ArticlePDF Available

Strategy Making in Novel and Complex Worlds: The Power of Analogy

August 2005
Strategic Management Journal 26(8):691 - 712

DOI:10.1002/smj.475

Authors:

Giovanni Gavetti

Dartmouth College

Daniel A. Levinthal

University of Pennsylvania

We examine how firms discover effective competitive positions in worlds that are both novel and complex. In such settings, neither rational deduction nor local search is likely to lead a firm to a successful array of choices. Analogical reasoning, however, may be helpful, allowing managers to transfer useful wisdom from similar settings they have experienced in the past. From a long list of observable industry characteristics, analogizing managers choose a subset they believe distinguishes similar industries from different ones. Faced with a novel industry, they seek a familiar industry which matches the novel one along that subset of characteristics. They transfer from the matching industry high-level policies that guide search in the novel industry.

Firm performance over time

…

Performance differential due to orthodoxy

…

. Effect of policy breadth on analogizer's long- run performance advantage

…

Figures - uploaded by Daniel A. Levinthal

Content may be subject to copyright.

Content uploaded by Daniel A. Levinthal

Content may be subject to copyright.

Strategic Management Journal

Strat. Mgmt. J.,26: 691– 712 (2005)

Published online 7 June 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/smj.475

STRATEGY MAKING IN NOVEL AND COMPLEX

WORLDS: THE POWER OF ANALOGY

GIOVANNI GAVETTI,1DANIEL A. LEVINTHAL2* and JAN W. RIVKIN1

Harvard Business School, Boston, Massachusetts, U.S.A.

The Wharton School, Philadelphia, Pennsylvania, U.S.A.

We examine how ﬁrms discover effective competitive positions in worlds that are both novel and

complex. In such settings, neither rational deduction nor local search is likely to lead a ﬁrm to a

successful array of choices. Analogical reasoning, however, may be helpful, allowing managers

to transfer useful wisdom from similar settings they have experienced in the past. From a long

list of observable industry characteristics, analogizing managers choose a subset they believe

distinguishes similar industries from different ones. Faced with a novel industry, they seek a

familiar industry which matches the novel one along that subset of characteristics. They transfer

from the matching industry high-level policies that guide search in the novel industry.

We embody this conceptualization of analogy in an agent-based simulation model. The model

allows us to examine the impact of managerial and structural characteristics on the effectiveness

of analogical reasoning. With respect to managerial characteristics, we ﬁnd, not surprisingly,

that analogical reasoning is especially powerful when managers pay attention to characteristics

that truly distinguish similar industries from different ones. More surprisingly, we ﬁnd that the

marginal returns to depth of experience diminish rapidly while greater breadth of experience

steadily improves performance. Both depth and breadth of experience are useful only when one

accurately understands what distinguishes similar industries from different ones. We also discover

that following an analogy in too orthodox a manner —strictly constraining search efforts to what

the analogy suggests—can be dysfunctional. With regard to structural characteristics, we ﬁnd

that a well-informed analogy is particularly powerful when interactions among decisions cross

policy boundaries so that the underlying decision problem is not easily decomposed. Overall,

the results shed light on a form of managerial reasoning that we believe is prevalent among

practicing strategists yet is largely absent from scholarly analysis of strategy. Copyright 2005

John Wiley & Sons, Ltd.

INTRODUCTION

Strategy making is most critical in times of change

and in unfamiliar environments. It is in such a con-

text that the strategy makers of a ﬁrm must identify

a viable new strategic position or face the potential

Keywords: analogy; cognition; complex systems; strate-

gic decision making; modularity

∗Correspondence to: Daniel A. Levinthal, The Wharton School,

University of Pennsylvania, 2028 Steinberg-Dietrich Hall,

Philadelphia, PA 19104, U.S.A.

E-mail: levinthal@wharton.upenn.edu

demise of their enterprise. Yet how are managers to

make intelligent choices in these novel contexts?

Popular accounts suggest that managers in such

settings fall back on past learning and their expe-

rience in a variety of business settings. Managers

with such experience may be capable of seeing

through surface features of the current context to

deeper truths that underlie it. This paper aims to

shed light on the experiential wisdom that enables

strategy makers to cope with novel environments.

We argue that the basis of this sort of experien-

tial wisdom lies in analogical reasoning. For past

Final revision received 24 December 2004

692 G. Gavetti, D. A. Levinthal and J. W. Rivkin

experience to be of value in a world of novelty,

actors must be able to generalize from prior set-

tings to the current environment. This process of

mapping from a source context of prior experi-

ence to the current, ‘target’ context is precisely

what constitutes analogical reasoning (Gick and

Holyoak, 1980). To shed light on strategy mak-

ing in the face of novelty, we tackle two related

tasks. First, we describe the anatomy of analogi-

cal reasoning, discussing in detail the components

that make such logic work in the context of busi-

ness strategy. Second, we build a simulation model

of ﬁrms that use analogies to seek good strate-

gic positions in unfamiliar settings. An analysis of

the model allows us to identify situations in which

analogical reasoning works well or poorly.

Our focus on analogical reasoning contrasts with

the two dominant perspectives on strategic choice

in prior literature. In one corner of the schol-

arly ring, the positioning school within strategic

management has emphasized the importance of

deductive reasoning and rational choice in the

strategy-making process. The imagery is that of a

general and staff in a situation room, coolly survey-

ing topographical maps of the business landscape

(Ghemawat, 1999), picking out a peak, and direct-

ing the troops to take the summit. This image is

hard to sustain, however, in the world that posi-

tioning scholars have increasingly painted of ﬁrms’

strategies embodied in intricate arrays of inter-

locking choices (Porter, 1996; Ghemawat, 1999;

Rivkin, 2000; Siggelkow, 2002). With this recent

emphasis on nuanced interactions among strategic

choices, the positioning school has created a bit of

a logical predicament for itself: it is precisely in

complex worlds of highly interactive systems that

deductive reasoning and rational choice seem least

able to pinpoint effective positions. If the choices

involved in a strategy are numerous and each

affects the pay-offs associated with many others,

the computational load created by a deductive pro-

cess can quickly outstrip the bounded processing

power (Simon, 1955) of any management team.1

We explore below whether analogical reasoning

can cope not only with novel settings, but also

with novel settings that are complex.

In the opposite corner of the ring are those

who argue that ﬁrms discover effective positions

1Indeed, Rivkin (2000) shows that strategic problems can

become NP-complete— that is, intractable to algorithmic solu-

tion— as the degree of interaction among component choices

passes a critical threshold.

through local, boundedly rational search and luck.

In any given industry, a broad set of ﬁrms begin

their lives with a wide variety of strategic choices.

Each ﬁrm’s decisions and its routines are grad-

ually perturbed in ways that enhance immediate

performance, in a process that behavioral theorists

of the ﬁrm (Cyert and March, 1963) and evolu-

tionary scholars (Nelson and Winter, 1982) depict

as largely automatic, experiential, and emergent

(Mintzberg, 1978). A few fortunate ﬁrms happen

upon highly effective sets of choices and survive

an ensuing shakeout. The imagery here depicts

blindfolded individuals spread out widely on a

rugged landscape, each engaged in local hill climb-

ing (Levinthal, 1997). The water level rises over

time, and a few lucky survivors discover effec-

tive positions not by virtue of deductive power but

simply because they started their search in a for-

tunate place and scrambled up the right inclines.

This imagery doubtlessly paints an accurate pic-

ture of some successes. Pascale’s (1984) well-

known description of how Honda discovered its

effective position in the U.S. motorcycle indus-

try, for instance, is marked by happenstance and

incremental response to unforeseen problems and

opportunities.

We hesitate, however, to ascribe all effective

positions to luck and local search. Human rational-

ity may be limited, but intendedly rational action

surely remains possible (March and Simon, 1958).

By simplifying the problems they face, managers

can bring problems within the bounds of their pro-

cessing power and possibly come up with effective

solutions (Simon, 1991). This is a central role of

managerial cognition. Cognition can be especially

powerful when coupled with local search (Gavetti

and Levinthal, 2000); cognition can raise the odds

that a ﬁrm begins its search near a high, promising

mountain rather than in the valley beneath a mere

foothill, and local search can complete the journey

uphill. We see cognition, then, as a means to cope

with complexity.

Novelty, however, remains a challenge. What

cognitive processes are available to managers

when they face unfamiliar problems? In novel sit-

uations, where deduction is likely difﬁcult, past

lessons from prior settings can be a tremendously

powerful source of wisdom (Neustadt and May,

1986). One process for transferring such wis-

dom is analogical reasoning. As Thagard (1996:

80) points out, ‘analogies can be computation-

ally powerful in situations when conceptual and

The Power of Analogy 693

rule-based knowledge is not available.’ Analogies

allow actors to take the insight developed in one

context and apply it to a new setting.

In this paper, we examine how managers can use

analogy to tackle the twin challenges of complexity

and novelty. The managers we consider discover

new positions neither by reasoning from ﬁrst prin-

ciples of economics nor by undertaking unguided

local search. Rather, when faced with a new and

complex setting, managers identify the features of

the setting that seem most pertinent, think back

through their experiences in other settings with

similar features, and recall the broad policies that

worked well in those settings. These broad policies

then form the starting point for a local search pro-

cess. Analogies to other settings, drawn from direct

or vicarious experience, guide the strategy-making

process. The image of the strategist here is that of a

wizened trail guide who might not know the details

of the local landscape, but can recognize types of

terrain well enough to give some general guidance.

Analogical reasoning gives managerial cognition a

signiﬁcant hand in strategy making, and it empha-

sizes aspects of strategy making like pattern recog-

nition, judgment, and even wisdom —aspects that,

in our view, are prominent among practicing strate-

gists but are understated in the academic literature

on strategy.

To begin to address this gap, we develop an

analytical structure with which we can explore

both the power and limits of analogical reason-

ing in discovering strategic positions. We do so

by ﬁrst describing qualitatively what we view as

the basic features of analogical reasoning in strat-

egy making. We next formalize this argument in

an agent-based simulation of ﬁrms that struggle to

ﬁnd superior competitive positions in unfamiliar

industries. The simulation clariﬁes our depiction

of analogical reasoning. Results of the simulation

identify the managerial and structural characteris-

tics that make analogical reasoning more or less

powerful. Our analysis highlights the critical role

played by a manager’s mental representations, the

conditions that make breadth and depth of man-

agerial experience valuable, the dangers of holding

too closely to the guidance offered by an analogy,

and the power of analogy in the face of poorly

decomposed decision problems.

In our concluding discussion, we suggest that the

approach taken here helps to bridge two perspec-

tives on strategy making: behavioral approaches

that emphasize limits to cognition and the impor-

tance of search processes, and the positioning

approach, with its emphasis on conscious, apri-

ori strategic choice. Each approach encompasses

important elements of reality. The strategy ﬁeld, at

an aggregate level, has recognized the importance

of pluralism, but we often lack individual analyt-

ical frameworks that embrace and meld multiple

perspectives. We suggest that our analytical struc-

ture has this property and therefore, beyond the

value of the particular results we offer, we view

the conceptual framework as a useful contribution

that may facilitate a more integrated analysis of

strategic decision making.

MAKING STRATEGY BY ANALOGY

Reasoning by analogy is a common form of logic

among business strategists. Facing a novel oppor-

tunity or predicament, strategists think back to

some similar situation they have faced or heard

about, and they apply the lessons from that previ-

ous experience. Analogies —to the past, to other

ﬁrms or industries, and to other competitive set-

tings like sports or war —come up frequently in

strategy discussions. Popular management writers

hail pattern recognition and associated analogical

thinking as the best way for managers to cope

with rapid change (e.g., Slywotzky and Morrison,

1999). The case method, perhaps the most popular

form of management education, is designed in part

to give students a rich base from which to draw

analogies. Through the case method, students ‘are

led to active consideration of a tremendous num-

ber of diverse and related real situations, which

it would take them at least a lifetime of experi-

ence to encounter, and they are thus given a basis

for comparison and analysis when they enter upon

their careers of business action’ (Gragg, 1940).

