flows do not have a significant im-

pact on the incidence of organizational improvisation

(b =

.100, t = 1.14, p > .10), thus failing to support H2. Finally,

organizational

memory has a negative effect on the level

of improvisation (b = -.687, t = -7.520, p < .001), in sup-

port of H3.

Factors Influencing the Effectiveness of

Improvisation

The split group analyses results appear

in Table 3, Part C.

We note there that high levels of environmental

turbulence

have a positive influence on improvisation's impact on de-

sign effectiveness (t(86)

= 2.83, p < .05). We find that when

environmental turbulence

is low, improvisation

has a nega-

tive effect on design effectiveness, but in the presence of

high environmental

turbulence,

improvisation

improves

de-

sign effectiveness. Environmental

turbulence

does not have,

however, a statistically significant moderating impact on

market effectiveness (t(86) = 1.21, p > .10). These results

thereby support H4a with respect to technical design and

quality

outcomes but not in terms of the product's

effective-

ness using market indicators.

Environmental

turbulence

has equally mixed effects on

the improvisation-process outcome relationships. Turbu-

lence improves the extent to which the team reports it

learned (t(86)= 3.42, p < .05) and functioned smoothly

(t(86)

= 1.98, p < .05) while taking improvisational

actions,

in support of H4b.

However, the improvisation-cost effi-

ciency relationship becomes weaker and more negative

when turbulence

is high, and the improvisation-time effi-

ciency relationship is not influenced at all, thus failing to

support H4b.

These results suggest important

trade-offs for

the use of improvisation in conditions of environmental

turbulence.

Organizational

real-time information

flows have a more

uniform positive influence on the extent to which impro-

vised new product

actions influence design (t(94)

= 6.21, p <

.05) and market (t(94)

= 4.76, p < .05) effectiveness, in sup-

port of H5a.

However, real-time information

flows do not

have the same positive influence on process outcomes, thus

failing to support

H5b.

Improvised

new product

actions do

not have a greater impact on cost efficiency (t(94)

= .66, p >

12 / Journal of Marketing, July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

TABLE 3

Tests of Hypothesized Relationships

A.The Incidence of Organizational Improvisation

Mean: 4.25

Standard Deviation: 1.98

Mode: 5.00

Median: 4.67

1.00

1.01-1.99

2.00-2.99

3.00-3.99

4.00-4.99

5.00-5.99

6.00-6.99

7.00

Frequency Percentage

14 13.1%

4 3.8

7 6.5

18 16.8

13 12.1

25 23.4

9 8.2

17 15.9

107 100 %

B. Factors Influencing the Incidence of Organizational Improvisation

Overall model, F(3,88)

= 20.97, p < .001, Adjusted R2 = .39

Independent Variables

Environmental Real-Time Information Organizational

Turbulence Level Flow Level Memory Level

Dependent Variable b t-value b t-value b t-value

Improvisation .252 (1.88**) .100 (1.14) -.687 (-7.52*)

C. Factors Influencing the Effectiveness of Organizational Improvisation

Moderating

Variables

Environmental Real-Time Information Organizational

Turbulence Level Flow Level Memory Level

(n = 86) (n = 95) (n = 94)

Dependent Variables Lowa High t-value Low High t-value Low High t-value

Product effectiveness

Design effectiveness -.17 .17 2.83* -.57 -.01 6.21* -.13 .08 1.65**

Market

effectiveness -.27 -.16 1.21 -.54 -.18 4.76* -.13 -.06 .64

Process effectiveness

Cost efficiency .02 -.22 -2.91* -.14 .09 .66 -.18 .09 2.83*

Time efficiency .07 .09 .16 .09 .02 -1.46 -.18 .17 3.15*

Team

functioning -.36 -.20 1.98* -.40 -.39 .10 -.41 -.23 1.98*

Team learning -.35 .00 3.42* -.59 -.28 3.81* -.32 -.11 1.98*

aNumbers

the

low and

high

columns

represent

the

standardized

beta coefficient

(b)

for

the impact

of improvisation

on each

dependent

vari-

able under

low and

high

moderating

conditions.

< .05.

**p<

.10.

.10), time efficiency (t(94)

= -1.46, p > .10), or team func-

tioning (t(94)= .10, p > .10) when real-time information

flows are

high rather than

low (see Table

3, Part

C). Only the

impact of improvisation

on team learning improves when

real-time

information

flows are high (t(94)

= 3.81, p < .05).

Finally, H6 predicts that high levels of organizational

memory

will increase

the likelihood that improvisation

will

generate effective products and processes in new product

development.