In this section, we lay out the key elements

of analogical reasoning. A few examples illustrate

how analogies are commonly used and give us con-

crete instances in which we ground the following

conceptual discussion:

•The supermarket has served repeatedly as the

basis for analogical reasoning. Charlie Merrill

fashioned Merrill Lynch’s distinctive approach

to retail brokerage after the practices he witness-

ed as a long-time manager in the supermarket

industry. ‘Although I am supposed to be an

694 G. Gavetti, D. A. Levinthal and J. W. Rivkin

investment banker,’ he confessed, ‘I think I am

really and truly a grocery man at heart’ (Perkins,

1999). Likewise, when Charles Lazarus founded

the toy superstore Toys ‘R’ Us in the 1950s, he

relied explicitly on an analogy to supermarket

retailing. Indeed, he called his retail outlet the

Baby Furniture and Toy Supermarket until the

signage at a new site demanded a shorter name

(Hast, 1992). And when Thomas Stemberg, a

former supermarket executive, established the

ofﬁce supply superstore Staples, he posed the

initial strategic insight as an analogical question:

‘Could we create a Toys ‘R’ Us for ofﬁce sup-

plies?’ (Stemberg, 1996). The basic supermarket

formula—exhaustive selection, low prices and

margins, and high volume—has been applied

in a wide range of retail categories.

•Starting in the 1970s, Circuit City thrived by

selling consumer electronics in large super-

stores. A wide selection of products, profes-

sional sales help, and a policy of not hag-

gling with customers distinguished the com-

pany’s stores. In 1993, Circuit City surprised

investors by announcing that it would open Car-

Max, a chain of used car outlets. The com-

pany argued that the used car industry of the

1990s bore a close resemblance to the elec-

tronics retailing environment of the 1970s. The

industry was dominated, for instance, by small

Mom & Pop dealers with questionable reputa-

tions. The company hoped that its success for-

mula from electronics retailing would work well

in an apparently analogous setting.

•In 1997, the Board of the Internet portal Lycos

decided to grow rapidly and become a full-

ﬂedged new-media company, in part by means

of acquisition. After making its ﬁrst acqui-

sition—of homepage builder Tripod —Lycos’s

managers faced a decision that they later identi-

ﬁed as the pivotal choice in the company’s his-

tory: should Lycos maintain Tripod as a separate

operation with its own brand name or integrate

Tripod more fully into Lycos-branded opera-

tions? The management team argued the issue

for weeks, and ultimately an analogy proved

decisive in the debate. Traditional media compa-

nies, the managers observed, tended to maintain

multiple divisions with separate brands but to

coordinate back-ofﬁce operations. Since Lycos

wanted to become a large media company,

the managers reasoned, the company should do

likewise. Accordingly, the company kept alive

the brands of Tripod and subsequent acquisi-

tions, but integrated the back ofﬁces (Gavetti

and Rivkin, 2004).

•Many factors contributed to Enron’s startling

failure, but headlong diversiﬁcation based on

loose analogies played an important role. After

apparent success in trading natural gas and elec-

tric power, Enron executives moved rapidly to

create markets for other goods ranging from

coal, steel, and pulp and paper to weather deriva-

tives and broadband telecom capacity. Ana-

logical reasoning seemed to drive the expan-

sion. Executives looked for markets with cer-

tain characteristics: fragmented demand, rapid

change due to deregulation or technological

progress, complex and capital intensive distribu-

tion systems, lengthy sales cycles, opaque pric-

ing, and mismatches between long-term supply

contracts and short-term ﬂuctuations in customer

demand (Salter, Levesque, and Ciampa, 2002).

On the broadband opportunity, for instance,

Enron Chairman Kenneth Lay said: ‘[Broad-

band]’s going to start off as a very inefﬁcient

market. It’s going to settle down to a business

model that looks very much like our business

model on [gas and electricity] wholesale, which

obviously has been very proﬁtable with rapid

growth’ (Gas Daily, 2000). However, the ana-

logical reasoning failed to appreciate important,

deeper differences between the market for nat-

ural gas and the market for bandwidth. The

broadband market was based on unproven tech-

nology and was dominated by telecom compa-

nies that largely resented Enron’s encroachment.

The underlying good, bandwidth, did not lend

itself to the kinds of standard contracts that made

efﬁcient trading possible in gas and electricity.

Perhaps worst, in broadband trading Enron had

to deliver capacity the ‘last mile’ to a customer’s

site, an expensive chllenge that gas wholesalers

did not face. Even according to Enron’s mis-

leadingly rosy ﬁnancial reports, the broadband

venture was disastrous (Salter et al., 2002).

In each of these cases, a management team borrow-

ed a broad set of related choices from one industry

and applied the system to a new industry that it

believed to be similar on some crucial dimensions.

In some instances (e.g., Merrill Lynch, Lycos), the

team was motivated by a problem that required a

solution. In other cases (e.g., Enron), the analogy

took the form of a solution seeking a problem. In

The Power of Analogy 695

the following discussion, we focus on the problem-

requiring-a-solution type of analogy, though we

suggest that the solution-seeking-a-problem form

of analogical reasoning is important as well, par-

ticularly as a source on entrepreneurial activity.

Well beyond the context of business, the use

of analogical reasoning has been of long-standing

interest among cognitive scientists (Gick and

Holyoak, 1980; Holyoak and Thagard, 1995; Tha-

gard, 1996). When reasoning by analogy, an indi-

vidual starts with a situation to be handled —the

target problem. The actor develops a mental

representation of the target problem, a lower-

dimensional sketch that, in the actor’s view, cap-

tures the salient characteristics of the situation. She

then uses some computational procedure to scour

other settings with which she is familiar, due to

either direct or vicarious experience, and identiﬁes

a setting that displays similar salient characteris-

tics. This setting serves as a source of a candidate

solution. The individual then transfers the candi-

date solution and applies it to the target problem.

Clearly the power of analogy depends on the valid-

ity of the similarity mapping between source and

target contexts as well as the quality of the solution

suggested by the source context.

The danger of ‘superﬁcial mappings’ between

the source and the target is a widely studied phe-

nomenon in cognitive psychology. Experimental

work and ﬁeld evidence strongly suggest that poor

analogies are typically based on representations

that capture only superﬁcial features of the prob-

lems (Holyoak and Thagard, 1995). Indeed, re-

search in this ﬁeld shows that superﬁcial similarity

can readily induce experimental subjects, even

well-educated individuals, to adopt poor analo-

gies.2We suspect the same is true of practicing

managers.

2In one study (Gilovich, 1981), for example, students of

international conﬂict at Stanford were told of a hypothetical

foreign policy crisis: a small democratic nation was being

threatened by an aggressive, totalitarian neighbor. Each student

was asked to play the role of a State Department ofﬁcial and to

recommend a course of action. The descriptions of the situation

were manipulated slightly. Some of the students heard versions

with cues that were intended to make them think of the crises

that preceded World War II. The President at the time, they were

told for example, was ‘from New York, like Franklin Roosevelt,’

refugees were ﬂeeing in boxcars, and the brieﬁng was held in

‘Winston Churchill Hall.’ Other students heard versions that

might have reminded them of Vietnam. The President was

‘from Texas, the same state as Lyndon Johnson,’ refugees were

escaping in small boats, and the brieﬁng took place in ‘Dean

Rusk Hall.’ Clearly, there is little reason that the home state

of the President, the vehicles used by refugees, or a brieﬁng

In the context of business strategy, the

observable characteristics of an industry may

constitute the dimensions of a representation.

Three features of any industry, for instance, are

the size of economies of scale, the size of customer

switching costs, and the heterogeneity of customer

tastes. Suppose a manager in a novel setting

opts to represent her target problem along these

dimensions. On the basis of an initial assessment

of the target, the manager judges that the target

industry is characterized by modest economies of

scale, large switching costs, and diverse customer

tastes. The manager then engages in a simple

computational procedure: where has she seen

modest economies of scale, large switching costs,

and diverse tastes before, in other industries?

The manager reviews her experience and realizes

that, in a speciﬁc industry that was similar

along these three dimensions, a particular niche

provider of high-end, premium products was

highly successful. She then transfers this solution

to the target industry, adopting a small-scale

manufacturing policy, a cream-skimming pricing

policy, a targeted sales policy, and so forth. If the

dimensions she chose to focus on are the ones that

best summarize the true drivers of performance,

the ﬁrm improves its odds of success.

The difﬁculty that faces the analogizing manager

is that there are innumerable dimensions along

which one can form a representation and some

dimensions may be misleading. Suppose, for

instance, that a different manager facing the

same target industry ignores economies of scale,

switching costs, and customer heterogeneity.

Instead he pays attention only to the prevalence

of Internet technology in the industry. He notes

that the target industry relies heavily on the

Internet for sales, marketing, and distribution.

room name should inﬂuence one’s recommendation. Yet the

surface features caused the two groups to reach very different

conclusions. Students in the ﬁrst group were signiﬁcantly more

likely to apply the lessons of World War II— that aggression

must be met with force—than were students in the second group,

which veered toward a hands-off policy inspired by Vietnam.

Not only were the subjects of the experiment lured by

superﬁcial likeness, but —perhaps more disturbing —they were

not even aware that they had been lured. After making a

recommendation, each student was asked, ‘How similar is the

situation to World War II?’ and ‘How similar is the situation

to Vietnam?’ The two groups gave identical answers. Small,

superﬁcial, irrelevant features led well-educated students to draw

their analogies from different sources. This caused them to

recommend very different candidate solutions. Yet the students

seemed unaware that they were drawing from different sources.

696 G. Gavetti, D. A. Levinthal and J. W. Rivkin

In his experience with Internet-based industries,

successful companies in such settings spend

aggressively in order to get big fast. He deploys

this candidate solution in the target industry. If

the prevalence of Internet technology does not

truly shape the target landscape, he may ﬁnd

that his mass-market product is far less appealing

to diverse customers than the offerings of niche

competitors and that his large-scale manufacturing

operations do not lower his costs. By focusing on

an irrelevant dimension and ignoring three other,

more pertinent characteristics, he has been led to

a poor analogy.

Analogical reasoning and strategic positioning

The target problem is the business situation the

strategist hopes to resolve. In our particular setting,

the challenge of interest is how to position a ﬁrm

in an industry that is novel—novel either for the

managers of the ﬁrm itself (e.g., the used car

business for Circuit City’s managers) or for all

managers (e.g., the Internet portal industry). To

position itself, the company must make a vast array

of detailed choices about how to develop, design,

produce, sell, deliver, and service products (Porter,

1985). These choices incur costs and generate

buyer value, and therefore they shape the economic

success of the ﬁrm.

We ﬁnd it useful to conceive of the target prob-

lem in terms of a high-dimensional performance

landscape (Kauffman, 1995; Levinthal, 1997;

McKelvey, 1999, and references therein). Each

detailed choice constitutes a horizontal dimen-

sion on this landscape, and the vertical dimension

records the economic success associated with each

combination of choices, thereby creating a surface.

The task of management is to discover a com-

bination of choices that, together, produce high

performance; in graphical terms, the challenge is

to ﬁnd a high peak on the performance land-

scape. Because the decisions may interact with one

another (that is, the choice made on each may alter

the marginal costs and beneﬁts associated with the

others), the performance landscape may contain

numerous local peaks. A local peak is a conﬁg-

uration of choices from which one cannot improve

performance by altering any single choice, even

though simultaneous change in many choices may

reposition a ﬁrm to higher ground.

In many business settings, detailed choices clus-

ter together to form policy domains. Choices about

automation, lot sizes, work ﬂow, factory scale,

and so forth may aggregate into a manufacturing

policy, for example, with combinations of these

detailed choices taking on labels that summarize

the policies (e.g., mass production, job shop oper-

ations, cell manufacturing, or batch production).

Obviously, the choice of a particular approach

for higher-order policies typically has an inﬂuence

on detailed choices. Three aspects of these pol-

icy domains shape the modeling choices we make

below. First, the interactions within the decisions

that make up a policy domain may be more intense

than the interactions across domains. This creates

a type of interdependence that Simon (1962) has

labeled near-decomposability. In our subsequent

analysis, we consider the effect of decomposability

on the efﬁcacy of analogy.

Second, the effect of one policy domain on

the efﬁcacy of another may depend not on the

detailed choices made in the ﬁrst domain, but on

the net effect of those choices. Take, for instance,

the example of Circuit City’s original success

in consumer electronics. Circuit City’s success

in marketing and selling consumer electronics

clearly had an effect on its policy for procuring

electronics. When the marketing and sales policy

produced strong demand among consumers in

the 1970s, the marginal value of large-scale

procurement increased dramatically. The effect of

marketing and sales on procurement depended

not on detailed marketing and sales choices such

as pricing, sales tactics, and sales compensation,

but only on the total demand that those choices

generated in aggregate.

Third, the high-level policies adopted by a com-

pany are often more visible to outsiders and per-

haps more memorable to insiders than the detailed

decisions that underpin them. In Circuit City’s

case, the consumer-friendly marketing and sales

policy that it created in the electronics retailing

market was highly visible, but the compensation

schemes and information technology that enabled

that policy remained more obscure. We assume

below that analogies provide guidance about high-

level policies, not detailed decisions per se.

Quality of analogically based solutions

On the basis of representational similarity, the

analogizing manager chooses a source industry

from her library of experience. Beyond the

choice of representation, we see three factors that

The Power of Analogy 697

inﬂuence the quality of the guidance provided

by an analogy. First, there is the breadth of

the manager’s experience. The larger the number

of other industries the manager has experienced,

personally or vicariously, the greater is the

chance that the manager will be familiar with

an industry that matches the target well along

the representational dimensions. It is this kind of

breadth that strategy consultants often purport to

bring to a client.

A second factor that affects the quality of an

analogy’s guidance is the depth of the manager’s

experience in the source industry. If the manager

has spent a great deal of time in the source

industry and understands deeply what distinguishes

a good position from a poor one in that setting,

it is more likely that the analogy will accurately

guide the manager’s ﬁrm to a favorable set of

policies. On the other hand, deep experience in

a source setting might be a double-edged sword

(Levinthal and March, 1993). If a manager couples

deep experience with a poorly chosen source, the

result can be disastrous: profoundly convinced that

she knows exactly what to do, the manager may

persist in applying the policies that succeeded

in a prior, but very different, setting and may

ignore negative feedback from the environment.

In this sense, Enron’s extensive experience with

gas and electricity trading may have impeded its

ability to adapt in its broadband venture. More

generally, it is useful to consider how seriously

managers take their analogies and how closely

they abide by the candidate solutions. In the

formal model we develop below, we consider both

orthodox analogizers, who consistently hold to the

set of policies recommended by their analogy, and

heterodox analogizers, who use the recommended

set only as a starting point for further exploration.

Analogies typically give high-level guidance

whose details must be worked out at the ground

level of the target industry. This points to a third

factor that inﬂuences the quality of the guidance

provided by an analogy: the degree to which a

high-level policy choice constrains the detailed

choices that comprise it. In some settings, a policy

choice may tightly constrain detailed choices,

while in others a policy choice may leave wide

latitude for what goes on below. Since candidate

solutions are transferred at the level of policies,

not detailed choices, an analogy provides sharper

guidance in the former setting than in the latter.

When an analogy is chosen well, we expect clearer

guidance to produce higher performance.

The subtlety of analogizing and its implications

for the establishment of effective competitive

positions suggest that a formal model of the

process may generate valuable insights. In the

following section, we describe such a model,

built as an agent-based simulation. The model

incorporates the features we have described here:

representations of variable quality; the possibility

of analogies based on superﬁcial similarity; target

and source industries of variable complexity; a

distinction between broad policies and detailed

choices; interactions across choices within policies

and interactions across policies; and managers

who differ in the breadth and depth of their

experience and in the orthodoxy with which

they follow their candidate solutions. The model

allows us to conﬁrm some of the relatively

straightforward arguments we have made here

(e.g., that better representations make analogical

reasoning more effective). It also reveals some less

obvious factors, such as the role of the underlying

interaction structure among policy choices, that

make analogical reasoning a more or less powerful

tool for discovering effective competitive positions

in novel and complex settings.

A MODEL OF SEARCH IN NOVEL AND

COMPLEX WORLDS

The simulation model undertakes two basic sets

of operations. First, it generates a family of

performance landscapes—a target landscape and

a set of potential sources —in which the modeler

can tune the relationship between the sources and

the target. The performance landscape itself can

be tuned with respect to the degree to which ﬁrm

choices on these landscapes are decomposable.

Second, the model permits ﬁrms to search for

effective positions on the target landscape in a

variety of ways. Firms may differ, for example,

in terms of the mapping between targets and

sources that each ﬁrm uses, the way in which

attractive candidate solutions are identiﬁed on each

source landscape, and the manner in which each

ﬁrm couples analogical reasoning with incremental

search. We discuss the generation of families of

landscapes and the search for effective positions in

turn. (Table 1 summarizes the model’s parameters

and symbols.)

698 G. Gavetti, D. A. Levinthal and J. W. Rivkin

Table 1. Parameters and symbols

Parameters related to generation

of families of landscapes

PThe number of high-level policy decisions that each ﬁrm faces on a

target or source landscape.

DThe number of detailed decisions each ﬁrm must make within each

high-level policy. P×Dis the total number of decisions each

ﬁrm makes. Each decision is binary so there are 2P×Dpossible

conﬁgurations of detailed decisions.

XThe total number of observable industry characteristics. Each

characteristic is binary, so there are 2Xindustries or landscapes in

a family: one target and 2X−1 potential sources.

KwThe probability that the performance contribution of a focal detailed

decision is affected by the resolution of each of the other D−1

detailed decisions within the focal decision’s policy. The ‘w’in

Kwsigniﬁes ‘within’.

KbThe probability that the performance contribution of a focal detailed

decision is affected by the resolution of each of the other P−1

high-level policies. ‘b’ signiﬁes ‘between’.

XPROB An X-digit vector of probabilities. The ﬁrst element is the probability

that the ﬁrst observable characteristic affects the performance

contribution of each detailed decision. Likewise for elements 2, 3,

...,X.

Parameters related to ﬁrm search DEPTH The portion of each source landscape that is evaluated in order to

generate the library of most promising policy conﬁgurations. The

higher this value is, the more likely it is that the library truly

identiﬁes the best policies for each source.

BREADTH The portion of the library of most promising policy conﬁgurations

that is available to a particular analogizing ﬁrm.

###Numbers within angle brackets refer to the representation of each

analogizing ﬁrm. The ﬁrst number is the observable characteristic

the ﬁrm’s managers deem the most important; the second number

is the second most important; and so forth.

Orthodoxy Each analogizing ﬁrm is set to be orthodox or heterodox. During

incremental search, orthodox ﬁrms never violate the policy

guidance provided by the analogical source, while heterodox ﬁrms

may ignore that guidance.

Symbols used to describe the

model

dA particular conﬁguration of detailed decisions; a P×Dstring of

zeros and ones.

diThe resolution, zero or one, of a speciﬁc detailed decision. iis

between 1 and P×D.

{###}Numbers within braces consistently refer to lists of P×Ddetailed

decisions. Within a list, vertical lines separate the Ppolicy

domains.

[# # #] Numbers within square brackets refer to lists of Phigh-level policy

choices.

(# # #) Numbers within parentheses refer to lists of Xobservable industry

characteristics.

Families of performance landscapes

We generate families of landscapes using an

adaptation of Kauffman’s (1993) NK model.3

3Developed in the biological sciences, Kauffman’s model

has now been used to explore a variety of issues related

to organizational and technological search. See, for instance,

Kauffman (1995), Westhoff, Yarbrough, and Yarbrough (1996),

Levinthal (1997), McKelvey (1999), Rivkin (2000), Gavetti and

Levinthal (2000), Rivkin (2001), and Rivkin and Siggelkow

(2002, 2003). Fleming and Sorenson (2001, 2004) have found

The NK structure highlights the interdependency

among choices. Settings in which choices are

more interdependent (a higher Kvalue in the

standard structure) result in more rugged, multi-

peak performance surfaces because a change in

one choice will have repercussions for many of

the Nchoices.

that subtle predictions of the NK model concerning patterns in

technological search are borne out in patent data.

The Power of Analogy 699

In our elaboration of Kauffman’s basic

framework, we introduce a hierarchy of choices.

Each ﬁrm is assumed to face Phigh-level policy

decisions, each of which includes Ddetailed

choices. A competitive position, then, is deﬁned by

P×Ddetailed choices. For simplicity, we assume

that each choice offers two options. A strategic

position can then be represented as a P×D-digit

string of zeros and ones: d={d1d2...d

P×D}with

di=0or1foralli. Therefore, a ﬁrm has 2P×D

positions available to it in the target industry.

Likewise, in each source industry, there are 2P×D

possible conﬁgurations of choices.

Policy choices are related to detailed choices by

a simple majority rule. Suppose, for instance, that

D=3. We say that a particular policy choice is

1 if the detailed decisions within the policy are

conﬁgured such that the majority of them are 1:

{111},{110},{101},or{011}.Itis0ifmostof

the detailed decisions are 0: {000},{001},{010},

or {100}. Each conﬁguration of detailed choices

gives a unique conﬁguration of policies, but each

conﬁguration of policies is consistent with many

conﬁgurations of detailed choices.4

In order to explore analogical reasoning and

possible similarity between source and target

contexts, we also need to characterize and be

able to tune the degree to which one business

context is like another. We postulate that associated

with all industries, including the target industry

and each potential source industry, is a set of

observable characteristics. These are the possible

dimensions along which representations can be

drawn. In particular, we assume that there are X

observable characteristics, each of which offers

two options. As a result, there are 2Xcombinations

of observable characteristics. Given a particular

speciﬁcation of the Xcharacteristics for the target

industry, there are 2X−1 possible combinations

of characteristics for potential source industries.5

4Suppose, for instance, that P=3andD=3. Then the detailed

conﬁguration {011|111|001}implies the policy choice [1 1 0].

But [1 1 0] is also consistent with {101|011|000},aswellas

many other conﬁgurations of detailed choices. For clarity, we

place policy conﬁgurations in square brackets and detailed

conﬁgurations in braces. Within detailed conﬁgurations, we will

separate policy domains by vertical bars. We require Dto be an

odd number so that the majority rule provides a unique policy

choice for each conﬁguration of detailed decisions.

5There are 2Xpossible combinations of characteristics; however,

one of them signiﬁes the target landscape itself.

Thus each industry has an X-digit ‘tag.’ Recall,

for instance, the hypothetical situation we de-

scribed earlier in which industries had four

observable characteristics: the size of economies

of scale (large or small), the size of customer

switching costs (large or small), the heterogeneity

of customer tastes (high or low), and the

prevalence of Internet technology (high or low).

The particular industry our two managers faced,

with small economies of scale, high switching

costs, diverse customers, and prevalent Internet

technology, might be represented as (0 1 1 1). In

choosing an analogical source, each manager paid

attention to only part of this vector; we return to

this consideration below.