The results indicate that

organizational

mem-

ory uniformly improves

the impact

of improvisation

on var-

ious process outcomes, including cost efficiency (t(94)=

2.83, p < .05), time efficiency (t(94)= 1.98, p < .05), team

functioning (t(94)

= 3.15, p < .05), and team learning (t(94)

1.98, p < .05). These results support

H6b.

Likewise, organi-

zational memory marginally

improves the extent to which

improvised new product

actions result in design effective-

ness (t(94)

= 1.65, p < .10). However, memory does not im-

prove the likelihood that improvised new product actions

will result in market

effectiveness (t(94)

= .64, p > .10), thus

providing

mixed support

for H6a.

Improvisation

in New

Product

Development

113

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

Discussion and Implications

Our conceptual work builds on prior interdisciplinary

re-

search and

suggests that

improvisation

can play a role in the

new product

development process.

An investigation

of mar-

keting literature

provides

few positive empirical

accounts

extemporaneous

action in managerial

action, suggesting in-

stead that

more fruitful action is planned

and then

executed.

In an attempt

to address this gap in the literature,

we have

documented

the incidence of improvisation

and the factors

that influence that incidence. Drawing on equivocalities in

literature,

we further

suggest that improvisation

is under-

stood best as perhaps

having

both positive and negative

out-

comes for firms. This mixed assessment draws us to try to

understand the conditions in which improvisation

might be

deployed effectively by organizations.

We propose and test

several such conditions.

In this section, we address the theoretical

and practical

potential

of our view of improvisation

in several

ways. First,

we discuss the limitations of our work. Second, we review

the pattern

of our results in more detail and with an eye to-

ward understanding

the conditions in which the moderators

(I) change improvisation's

effect from negative to positive

or (2) reduce the negative impact of improvisation.

Third,

we discuss several trade-offs

in managing

the incidence

and

effectiveness of improvisation

in organizations

and high-

light the implications

of these trade-offs.

Limitations

Our research

has several limitations.

First,

despite the longi-

tudinal

approach

we adopted,

the use of two firms limits the

generalizability

of our results.

Similar to Hutt,

Reingen,

and

Ronchetto

(1988), we initially sought

to control

for many of

the firm factors that might influence either the rate of im-

provisation

or its impact

by limiting ourselves to two firms.

Furthermore,

the challenges of longitudinal

access to orga-

nizations, especially to proprietary

activities such as new

product development processes, made our method choice

more reasonable. This is particularly

true given that we se-

lected actions from among those that occurred

during

week-

ly product development meetings, all of which were

attended, recorded,

and transcribed.

Other approaches

that

may be more

externally

valid might

have created

other

prob-

lems in generating an unbiased sample of actions. There-

fore, our approach

offers solid, internally

valid evidence of

improvisation

that further

research

might examine in more

firms using a less sensitive methodology.

Second, our hypotheses focus on three information

fac-

tors that

have been discussed in prior

literature

on improvi-

sation and that deserve empirical attention.

Although still

limited, these factors

include the effect of different

informa-

tion sources (internal and external) and types (stocks and

flows) on improvisation. Further

research could involve

considering a more comprehensive

study of improvisation

that extends the connections between information

and im-

provisation presented

here. This approach

could develop a

more general framework of the antecedents and conse-

quences of improvisation

in new product

development, us-

ing relevant industry, firm, product, environmental, and

individual team factors. Many of the factors examined by

Hutt, Reingen, and Ronchetto

(1988) would be fruitful av-

enues for such a framework. For example, the role of cul-

ture, structure,

boundary

spanners,

and product

champions

as catalysts for and facilitators of improvisation

would be

appropriate

factors. Additional research also could involve

examining the role of individual

improvisation

in organiza-

tional improvisation,

a factor we did not investigate.

Third,

though this research involved examining the im-

pact of improvisation

on short-term

new product

effective-

ness, it also would be valuable to examine the impact of

improvisation

on long-term organizational

outcomes. Our

research also was limited to new product development ac-

tions. However,

many other

marketing

contexts are also rel-

evant contexts for the study of improvisation.

Advertising

and personal selling stand out as areas in which we would

expect improvisation

to occur at even higher levels. Addi-

tional research could examine the incidence of improvisa-

tion and attempt

to demonstrate

the generalizability

of our

findings across multiple

contexts.

Finally, further research could address the possibility of

common methods variance

influencing our results, because

the same informants

rated

aspects of both actions and out-

comes. However, because observers independently identi-

fied the action, the ratings of actions and outcomes were

accomplished at significantly different times, and infor-

mants had no record of prior

ratings

when performing

their

outcome ratings, the chance of informant preconceptions

producing

the results

here is reduced.