Our goal is to create families of landscapes

in which (a) we control the pattern of interac-

tion among detailed choices within and across

policy domains, (b) pairs of landscapes with sim-

ilar observable characteristics—similar tags—are

more alike than are pairs with very different char-

acteristics, and (c) some observable characteris-

tics have more inﬂuence than others on landscape

topography.

To meet this goal, we assume that the

contribution to performance of each detailed

decision on each landscape depends not only on

the resolution of that decision (0 or 1), but also,

possibly, on the resolutions of other decisions

within the focal decision’s policy, the resolutions

of other policies, and the states of observable

characteristics. The pattern of dependence is

inﬂuenced by three parameters. Kw, a number

between 0 and 1, is the probability that each

detailed choice’s contribution depends on the

resolution of each other choice within the focal

choice’s policy domain. Kb, a number between 0

and 1, controls interdependence between policy

domains. Speciﬁcally, it is the probability that

the focal choice’s contribution depends on the

resolution of each other policy. XPROB , a vector

of Xnumbers each between 0 and 1, controls

the probability that each observable characteristic

inﬂuences the contribution of the focal choice. An

observable characteristic with a large component in

XPROB has a large impact on landscape topography.

For a given set of values for P,D,X,Kw,

Kb,andXPROB , the simulating computer generates

an inﬂuence matrix—a matrix that records which

choices, policies, and observable characteristics

inﬂuence the contribution of each detailed choice.

(See Figure 1 for an example.) This inﬂuence

700 G. Gavetti, D. A. Levinthal and J. W. Rivkin

Figure 1. Example of an inﬂuence matrix

matrix is then used to generate target and source

landscapes. We lay out the details of this procedure

in an Appendix, and we describe the upshot of the

procedure in intuitive terms in the following two

paragraphs.

•First consider observable characteristics. Each

of the 2Xcombinations of characteristics deﬁnes

a particular business context. In the example we

used above, the target context was characterized

by the vector (0111). There are 15 possible

source settings with proﬁles of characteristics

such as (0001), (0010), (0011), and so

forth. Suppose the inﬂuence matrix looks like

Figure 1. The fourth observable characteristic

does not inﬂuence the contribution of any

choice. Therefore, on source landscape (0110),

the contribution of every decision is the same

as it is on the target landscape (0111); the

difference in the fourth characteristic alters

nothing. Source landscape (0110) is identical

to the target landscape in every way and

is therefore an excellent source for analogy.

In contrast, the ﬁrst observable characteristic

inﬂuences the contribution of every choice.

On source landscape (1111), which differs

from the target (0111) only along the ﬁrst

characteristic, the contribution of every decision

is different from what it is on the target.

The difference in the ﬁrst characteristic shifts

each contribution altogether so that (1111)

bears no resemblance whatsoever to the

target (0111). More generally, the larger is

an element of XPROB, the more profoundly

does a difference in that characteristic alter

the landscape. In choosing a source for an

analogy, it is crucial to pay attention to

observable dimensions with high values of

XPROB. Our procedure for constructing families

of landscapes ensures that landscapes with

similar observable characteristics have similar

shapes. Moreover, it allows us to control which

characteristics strongly delineate groups of

landscapes with similar topographies and which

characteristics are weak, or even irrelevant,

delineators.

•Next consider the effects of Kwand Kb.When

Kw=Kb=0, the contribution of each choice

depends only on the resolution of that choice

and the state of observable characteristics. On

any given landscape, i.e., for any particular

set of observable characteristics, the decision

problem facing a ﬁrm is easy to address; a

change in any choice alters the contribution of

no other choice so a ﬁrm can ﬁnd the optimal

conﬁguration of choices simply by adjusting

each choice to the resolution, 0 or 1, that

makes the greater contribution in isolation. The

landscape is smooth, single peaked, and easy to

scale. When Kwis high and Kbremains low, the

system is nearly decomposable (Simon, 1962).

A change in any choice alters the contributions

of other choices within the same policy, but it

is possible for the ﬁrm to tackle its decision

The Power of Analogy 701

problem policy by policy. As Kbrises, however,

decomposition becomes more difﬁcult because a

change that alters a policy now has ramiﬁcations

for choices in other domains. Graphically, we

ﬁnd it useful to think of Kwand Kbin terms of

plateaus. When Kwand Kbare low, landscapes

are smooth. As Kbrises, they develop distinct

plateaus, with the edges of plateaus deﬁned by

changes in policies. As Kwrises, the surfaces of

individual plateaus become internally rugged.

Having generated a family of landscapes—target

and potential sources—whose relations to one

another are well controlled, the computer pinpoints

an attractive candidate solution on each of the

(2X−1)potential sources. Speciﬁcally, for a

given source landscape, the computer considers

each of the 2Pcombinations of policies, surveys

a fraction DEPTH of the detailed conﬁgurations

consistent with each policy conﬁguration, assesses

the performance of each detailed conﬁguration

it surveys, and remembers the policy that

produces the best results on average for the

detailed conﬁgurations considered. The result is

an exhaustive library of the most promising policy

conﬁgurations on each and every potential source

landscape. Table 2 shows what the library might

Table 2. Example of a library of source landscapes and

promising policies

Source

landscape

Most promising

policy

conﬁguration

Average performance

within most promising

conﬁguration

(0 0 0 0) [1 0 0] 0.65

(0 0 0 1) [1 0 0] 0.65

(0 0 1 0) [0 1 1] 0.62

(0 0 1 1) [0 1 1] 0.62

(0 1 0 0) [0 0 0] 0.55

(0 1 0 1) [0 0 0] 0.55

(0 1 1 0) [0 1 0] 0.74

(1 0 0 0) [1 0 0] 0.61

(1 0 0 1) [1 0 0] 0.61

(1 0 1 0) [1 1 0] 0.65

(1 0 1 1) [1 1 0] 0.65

(1 1 0 0) [0 1 0] 0.73

(1 1 0 1) [0 1 0] 0.73

(1 1 1 0) [1 1 0] 0.70

(1 1 1 1) [1 1 0] 0.70

Note that in this example, pairs of source landscapes that

differ only in the last characteristic, which has no inﬂuence on

performance levels, always have the same most promising policy

conﬁguration and the same average performance within the most

promising policy conﬁguration (assuming DEPTH =1).

look like for the example used above. The library

includes the information, for instance, that the

policy [1 0 0] produced the best results on the

source landscape with observable characteristics

(0001). The promising policy conﬁgurations

will later serve as candidate solutions. DEPTH

parameterizes the depth of experience on which

candidate solutions are based.

Search for effective positions

The stage is now set for a discussion of how

ﬁrms search the target landscape. We consider

two basic classes of search strategies: local search

and analogical reasoning. Local searchers rely

on simple hill-climbing and make no use of

analogy. Each is given an initial conﬁguration of

detailed choices and a corresponding set of policy

choices by chance. That is, each is released at a

random location on the target landscape. In each

subsequent period, each local searcher considers

the P×Dalternatives to its current conﬁguration

that involve a change in a single detailed choice.

If it spots opportunities for improvement, the

local searcher pursues one of the opportunities.

Otherwise, it has arrived atop a local peak and

remains there for the duration of the simulation.

Analogizers, in contrast, base their initial

conﬁgurations on the library that was generated

earlier. Each analogizer has access to a fraction

BREADTH of the full library of potential source

landscapes. The particular source landscapes to

which an analogizer has access are chosen

at random. In the example above with 15

possible sources, an analogizer with BREADTH =

0.33 might be familiar with landscapes whose

observable characteristics are (1110), (1101),

(1010), (0100), and (1100). A less experienced

ﬁrm with BREADTH =0.2 might know only the

landscapes with the observable characteristics

(0110), (1001), and (101 0).

Each analogizing ﬁrm also has an ordered list

of the representational dimensions, or observable

characteristics, that it considers to be important.

We consider this list to be the ﬁrm’s represen-

tation of its problem context. The representation

321, for instance, implies that the ﬁrm consid-

ers the third observable characteristic to be most

important, the second characteristic to be second-

most important, the ﬁrst characteristic to be third-

most important, and the fourth characteristic alto-

gether irrelevant. The representation 4reﬂects

702 G. Gavetti, D. A. Levinthal and J. W. Rivkin

a belief that only the fourth characteristic mat-

ters. The ﬁrst manager we described before—who

considered scale economies, switching costs, and

customer heterogeneity—might have representa-

tion 321while the second manager—who paid

attention only to the prevalence of Internet tech-

nology—might have representation 4.

The analogizer uses its representation to choose

among source landscapes within its breadth of

experience. It does so in a lexicographic manner.

Take, for instance, a ﬁrm familiar with source

landscapes (1110), (1101), (1010), (0100),

and (1100). Assume that the ﬁrm has the

representation 321, and the target landscape is

(0111), as in our earlier example. The ﬁrm ﬁrst

looks for a source landscape that matches the

target perfectly on characteristics 3, 2, and 1. It

ﬁnds none. It then disregards the characteristic

it considers least important, characteristic 1, and

looks for a source landscape that matches the

target on characteristics 3 and 2. It ﬁnds one such

landscape, (1110), and chooses to focus on that

source.

Having chosen a source, the analogizer transfers

the candidate solution—in this example and in

line with Table 2, [1 1 0] —back to the target.

Consistent with our discussion in the previous

section, candidate solutions carry only high-level

policy guidance. The ﬁrm chooses, at random, a

set of detailed choices that abide by the policy

guidance. In the example, the ﬁrm may choose

{110|111|010}. This conﬁguration serves as the

ﬁrm’s starting point for subsequent search. In

following periods, the analogizing ﬁrm engages in

incremental improvement just as a local searcher

would. If we force the ﬁrm to abide by its

analogy in an orthodox manner, it never considers

incremental improvements that violate the initial

policy guidance. From {110|111|010}, for instance,

an orthodox analogizer would never consider

{010|111|010}because to do so would switch the

ﬁrst policy from 1 to 0. On the other hand, if

we allow the ﬁrm to be heterodox in its use of

analogy, its incremental improvement efforts are

not constrained in this manner.

Note the crucial role that the ﬁrm’s represen-

tation plays in this process. If the representation

matches XPROB closely, in the sense that observ-

able characteristics with high inﬂuence are early

in the ﬁrm’s list, the ﬁrm is likely to focus on a

source landscape that closely resembles the target.

The candidate solution will then guide the ﬁrm

to a promising starting point. On the other hand,

if the representation causes the ﬁrm to pay atten-

tion to industry characteristics that are secondary

or irrelevant, the source will typically bear little

resemblance to the target, and the candidate solu-

tion will likely provide poor guidance.

RESULTS

Our simulation results identify factors that

make analogical reasoning especially powerful

as well as relationships among those factors.

We start by looking at factors related to the

analogizing management teams: the quality of

their representations, the breadth and depth of

their experience, and the orthodoxy with which

they use the candidate solution. We then turn to

structural characteristics of the industries they face,

especially the decomposability of the problem

space and the breadth of each policy domain.

Throughout this section, we examine families of

landscapes in which there are four observable

characteristics (X=4), three of which truly affect

landscape structure and one of which is irrelevant

(i.e., XPROB =(0.5,0.5,0.5,0.0)).

We compare the position-seeking success of

ﬁve types of ﬁrms: a local searcher and four

analogizers that differ in their representations.

The ﬁrst analogizer, with representation 123,

correctly surmises that the ﬁrst three observable

characteristics affect the choice/payoff mapping.

The second, with 1, heeds only one of the three

relevant characteristics. The third, with 432,

gives primacy to the irrelevant characteristic,

but also heeds some relevant factors. The ﬁnal

analogizer, with 4, heeds only the characteristic

that has no bearing on the landscapes. Among

representations that include a subset of observable

characteristics, these four cover the spectrum from

the longest (123and 432) to the shortest

(1and 4) as well as the range from the most

accurate given length (123and 1) to the least

accurate (432and 4).6Results for ﬁrms with

other representations fall between results for these

four ﬁrms in an intuitive way; the performance

of a ﬁrm with representation 12, for instance,

lies between that of the ﬁrm with 1and that of

6Our premise is that boundedly rational managers cannot attend

to all aspects of their environment. For this reason, we do not

examine ﬁrms that consider all four observable characteristics.

The Power of Analogy 703

Figure 2. Firm performance over time

the ﬁrm with 123. Results for ﬁrms with other

representations are available from the authors on

request.

We report ﬁrm performance as a portion of

the highest performance attainable on the target

landscapes that were explored. By this measure,

a type of ﬁrm that always attains the global

maximum will achieve a performance of 1.00.

Unless otherwise noted, each result is an average

over 1,000 families of landscapes, each explored

by 10 ﬁrms of each type. Each of the 1,000 families

of landscapes shares the same structural parameters

but constitutes an independent draw from the same

underlying distribution.

Management characteristics

We ﬁrst examine a situation in which P=3

and D=3. To highlight the effect of choosing

the correct observable characteristics on which

to base one’s representation, we choose a set

of baseline parameter settings in which cognition

is most powerful. DEPTH =BREADTH =1forall

ﬁrms, and all analogizing ﬁrms use candidate

solutions in an orthodox manner. In addition, we

set the probability of interaction across policy

choices, Kb, equal to one and the probability of

interaction among decisions within a policy, Kw,

to zero.7For the sake of comparability, we model

7The results we obtain from this ﬁrst set of analyses are robust to

different speciﬁcations for the interaction structure. We explore

a local searcher that abides by its initial, randomly

assigned policy choices in an orthodox manner.

Figure 2 shows the average performance of each

type of ﬁrm over the course of the simulation.

On average, ﬁrms with better representations

achieve higher performance in the initial period

than do ﬁrms with worse representations because

their initial policy choices are guided by a

source that more closely resembles the target.

Within the constraints of these initial policy

choices, the sets of detailed decision choices

are randomly speciﬁed in the initial period. As

a result, there is considerable room for each

ﬁrm to adapt and improve its performance over

subsequent time periods, even while it maintains

its set of overarching policies. Higher-quality

representations place ﬁrms in more opportune

locales, as measured by their initial superior ﬁtness

as well as by their higher asymptotic value.

While all ﬁrms search the landscape to identify

more favorable sets of decisions, ﬁrms with better

representations carry out this search in more

favorable regions of the landscape.

An intriguing aspect of Figure 2 is that the

ﬁrm with representation 4, the ﬁrm that heeds

only an irrelevant dimension, fares substantially

better than the local searcher. Even a ‘false’

the impact of the interaction structure in the next set of analyses.

We choose this particular combination of structural parameters

because it yields the greatest power of cognition. Below we

explore how the power of cognition depends on the structure of

the decision problem.

704 G. Gavetti, D. A. Levinthal and J. W. Rivkin

analogy has power. This result arises because, in

choosing a source landscape on the basis of an

irrelevant characteristic, the ﬁrm may, by chance,

focus on a source landscape that happens to share

relevant characteristics with the target as well.

Suppose, for instance, that the target landscape

has characteristics (0111), as in the example in

the previous section. The ﬁrm with representation

4might choose as its source a landscape such

as (1111) or (0011) or (0101)—a landscape

that does overlap with the target on some relevant

dimensions. Based on this ‘false’ analogy, the

ﬁrm identiﬁes a set of choices coherent with this

representation which, purely by chance, may prove

useful in the target problem. In the discussion

section below, we return to the virtues of false

analogies.

We turn next to the effects of experience

on the power of analogy. Table 3 shows the

long-run differential between the performance

of each analogizer and the performance of the

local searcher for various levels of BREADTH,the

portion of the source library to which the ﬁrm

has access. Table 4 presents similar results for

DEPTH, the portion of each policy domain that

is explored to generate the library of candidate

solutions. There are at least two striking features

of these results. First, while greater BREADTH

of experience steadily improves performance, the

marginal returns to DEPTH of experience diminish

rapidly; a management team with very limited

experience on each source landscape fares nearly

as well on the target as does a team with extensive

experience.8This result suggests that it is far more

important to identify an analogical source that

shares the structure of the target problem than

to pinpoint the absolute best solution within the

source problem context. Second, greater breadth

and depth of experience boost performance more

when a ﬁrm has a high-quality representation.

Put differently, broader and deeper experience

does not improve a ﬁrm’s lot if it draws from

that experience on the basis of less relevant

characteristics.

Tables 3 and 4 obscure a hazard of experience

we discussed above: broad and especially deep

experience may motivate a management team to

abide by its candidate solution even though the

subsequent local search process identiﬁes superior

8In each row, the results for ﬁrms with DEPTH =0.25, DEPTH =

0.50, and DEPTH =1.00 are statistically indistinguishable.

Table 3. Effect of breadth of experience on analogizer’s long-run performance advantage

Long-run performance

advantage over a local

BREADTH =

searcher for an analogizer

with representation ... 0.10 0.25 0.50 0.75 1.00

1230.069 (0.002) 0.086 (0.002) 0.111 (0.001) 0.132 (0.001) 0.146 (0.001)

10.065 (0.002) 0.074 (0.002) 0.081 (0.002) 0.086 (0.002) 0.090 (0.002)

4320.055 (0.002) 0.055 (0.002) 0.066 (0.002) 0.069 (0.002) 0.073 (0.002)

40.054 (0.002) 0.053 (0.002) 0.056 (0.002) 0.057 (0.002) 0.059 (0.002)

Kb=1, Kw=0, DEPTH =1, all ﬁrms are orthodox. Each ﬁgure is an average over 10,000 ﬁrms on 1,000 families of landscapes.

Each number in parentheses is the standard deviation of the average ﬁgure to the left of it.

Table 4. Effect of depth of experience on analogizer’s long-run performance advantage

Long-run performance advantage DEPTH =

over a local searcher for an

analogizer with representation ... 0.01 0.10 0.25 0.50 1.00

1230.119 (0.001) 0.141 (0.001) 0.146 (0.001) 0.146 (0.001) 0.146 (0.001)

10.083 (0.001) 0.090 (0.001) 0.093 (0.001) 0.093 (0.001) 0.092 (0.001)

4320.060 (0.001) 0.072 (0.001) 0.072 (0.001) 0.071 (0.001) 0.072 (0.001)

40.054 (0.001) 0.059 (0.001) 0.062 (0.001) 0.061 (0.001) 0.060 (0.001)

Kb=1, Kw=0, BREADTH =1, all ﬁrms are orthodox. Each ﬁgure is an average over 20,000 ﬁrms on 2,000 families of landscapes.

Each number in parentheses is the standard deviation of the average ﬁgure to the left of it.

The Power of Analogy 705

Figure 3. Performance differential due to orthodoxy

solutions. To size up this danger, we repeat the

experiment shown in Figure 2, but now allow

ﬁrms to be heterodox; that is, when improving

incrementally, ﬁrms are permitted to make changes

in detailed decisions that alter high-level policies.

We then plot in Figure 3 the difference between

the performance of the orthodox ﬁrm and the

performance of the corresponding heterodox ﬁrm.

One can think of this difference as the hidden cost

a ﬁrm bears if it takes its candidate solution too

seriously. The ﬁgure shows that this cost depends

sensitively on the quality of a ﬁrm’s representation.

Firms with very good representations incur no

costs of orthodoxy; they are in such promising

parts of the target landscapes that they can perform

well without crossing policy borders. Heterodox

ﬁrms have little incentive to cross policy borders,

so the search behaviors of orthodox and heterodox

ﬁrms are effectively the same. Firms with poorer

representations, in contrast, pay a heavy price for

holding on too ﬁrmly to their analogies. Indeed,

in results not reported here, we ﬁnd that ﬁrms

that take poor analogies seriously typically fare

worse than heterodox local searchers; the penalty

of orthodoxy outweighs the beneﬁt of analogical

guidance.

Structural characteristics

The power of analogy depends not only on the

representations, experience, and orthodoxy of the

management team, but also on the structure of

the target landscape. We explore the effect of

both ‘vertical’ and ‘horizontal’ structural changes

on our results. By vertical structure, we mean

the relationship between the number of detailed

decisions per policy (D) relative to the number of

higher-order policies (P). By horizontal structure,

we mean the intensity of interaction within each

policy domain (Kw) and the intensity of interaction

across domains (Kb).

To explore vertical structure, we compare the

performance of analogizing ﬁrms to that of local

searchers in two settings: P=5, D=3vs.P=3,

D=5. The total number of detailed decisions

is 15 in both cases, but the high-level guidance

given by analogy is more thorough when policy

domains are numerous and narrow (P=5, D=

3). Analogical reasoning is more powerful in

this situation, as the results in Table 5 show.

Comparing across rows in Table 5, we see that

the differential created by more thorough guidance

diminishes as the quality of the representation

deteriorates. Detailed guidance is less valuable if

it is based on misleading keys.

To explore horizontal structure, we ﬁrst examine

the effects of Kband Kwon landscape topography

and then turn to their effects on the performance

of analogizing ﬁrms. One can think of each

landscape as consisting of 2Ppolicy domains,

with 2Dconﬁgurations of detailed decisions within

each domain. Tables 6a and 6b reveal typical

topography within and across policy domains, as

a function of Kband Kw. To construct the tables,

706 G. Gavetti, D. A. Levinthal and J. W. Rivkin

Table 5. Effect of policy breadth on analogizer’s long-

run performance advantage

Long-run performance

advantage over a local

searcher for an analogizer

with representation ...

P=5

D=3

P=3

D=5

Differential

1230.162 0.109 0.053∗

(0.002) (0.002)

10.102 0.073 0.029∗

(0.002) (0.002)

4320.085 0.054 0.031∗

(0.002) (0.002)

40.066 0.048 0.018∗

(0.002) (0.002)

Kb=1, Kw=0, BREADTH =DEPTH =1; all ﬁrms are orthodox.

Each ﬁgure is an average over 3000 ﬁrms on 300 families of

landscapes. Each number in parentheses is the standard deviation

of the average ﬁgure above it. ∗Indicates that a differential is

statistically signiﬁcant at the 0.1% level.

Table 6a. Absolute change in performance associated

with a within-policy change in a detailed decision

Kw=0Kw=0.5Kw=1

Kb=0 0.056 0.074 0.087

Kb=0.5 0.051 0.069 0.085

Kb=1 0.050 0.068 0.083

P=3, D=3; each ﬁgure an average over 40 ﬁrms walking at

random for 125 periods on 100 landscapes.

Table 6b. Standard deviation across average per-

formance in each policy domain for a typical landscape

Kw=0Kw=0.5Kw=1

Kb=0 0.069 0.060 0.055

Kb=0.5 0.092 0.081 0.061

Kb=1 0.110 0.088 0.071

P=3, D=3; each ﬁgure an average over 40 ﬁrms walking at

random for 125 periods on 100 landscapes.

we release ﬁrms in each policy domain on a large

number of landscapes, let each walk at random

within a domain, and record performance during

the random walk. Table 6a shows the average

absolute value of the performance change caused

by altering one detailed decision within a policy

domain; this reﬂects within-policy ruggedness.

Table 6b reports the standard deviation across the

average level of performance within each domain

for the typical landscape; this captures across-

policy ruggedness. Within-policy ruggedness is

driven upward by increases in Kw, while across-

policy ruggedness is boosted by Kband mitigated

by Kw.