Further

research

could

use multiple measures of actions and outcomes to ensure

even further

the lack of common methods

or informant

bias.

Pattern of Improvisation Results

Considerable

research on organizations

suggests that for-

malized organizations

with well-developed product devel-

opment procedures are relatively unlikely to engage in

improvisation

(Scott 1987). Contrary

to that research, our

first basic finding is that,

even in two well-established

orga-

nizations with formal structures,

roles, and procedures,

im-

provisation

occurs with substantial

regularity

in the product

development process.

Prior research also has tended to highlight either the

dangers of improvisation

or its potential for helping firms

adapt. Our pattern of results supports a more contingent

view of improvisation.

For example, we provide some sup-

port

for the traditional

concern

regarding

the risks of impro-

visation because, in five of the specific outcome variables

we observed,

our moderating

variables

had a positive effect

but

worked

by reducing

the negative

effect of improvisation.

Thus, though

the moderating

conditions enhanced the value

of improvisation,

as we had expected, the conditions were

not strong enough to make the net effect of improvisation

positive.

In other cases, however, the moderating

conditions re-

versed the negative impact of improvisation.

For example,

in conditions with low organizational

memory, improvisa-

tion had a negative effect on design effectiveness, cost effi-

ciency, and time efficiency. However, in the presence of

high organizational

memory, improvisation

had a positive

effect on these outcomes. These findings support

the gener-

al argument

that

emergent processes might have value in un-

certain or ambiguous conditions (Burgelman 1983; Miner

14/ Journal of Marketing,

July 1998

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

1987). They also support the more recent arguments of

scholars

who claim that improvisation

represents

an impor-

tant

competency that

can produce

value for organizations

certain conditions (Eisenhardt

and Tabrizi 1995; Moorman

and Miner 1998; Weick 1987).

Trade-Offs in Improvisation

A common theme that runs through

our conceptual frame-

work and results is that

improvisation,

similar to most strate-

gic actions, involves trade-offs and potential synergies for

marketing organizations.

This theme is evidenced in several

ways.

Recall that our findings show that memory reduces the

likelihood of improvisation,

but it also increases the effec-

tiveness of improvisation

when improvisation

does occur.

Therefore,

the same organizational

feature that makes im-

provisation

effective is likely to reduce

the chances

of its oc-

currence. Too powerful a memory, then, can remove

improvisation

from the organization's

repertoire,

whereas

too little memory can render the improvisation

that occurs

ineffective. This result suggests that there is a threshold

memory level at which improvisation

is a valuable organi-

zational activity, with levels below or above this threshold

reducing

the chances of such an impact.

This trade-off

im-

plies that organizations

must minimize the fixating aspects

of memory

when improvisation

is needed and

evoke memo-

ry as improvisation

is unfolding if it is to be effective. This

delicate balance of restraining

and infusing memory at cer-

tain times requires a greater understanding

of memory, its

forms, and the degree to which these forms restrict extem-

poraneous

actions in organizations.

For example, in a forth-

coming article (1998), we distinguish

between the effects of

declarative organizational

memory (facts and theories) on

the novelty of organizational

outcomes versus the effects of

procedural

organizational

memory (skills and routines) on

the timeliness of organizational

outcomes.

Another trade-off associated with memory that is evi-

dent in our results is that organizational

memory facilitates

the impact

of improvisation

on all new product

and process

outcomes, except those associated with external

market ef-

fectiveness. Although this is consistent with research that

suggests an internal

firm focus should reduce market

suc-

cess (Day and Nedungadi 1994; Deshpande, Farley, and

Webster

1993), none of our memory

measures

involve skills

and knowledge regarding

how a new product activity fits

with customer needs and wants. Therefore, the value of

memory appears

to be linked tightly

to its measured

content.

Real-time information

flows moving through

organiza-

tions present

a different

set of trade-offs

for firms.

Real-time

flows were found to increase

the extent to which improvisa-

tion produces

effective new products

but to reduce

the pos-

itive effect of improvisation on process outcomes

(excluding learning).

Improvisations

were less cost and time

efficient, and groups

engaging in them appeared

to function

less effectively when the level of real-time information

transfer

was high. Despite the inefficiencies associated

with

real-time information flows, improvisation continued to

promote design and market effectiveness when real-time

flows were high. Therefore,

similar to our prior

discussion

of memory, our results appear to recommend a restricted

zone of real-time information flow in which sufficient

amounts of real-time information

are needed to promote a

positive relationship

between improvisation

and product

ef-

fectiveness. However, flows cannot be so high as to create

negative improvisation-process

effectiveness relationships,

which, over time, might undermine

product effectiveness

levels.

Finally, as with the other informational

moderators,

the

influx of information about environmental

change brings

with it certain trade-offs. In particular,

our results suggest

that

high levels of information about

environmental

changes

during improvisation

result in increased product

design ef-

fectiveness. Across all of the informational moderators,

product

design effectiveness has the most to gain from high-

ly improvisational

actions when levels of information are

high. However, consistent with the other

results, firms have

to accept that

the influx of the high levels of environmental

change information

might have corresponding

risks, partic-

ularly higher costs. Such trade-offs

also appear

in other re-

search on new products

(Griffin and Page 1993; Moorman

1995).

In summary,

these results suggest that

improvisation

is a

strategy

of emergent

learning

(Mintzberg

1996) that can be

employed as a substitute

for planning

(Weick 1987). How-

ever, our results clearly suggest that improvisation

is not

necessarily a free good, nor is it one that translates

into ef-

fective outcomes in all conditions. On the contrary,

our re-

sults consistently emphasize that improvisation must be

directed

explicitly, its trade-offs

and tensions acknowledged

and managed,

and the conditions in which it is effective un-

derstood

and nurtured

by organizations.

Future Research Issues

There are many issues that our initial inquiry

into improvi-

sation did not consider.

We discuss several here as a way of

establishing

a strategy

for additional

research

on this topic.

Regarding

improvisation

generally,

we recommend

that fur-

ther research

consider whether improvisation

is driven by

firm mismanagement,

environmental

change, or the deci-

sion to use improvisation

purposively

as part

of firm strate-

gy. In the domain of product development, we encourage

investigation of improvisation's

occurrence and impact in

different (1) project

phases, (2) product

categories, and (3)

industries. Whether a product development project repre-

sents an incremental

or a radical

change from a prior

prod-

uct is an especially important

contingent worthy of further

research.

The risks for improvisation

intuitively seem high-

er in radical product

development projects because of the

probable

lack of relevant organizational

memory to inform

the product

development

process. At the same time, impro-

visation may be more likely to occur in radical

product

de-

velopment

projects

that lack memory.

Teasing

out these and

other possibilities is an important

next step.

Another topic we did not address is the nature of the

group or individual

factors that

spawn improvisation

during

the new product

development process. The research

tradi-

tion on product

championing

includes

projects

that

are initi-

ated outside the formal new product

development process

(Burgelman 1983; Hutt, Reingen, and Ronchetto 1988). It

would be interesting

to consider what

motivates these infor-

Improvisation

in New

Product

Development

/15

This content downloaded from 152.3.153.148 on Mon, 23 Sep 2013 16:38:04 PM

All use subject to JSTOR Terms and Conditions

mal improvisation

efforts. What form does improvisation

take when the focus is on a major

strategic

decision that

in-

volves multiple

stakeholders across business units?

The fo-

cus of the present

study is restricted

to improvisation

in the

narrow confines of a structured

development process.

Our focus was the impact of improvisational

activities

on the project

in which they occurred.

However, it is possi-

ble that improvisations have long-term impacts as well.

Team

members sometimes observed

that a particular

impro-

visational action not only worked in the current

project

but

also could be used in other settings or future

projects.

Our

quantitative

results support

such an impact

and suggest that

improvisation

positively affected team learning

outcomes in

high information conditions. Additional research could

fruitfully

investigate

whether

improvisation

serves as a sys-

tematic form of unplanned experimentation

in organiza-

tions. If this function is confirmed, the potential

"second-order"

impact creates an additional factor in the

calculus of improvisation's

value to organizations.

Each im-

provisation might

have, on average,

a low expected value as

a possible new routine for the organization.

But on rare

oc-

casions, an improvisational act (or "local experiment")

might represent

a real improvement

over prior

practices

and

thus be a very useful experiment (Miner, Moorman, and

Bassoff 1997).

Finally, our qualitative

observations

lead us to suspect

that, in addition

to building

a baseline model of factors

that

can move improvisation

from a hindrance to a potential

ad-

vantage,

contemporary

researchers

should

entertain

the pos-

sibility that the boundary conditions for organizational

improvisation

mightbe changing.

Corporate

intranets,

com-

puter-aided

design, and manufacturing

and point-of-sale

da-

ta can change the temporal

links between actions in ways

that previously were not possible. In many ways, this

change enhances the potential for accurate

planning;

how-

ever, it also might enhance

the possibility

of fusing planning

and

acting.

Therefore,

it appears

that

improvisation's

bound-

ary

conditions

are

The Convergence of Planning and Execution: Improvisation in New Product Development

Abstract