Together, Tables 6a and 6b paint a clear picture

of landscape topography. When Kb=Kw=0,

there are no interactions among decisions and the

landscape is relatively smooth, with changes in

individual decisions having relatively little impact

on overall performance. As Kbrises with Kwstill

0, the surface becomes more rugged. In particular,

plateaus deﬁned by policy conﬁgurations arise.

Within a policy conﬁguration, the surface is

relatively smooth; a change in a detailed choice

that does not change a policy alters only the

contribution of that decision in isolation. In

contrast, a change in a detailed choice that does

alter a policy—a step over the edge of the

plateau—causes many contributions to change.

The result is a surface with internally smooth

plateaus of quite different elevations. In contrast,

as Kwrises with Kbﬁxed, each plateau becomes

internally rugged.

This sets the stage for understanding the effects

of Kband Kwon analogical reasoning. The surface

plot in Figure 4 shows, for various combinations of

Kband Kw, the long-run performance differential

between an analogizing ﬁrm with representation

123and a local searcher. (P=3, D=3, and

DEPTH =BREADTH =1.) When Kb=Kw=0, the

landscape is smooth, both the analogizer and

the local searcher fare well, and as a result the

differential is modest. Indeed, a differential exists

only because ﬁrms are constrained to abide by their

original policies. If ﬁrms were heterodox, both

would attain the global peak, and the differential

would be precisely zero.9As Kbrises with Kw

9While it is true that analogy offers no long-run advantage

when target landscapes are smooth (Kb=Kw=0) and ﬁrms

are heterodox, this outcome masks important dynamics. Though

all ﬁrms eventually scale the global peak in such a situation,

analogical reasoning helps ﬁrms with good representations to

do so quickly. As a result, analogizing ﬁrms enjoy a transient

advantage. The better is the ﬁrm’s representation, the bigger

is this advantage. (Simulation results that illustrate this effect

are available from the authors on request.) There are numerous,

unmodeled reasons to believe that, in reality, such an advantage

might be of lasting importance. Discovering a strong competitive

position ﬁrst (Lieberman and Montgomery, 1988) may be

especially important if the ﬁrst ﬁrm to arrive at the position

can deter others from copying its conﬁguration of choices; if

customers are loyal and reluctant to switch once others reproduce

the position; if the selection environment is so vigorous that slow

The Power of Analogy 707

Figure 4. Long-run advantage of analogizer with representation 123as a function of Kwand Kb

low, internally smooth plateaus of different heights

emerge on the target landscape. In such a setting,

it is quite valuable to start one’s local search

in a favorable policy domain—on a promising

plateau. Accordingly, analogy is very powerful for

high Kband low Kw. In contrast, an increase in

Kwundermines the power of analogy: interactions

within policy domains make each plateau rugged,

and even a ﬁrm with a good representation,

which begins its local search in a favorable policy

domain, is likely to get stuck on a low local

peak. Good analogies in our model give policy-

level guidance, but if guidance at that level is not

sufﬁciently speciﬁc to lead a ﬁrm to success, then

analogical reasoning loses much of its force.

Overall, we ﬁnd that analogy is least powerful

in fully decomposed settings (Kb=0,K

w=1) or

nearly decomposed settings (Kblow, Kwhigh);

in such settings, the primary challenge facing

management is to achieve success within each of

several, fairly independent policy domains, and

analogy—at least as we have modeled it —offers

little toward this end. Analogical reasoning is also

of limited use in situations of full independence

(Kb=Kw=0), where incremental logic alone

will deliver a good conﬁguration of choices,

movers don’t survive their temporary disadvantage; or if the

environment changes so rapidly that one must reap the beneﬁts

of an advantage quickly. We speculate that analogical reasoning

might be especially valuable in such settings, even in the absence

of interactions among decisions.

and in situations of very high interdependence of

detailed decisions (Kw=1), where proliferating

local peaks trap analogizers and local searchers

alike. Analogy is most powerful when cross-policy

interactions dominate (Kbhigh, Kwlow)—that is,

when the plateaus deﬁned by policy conﬁgurations

are internally smooth, but heights are sufﬁciently

varied across plateaus that there is a danger

of being stranded on a low plateau. A number

of scholars have argued that piece-by-piece,

subsystem-by-subsystem learning is most effective

in nearly decomposable systems (Simon, 1962;

Baldwin and Clark, 2000; Frenken, Marengo,

and Valente, 1999). Our results suggest that

analogical reasoning offers its greatest relative

beneﬁt precisely where these approaches fail, in

systems that are not nearly decomposable.

DISCUSSION AND CONCLUSION

Discovering an effective competitive position is

a difﬁcult endeavor. It is difﬁcult because the

mapping from strategic decisions to performance

is typically complex and, especially in novel

settings, unknown to the decision-makers. Though

cognizant of such difﬁculties, positioning scholars

emphasize the role of deductive reasoning and

rational choice in the origin of positions (Porter,

1996; Ghemawat, 1999). In contrast, evolutionary

theorists highlight the bounds of individual

708 G. Gavetti, D. A. Levinthal and J. W. Rivkin

rationality and posit that effective positions emerge

through a mix of luck and experiential, local

search, thus leaving little space for the cognition

of managers (Nelson and Winter, 1982). Although

recognizing the merits of both such perspectives,

we hesitate to ascribe all effective positions

to either the omniscience assumed in economic

analysis or the myopia of experiential learning.

We put forth a model of the origin of strategies

that, while recognizing the limits of managerial

rationality and the intelligence of local search,

also respects the power of managerial cognition.

Indeed, the notion of bounded rationality, which

we take as a cornerstone of our conceptual

apparatus, does not rule out the possibility of

intendedly rational choice (March and Simon,

1958). Bounded rationality suggests that thinking

is typically premised on simpliﬁed cognitive

representations of the world (Simon, 1991).

As boundedly rational actors, managers create

cognitive simpliﬁcations of their decision problems

and come up with solutions on the basis of

such simpliﬁcations. These solutions, in turn, may

imprint subsequent efforts at local search, thus

playing a central role in the discovery of strategic

positions (Gavetti and Levinthal, 2000). This

perspective, which represents a middle ground

between positioning and evolutionary arguments,

suggests that the roots of superior competitive

positions may lie in the cognition of managers,

particularly in the way they represent the world.

The effect of cognition on managerial action

is an enormous area of inquiry. We cut into this

topic by looking at a particular, important type of

action: the creation of a competitive position in an

unfamiliar industry. This type of action confronts

managers with daunting complexity so it is natural

to expect cognition—a fundamental instrument to

handle complexity—to play a major role. It also

forces decision-makers to cope with novelty. In

novel situations, wisdom from prior experience in

other contexts can be particularly powerful. For

this reason, we focus on a particular vehicle for

transporting experience across contexts: analogy.

Our conceptual model of analogical reasoning

is straightforward. We surmise that, in any set

of industries, a large number of underlying

characteristics drive the relationship between ﬁrm

action and performance. There are so many

characteristics and their effects are so difﬁcult

to discern that boundedly rational actors focus

their reasoning efforts on a subset of the

characteristics. These subsets form representations.

Some representations are more effective bases of

reasoning than are others. A good representation

distinguishes sets of similar payoff functions

from one another, while a poor representation

leaves very different payoff functions in the same

category. In other words, a representation is a

classiﬁcation scheme. An effective scheme puts

similar objects in the same class and different

objects in distinct classes.

Armed with a good representation and adequate

experience, a ﬁrm is well prepared to draw a can-

didate solution from a germane source and apply

it to a target industry. Accordingly, our simulation

model shows the best performance among ﬁrms

with high-quality representations—good classiﬁ-

cation schemes. That said, we also ﬁnd it better to

take guidance from some source, even one based

on a poor representation, than to start one’s local

search at a randomly assigned conﬁguration. There

is some chance that, purely by luck, the source will

prove to be a good one even though it is based

on a poor representation. This power of ‘false’

analogies is reminiscent of Weick’s (1990) tale of

a Hungarian military reconnaissance unit. Lost in

the snowy Alps, the troops prepare for the worst.

Then one soldier discovers a map in his pocket.

Once equipped with the map, the unit outlasts the

storm, ﬁnds its bearings, and returns to safety.

Only later does a commanding ofﬁcer realize that

the map is of the Pyrenees, not the Alps. The tale

is often interpreted in terms of presence of mind

and motivation: the map calmed the troops and

moved them to take coordinated action. We would

suggest a second possibility: perhaps parts of the

Pyrenees and parts of the Alps truly do resem-

ble one another. The soldiers may have received

a good candidate solution, though only by chance,

just as do some (though not all) of our ﬁrms with

poor representations. In business also, it is pos-

sible to have a poor representation yet obtain a

good candidate solution. For instance, Lycos’s par-

tial integration of Tripod worked out well, but even

some supporters of the decision felt the analogical

reasoning itself was dubious (Gavetti and Rivkin,

2004). An intriguing avenue for future research is

to examine what makes some analogies useful even

if based on similarity along irrelevant dimensions.

We suspect the key is to focus on dimensions that,

even if irrelevant, are correlated with true drivers

of success.

The Power of Analogy 709

Our results also shed light on the roles of

broad and deep experience. We ﬁnd breadth

and depth of experience to be valuable only

if a manager has a good representation —a

valid system for categorizing environments and

classifying lessons learned. In addition, beyond

a modest level of depth, performance is not

sensitive to depth. As a result, if top management

of a diversiﬁed ﬁrm believes that business unit

managers have a good grasp of what factors

drive the choice–performance relationship in their

particular business, they should be more willing to

rotate managers across divisions, to invest in and

exploit breadth of experience. Otherwise, it may

be better to keep each business unit manager in

a single business for a longer period, to develop

depth.

A possibility we do not model, worthy of

future research, is that experience in a wide

range of industries may help a manager to build

an understanding of what drives the relationship

between choice and performance. Put simply,

breadth may improve representations. As business

school instructors who teach by the case method,

we are sympathetic to this perspective. Much of

what goes on in business strategy courses, we

believe, is representation building. We aim to focus

student attention on dimensions of environments

that shape the relationship between action and

outcome, and we do so in part by exposing

students to a broad variety of cases. A framework

such as the Five Forces (Porter, 1980) is helpful

because it synthesizes across cases and identiﬁes

the dimensions that are most salient across the

range of particulars. In doing so, the framework

enhances pattern recognition. Frameworks that

classify environments and ﬁrms along different

dimensions—say, by the durability of underlying

resources (Williams, 1992) or by the character of

capabilities (Teece, Pisano, and Shuen 1997) rather

than by the nature of competitive forces—might

lead managers to see very different patterns.

More broadly, the pervasive impact of repre-

sentations in our ﬁndings afﬁrms the importance

of research that examines the mental models of

strategists (e.g., Porac, Thomas, and Baden-Fuller,

1989; Huff, 1990; Huff and Jenkins, 2002). Espe-

cially intriguing is the question of where repre-

sentations come from (other than business school

curricula). Future empirical efforts might exam-

ine the processes by which management teams

develop, track, and alter their beliefs about what

characteristics distinguish similar settings from dif-

ferent ones. Modeling efforts might expose simu-

lated ﬁrms to a series of decision problems, not the

single problem we model here, and allow manage-

ment teams to alter their representations as they

update their beliefs about the true XPROB . Models

might also incorporate shocks to XPROB . Our spec-

ulation is that established teams with solid beliefs

about XPROB will ﬁnd such shocks especially haz-

ardous. Such a team may very well persist in pay-

ing attention to observable characteristics that used

to be relevant but are no longer. As a result, they

may draw analogies based on similarities that have

become superﬁcial.

Assessing the role and performance implications

of analogies also requires analyzing how analogies

are used. In the context of our model, we

distinguish between orthodox and heterodox uses

of analogy. Not surprisingly, orthodoxy is most

costly for ﬁrms with poor representations. More

surprisingly, orthodoxy provides no advantage

over heterodoxy even when the analogy is based on

a good representation. Good representations seed

subsequent search efforts on such promising areas

of the landscape that heterodox analogizers are

not motivated to violate the policies identiﬁed by

the analogy. This result suggests that analogy is

more effective as an instrument to seed search

efforts than as a means to constrain them. The

psychological literature on analogy focuses on

representations and experience as central elements

underlying the quality of analogies (Holyoak and

Thagard, 1995). Particularly in the context of

organizational search efforts, we believe, that in

addition to these fundamental elements, managers

should pay explicit attention to the orthodoxy with

which analogies are used.

Our ﬁnal set of results shows that analogical

reasoning produces the greatest long-run advantage

over incremental local search in settings that

are poorly decomposed. In these contexts,

the high-level policy guidance offered by

analogy is necessary to identify consistent policy

conﬁgurations—policy conﬁgurations that cannot

be recreated via incremental search at the level of

detailed choices. Others have touted the virtues of

decomposable or nearly decomposable systems of

choices (Simon, 1962; Baldwin and Clark, 2000).

In nearly decomposable or modular systems, one

can learn by means of parallel experiments at

the subsystem level (Baldwin and Clark, 2000).

Moreover, boundedly rational managers might be

710 G. Gavetti, D. A. Levinthal and J. W. Rivkin

able to apply deductive reasoning at the subsystem

level even if the system as a whole outstrips

their processing abilities. However, not all systems

of choices are neatly decomposable. We suggest

that, for dealing with systems of choices that

are inherently non-decomposable or have not

yet been decomposed, reasoning by analogy

is particularly powerful. Good representations

underlie the transfer of powerful strategic solutions

from managers’ past experience. These solutions,

in turn, allow a rich appreciation of the architecture

of the strategic problem (Henderson and Clark,

1990). It is in poorly decomposed systems that

this kind of architectural wisdom, and therefore the

quality of the cognitive representations underlying

the analogy, play a particularly important role.

These results emphasize that analogy and, more

generally, cognition are especially powerful in

settings where parallel local search fails to guide

organizational adaptation effectively.

More broadly, our analyses may help us better

understand observed variation in the processes

that underlie the origins of successful strategic

positions. Our ﬁndings lead us to hypothesize,

for instance, that cases in which local search is

the main process guiding ﬁrms toward successful

positions may well reﬂect some select structural

characteristics of the industries where such

companies succeed. In particular, such settings are

likely to have intense interactions among decisions

within policies. Rich sets of interactions among

the detailed decisions lying below the surface of

higher-level policies limit the power of apriori

cognition in these settings. We believe that this

linkage between the structural characteristics of

the context and the mechanisms underlying the

development of successful positions is a crucial

one—one that future empirical studies should

consider carefully.

Cognition in complex worlds inevitably involves

simpliﬁcation. The precise basis of simpliﬁcation,

however, is not inevitable. As academics and

practitioners, we are often apologetic about

operating in the space of such simpliﬁcations—the

commonly ridiculed ‘two-by-twos’ of the strategy

ﬁeld. But as boundedly rational individuals, we

cannot think in high-dimensional spaces. The

relevant question is not whether we conceive

of complex strategic problems in terms of a

few overarching variables, but rather what those

variables will be. The choice of variables can

have a major impact on performance, both directly

and by altering the inﬂuence of other factors

such as managerial experience and orthodoxy.

Our ﬁndings support this perspective in the

case of analogical reasoning, and we speculate

that it will also hold true for other forms of

managerial cognition. Our hope is that rigorous

analysis of cognition will help bridge the

chasm between rational, positional perspectives on

strategy and behavioral, evolutionary approaches.

Understanding how ﬁrms identify effective

competitive positions requires both perspectives.

With the current work, we try to provide some

substantiation of that link and a platform on which

others can build.

ACKNOWLEDGEMENTS

For helpful comments, we are grateful to Nicolaj

Siggelkow, anonymous referees, and seminar

audiences at Babson, Harvard Business School,

MIT’s Sloan School, the University of Calgary,

UCLA, the University of Toronto, and the Wharton

School. We also thank Howard Brenner for

computer programming efforts, Simona Giorgi and

Elizabeth Johnson for research assistance, and the

Division of Research of Harvard Business School

for generous funding. Errors remain our own.

REFERENCES

Baldwin CY, Clark KB. 2000. Design Rules: The Power

of Modularity. MIT Press: Cambridge, MA.

Cyert R, March J. 1963. A Behavioral Theory of the Firm .

Prentice-Hall: Englewood Cliffs, NJ.

Fleming L, Sorenson O. 2001. Technology as a complex

adaptive system: evidence from patent data. Research

Policy 30: 1019–1039.

Fleming L, Sorenson O. 2004. Science as a map in

technological search. Strategic Management Journal ,

Special Issue 25(8–9): 909 –928.

Frenken K, Marengo L, Valente M. 1999. Interdependen-

cies, near-decomposability and adaptation. In Com-

putational Techniques for Modelling Learning in Eco-

nomics, Brenner T (ed). Kluwer: Dordrecht; 145 –165.

Gas Daily. 2000. Broadband to be as big as Enron in ﬁve

years. 5 April.

Gavetti G, Levinthal D. 2000. Looking forward and

looking backward: cognitive and experiential search.

Administrative Science Quarterly 45: 113 –137.

Gavetti G, Rivkin J. 2004. Rationality and plasticity over

time: toward a grounded theory of the origin of

strategies. Working paper, Harvard Business School.

Ghemawat P. 1999. Strategy and the Business Landscape.

Addison-Wesley: Reading, MA.

The Power of Analogy 711

Gick ML, Holyoak KJ. 1980. Analogical problem

solving. Cognitive Psychology 12: 306 –355.

Gilovich T. 1981. Seeing the past in the present: the

effect of associations to familiar events on judgments

and decisions. Journal of Personality and Social

Psychology 40: 797–808.

Gragg CI. 1940. Because wisdom can’t be told. Harvard

Alumni Bulletin 43(2): 78 –84.

Hast A (ed). 1992. International Directory of Company

Histories. St James Press: London.

Henderson RM, Clark KB. 1990. Architectural innova-

tion: the reconﬁguration of existing product technolo-

gies and the failure of established ﬁrms. Administrative

Science Quarterly 35: 9 –30.

Holyoak K, Thagard P. 1995. Mental Leaps: Analogy in

Creative Thought. MIT Press: Cambridge, MA.

Huff AS. 1990. Mapping Strategic Thought. Wiley: New

York .

Huff AS, Jenkins M. 2002. Mapping Strategic Knowl-

edge. Sage: Thousand Oaks, CA.

Kauffman SA. 1993. The Origins of Order: Self-

Organization and Selection in Evolution. Oxford

University Press: Oxford.

Kauffman SA. 1995. At Home in the Universe. Oxford

University Press: Oxford.

Levinthal D, March J. 1993. The myopia of learning.

Strategic Management Journal, Winter Special Issue

14: 95–112.

Levinthal D. 1997. Adaptation on rugged landscapes.

Management Science 43: 934 –950.

Lieberman MB, Montgomery DB. 1988. First-mover

advantages. Strategic Management Journal, Summer

Special Issue 9: 41–58.

March J, Simon HA. 1958. Organizations. Wiley: New

York .

McKelvey B. 1999. Avoiding complexity catastrophe

in coevolutionary pockets: strategies for rugged

landscapes. Organization Science 10: 294 –321.

Mintzberg H. 1978. Patterns in strategy formation.

Management Science 24: 934 –948.

Nelson R, Winter S. 1982. An Evolutionary Theory of

Economic Change. Belknap: Cambridge, MA.

Neustadt RE, May ER. 1986. Thinking in Time: The Uses

of History for Decision-Makers . Free Press: New York.

Pascale RT. 1984. Perspectives on strategy: the real

story behind Honda’s success. California Management

Review 26(3): 47–72.

Perkins EJ. 1999. Wall Street to Main Street. Cambridge

University Press: New York.

Porac JF, Thomas H, Baden-Fuller C. 1989. Competitive

groups as cognitive communities: the case of Scottish

knitwear manufacturers. Journal of Management

Studies 26: 397–416.

Porter ME. 1980. Competitive Strategy: Techniques for

Analyzing Industries and Competitors. Free Press:

New York.

Porter ME. 1985. Competitive Advantage: Creating and

Sustaining Superior Performance. Free Press: New

York .

Porter ME. 1996. What is strategy? Harvard Business

Review 74(6): 61–78.

Rivkin JW. 2000. Imitation of complex strategies.

Management Science 46: 824 –844.

Rivkin JW. 2001. Reproducing knowledge: replication

without imitation at moderate complexity. Organiza-

tion Science 12: 274 –293.

Rivkin JW, Siggelkow N. 2002. Organizational sticking

points on NK landscapes. Complexity 7: 31 –43.

Rivkin JW, Siggelkow N. 2003. Balancing search

and stability: interdependencies among elements

of organizational design. Management Science 49:

290–311.

Salter MS, Levesque LC, Ciampa M. 2002. Innovation

corrupted: the rise and fall of Enron. Harvard Business

School Working Paper 02– 102.

Siggelkow N. 2002. Evolution towards ﬁt. Administrative

Science Quarterly 47: 125 –159.

Simon HA. 1955. A behavioral model of rational choice.

Quarterly Journal of Economics 69: 99 –118.

Simon HA. 1962. The architecture of complexity.

Proceedings of the American Philosophical Society

106: 467–482.

Simon HA. 1991. Bounded rationality and organizational

learning. Organizational Science 2: 125 –134.

Slywotzky AJ, Morrison DJ. 1999. Proﬁt Patterns: 30

Ways to Anticipate and Proﬁt from Strategic Forces

Reshaping Your Business. Times Business: New York.

Stemberg TG. 1996. Staples for Success: From Business

Plan to Billion-dollar Business in Just a Decade.

Knowledge Exchange: Santa Monica, CA.

Teece DJ, Pisano G, Shuen A. 1997. Dynamic capabili-

ties and strategic management. Strategic Management

Journal 18(7): 509–533.

Thagard P. 1996. Mind: Introduction to Cognitive

Science. MIT Press: Cambridge, MA.

Weick K. 1990. Cartographic myths in organizations.

In Mapping Strategic Thought, Huff AS (ed). Wiley:

New York; 1–10.

Westhoff FH, Yarbrough BV, Yarbrough RM. 1996.

Complexity, organization, and Stuart Kauffman’s The

Origins of Order.Journal of Economic Behavior and

Organization 29: 1– 25.

Williams JR. 1992. How sustainable is your competitive

advantage? California Management Review 34(3):

29–51.

APPENDIX: DETAILED DESCRIPTION

OF HOW FAMILIES OF LANDSCAPES

ARE GENERATED

For a given set of values for P,D,X,Kw,Kb,

and XPROB, the computer generates an inﬂuence

matrix—a matrix that records which choices,

policies, and observable characteristics inﬂuence

the contribution of each detailed choice. Figure 1

shows what the matrix might look like for a case

in which P=3, D=3, X=4, Kw=1, Kb=

0.33, and XPROB =(1,0.5,0.5,0). A bullet (•)in

the matrix indicates that the column element of

712 G. Gavetti, D. A. Levinthal and J. W. Rivkin

the matrix inﬂuences the contribution of the row

element. The second row, for instance, indicates

that the contribution of the second detailed choice

is affected by the resolutions of choices 1, 2, and 3,

the resolution of policy 3, and the states of the ﬁrst

and third observable characteristics. Because Kw=

1, there is thorough interaction within each policy

domain. Because Kb=0.33, the typical decision

is affected by one-third of the other policies. The

ﬁrst observable characteristic is highly inﬂuential,

affecting the contributions of all detailed decisions

because the ﬁrst element of XPROB is 1. The fourth

observable characteristic is altogether irrelevant

since the fourth element of XPROB is 0.

Once the inﬂuence matrix is set, the computer

generates target and source landscapes. That

is, it assigns a pay-off to each of the 2P×D

possible conﬁgurations of choices on each of

the 2Xlandscapes. The contribution of each

detailed choice depends on the resolution of

that choice, the resolution of other choices and

policies, and the state of observable characteristics.

For each possible combination of the inﬂuential

factors, the computer draws a contribution at

random from a uniform U[0, 1] distribution.

In Figure 1, for instance, the second choice is

affected by six factors (as indicated by the

six bullets in the second row). Each of these

factors can be resolved in two ways, so the

computer draws 26possible contributions for the

second choice: one corresponding to each of the

possible conﬁgurations of the six factors. The

overall performance associated with a particular

conﬁguration of detailed choices and observable

characteristics is then the average over the P×D

contributions:

Performance(dgiven a set of observable

characteristics)=P×D



i=1

contribution of decision i

given dand observable characteristics(P ×D).

A Scientific Approach to Entrepreneurial Decision-making: an introduction and potential extension in China

Article

Full-text available

Jan 2022

We illustrate the scientific approach to entrepreneurial decision-making, which is an approach that begins with formulating problems clearly, developing theories about the implications of actions, and testing these theories using existing data or data drawn from well-defined experimental designs. After describing the mechanism through which the scientific approach to entrepreneurial decision-making operates, that is a more effective process of to form, test, and update beliefs about the worth of entrepreneurial ideas, we elaborate on the effects of the scientific approach on three key entrepreneurial outcomes: idea termination, pivoting, and performance. We then refer to the emerging theoretical, empirical, and practitioner-oriented literature on the scientific approach to decision-making, articulating its potential developments. We discuss the potential benefits and challenges of applying the scientific approach to the Chinese entrepreneurial context.

Reproducing and Extending Experiments in Behavioral Strategy with Large Language Models

Preprint

Full-text available

Oct 2024

In this study, we propose LLM agents as a novel approach in behavioral strategy research, complementing simulations and laboratory experiments to advance our understanding of cognitive processes in decision-making. Specifically, we reproduce a human laboratory experiment in behavioral strategy using large language model (LLM) generated agents and investigate how LLM agents compare to observed human behavior. Our results show that LLM agents effectively reproduce search behavior and decision-making comparable to humans. Extending our experiment, we analyze LLM agents' simulated "thoughts," discovering that more forward-looking thoughts correlate with favoring exploitation over exploration to maximize wealth. We show how this new approach can be leveraged in behavioral strategy research and address limitations.

Does effectiveness of internal corporate governance and top management influence corporate performance The role of ownership concentration

Article

Jan 2024

This study investigates the relationship between the efficacy of internal corporate governance and the performance of top management and the corporation. In addition, the current study investigates the moderating influence of ownership concentration on the link between the efficacy of internal corporate governance, top management, and corporate performance. This analysis included listed Saudi Arabian firms between 2014 and 2018. Moreover, this study utilised the OLS regression to examine the direct and indirect link. The findings of this study indicate that the efficacy of internal corporate governance and ownership concentration have a positive and statistically significant relationship with corporate performance (ROA). Furthermore, this study demonstrates that internal corporate governance effectiveness has a negative and statistically significant relationship with business performance (TQ). In addition, this research demonstrates that ownership concentration has a favourable correlation with corporate performance. This study is the first to examine how ownership concentration affects the relationship between efficacy of top management, corporate governance, and corporate performance. It demonstrates the importance of taking the ownership structure into account when assessing the efficiency of top management and corporate governance. In addition, the paper provides a comprehensive synthesis of top management and internal governance procedures

User Language and Cultural Product Innovation: Insights from the Global Mobile Gaming Industry

Article

Full-text available

Sep 2024
J INT BUS STUD

Traditional studies on cultural industries have emphasized localized innovation rooted in cultural products’ countries of origin. While this research acknowledges the significance of production origin, the digital transformation of cultural industries has shifted the landscape of cultural product innovation from traditional localized processes to a more globalized and democratic approach by engaging users worldwide. However, there is a notable gap in understanding demand-side variations in users from different countries, particularly as digitalization allows global users’ languages to convey diverse linguistic inputs. To address this gap, we investigate how users’ future-time reference (FTR) across countries influences the pace of cultural product innovation in the mobile gaming industry. Analyzing a global sample of 7,787 mobile games, we find that in countries where gamers predominantly use weak FTR languages, their communication exhibits proximate temporal framing, prompting the introduction of new gaming content at a faster pace. Further, the effectiveness of such FTR framing becomes more pronounced when publishers pay close attention to or are familiar with gamers’ languages. These findings contribute important insights for research on country of origin and cultural industries and for managers to better engage users to drive cultural product innovation.

Behavioral Strategy: Exploring Microfoundations of Competitive Advantage

Book

Nov 2024

Behavioral strategy has emerged as one of the most important currents in contemporary strategic management. But, what is it? Where does it come from? Why is it important? This Element provides a review of key streams in behavioral, interpreting behavioral strategy as a consistently microfoundational approach to strategy that is grounded in evidence-based insight in behaviors and interaction. We show that there is considerable room for furthering the microfoundations of behavioral strategy and point to research opportunities and methods that may realize this aim. The Element is of interest to strategy scholars in general, and to Ph.D. students in strategy research in particular.

Making business model decisions like scientists: Strategic commitment, uncertainty, and economic performance

Article

Jul 2024

Research Summary This study abductively investigates how a firm's degree of business model development—the extent to which strategic choices are crystallized—moderates the impact of a scientific approach to decision‐making on performance. We present findings from a field experiment involving 261 entrepreneurs, where treated entrepreneurs learn to apply a scientific approach, while control counterparts receive comparable content without this approach. Results show that the effect of scientific decision making varies with business model development. Treated entrepreneurs with higher degrees of business model development elaborated their theories of value focusing on lower‐level choices, achieving superior economic performance compared to controls. Conversely, treated entrepreneurs with lower levels of business model development reevaluated fundamental aspects, resulting in increased epistemic uncertainty and less favorable short‐term economic outcomes compared to controls. Managerial Abstract Using a field experiment with 261 entrepreneurs, we explored how the degree of business strategy definition influences the benefits of adopting a scientific approach to decision‐making. In the experiment, half of the entrepreneurs were taught to use a scientific approach for making decisions (the treated group), while the others received similar training without the scientific approach (the control group). Results show that treated entrepreneurs with already defined strategies benefited more, experiencing improved performance even in the short term. Conversely, treated entrepreneurs with strategies still under definition experienced more uncertainty and lower short‐term economic performance, as the scientific approach prompted them to reassess and adjust their core strategic decisions.

Strategy Theory Using Analogy: Rationale, Tools and Examples

Article

Oct 2024

We discuss why analogical reasoning is widely used in strategy formulation, implementation, and evaluation despite its assessment by many logicians as a weak form of logical argumentation. Given this popularity, we consider how analogies might be used to assist in strategic decision making to develop better strategy arguments, including those consistent with the theory-based view of strategy. We develop an extended analogy between the online platforms Tripadvisor and Glassdoor to illustrate our main points. We also describe in steps how to build and evaluate strategy analogies and we suggest why working with analogies may be helpful to those practicing the theory-based view of strategy.

Theory, Search, and Learning

Article

Oct 2024

Olav Sorenson

When searching for a solution to a problem, having a theory—an underlying causal structure that explains outcomes as a consequence of antecedents and that allows for the prediction of potential consequences of combinations of choices not yet tried—changes the way in which people explore the solution space. Whether a theory proves useful to search, however, depends not just on its predictive precision. This essay argues that the internal structures of theories—their size, complexity, the extent of their elaboration, and the confidence that their users have in the assumptions—also influences how people search for solutions and the efficiency of their search processes. It offers several conjectures about how theory and theory structure influence search and about which types of theories prove most useful to success.

External representations in strategic decision‐making: Understanding strategy's reliance on visuals

Article

May 2024

Research Summary External representations, particularly visuals, are important in strategic decision‐making. However, their pervasiveness and impact are not well understood in the strategy literature. Based on cognitive science research, we identify four cognitive functions crucial to strategic decision‐making that benefit from using external representations. We also propose a conceptual model and propositions that explain how the quality of strategic decision‐making depends on the interactions among task environment, external representations, and managers. We show that external representations influence in predictable ways the boundedly rational process of searching for new strategies. Key determinants include the manager's representational capability and the usability and malleability of the external representation. We discuss implications for users, designers, and teachers of external representations in strategy, as well as suggest avenues for future research. Managerial Summary This research points to the pivotal role of external representations, especially visuals, in strategic decision‐making. Drawing from cognitive science, this study identifies four critical cognitive functions that benefit from these external representations—working memory, long‐term memory, pattern recognition, and knowledge transfer. Further, the study highlights that external representations significantly influence the process of strategic decision‐making in predictable ways. Finally, we show that not all external representations are alike in their ease of use and a managers' ability to operate on an external representation, referred to as representational capability, greatly affects the decision‐making quality. The implications extend to users, designers, and educators of external representations, urging attention to the design and use of external representations for improved decision outcomes.

Advancing Small Business Inclusion in Public Procurement: Evidence From U.S. Federal Government R&D Contracts

Article

Sep 2024

To encourage small business inclusion in public procurement, the U.S. federal government has established the set-aside program that mandates government agencies to award a portion of their contracts to small businesses. We study whether and to what extent the performance of R&D contracts awarded through this program differs from those awarded through open competition. Analyzing a large dataset of federal R&D contracts, we find that despite restricting competition to small businesses, set-aside R&D contracts experience lower schedule and cost overrun than R&D contracts awarded through open competition. Furthermore, although set-aside R&D contracts experience lower schedule and cost overrun when they are awarded to more experienced contractor firms, this benefit arises primarily from a contractor firm's experience in executing R&D contracts across different agencies compared to the firm's experience with the same agency. Finally, set-aside R&D contracts awarded early in a fiscal year experience lower schedule and cost overrun than those awarded later. Post-hoc analysis examining the underlying dimensions of different-agency experience highlights the asymmetric effects of related-agency experience and unrelated-agency experience of contractor firms on the performance of set-aside R&D contracts awarded by the Department of Defense. While related-agency experience improves contract performance, unrelated-agency experience has a detrimental effect on contract performance. These findings demonstrate that small business inclusion policies may not necessarily compromise contract performance. Importantly, they emphasize the need for federal agencies and contracting officers to consider the underlying dimensions of contractor firm experience and contract award timing to improve contract performance and taxpayer money utilization.

Design Rules: The Power of Modularity

Book

Jan 2000

The Origins of Order: Self-Organization and Selection in Evolution

Book

Jun 1993

Stuart A Kauffman

Stuart Kauffman here presents a brilliant new paradigm for evolutionary biology, one that extends the basic concepts of Darwinian evolution to accommodate recent findings and perspectives from the fields of biology, physics, chemistry and mathematics. The book drives to the heart of the exciting debate on the origins of life and maintenance of order in complex biological systems. It focuses on the concept of self-organization: the spontaneous emergence of order widely observed throughout nature. Kauffman here argues that self-organization plays an important role in the emergence of life itself and may play as fundamental a role in shaping life's subsequent evolution as does the Darwinian process of natural selection. Yet until now no systematic effort has been made to incorporate the concept of self-organization into evolutionary theory. The construction requirements which permit complex systems to adapt remain poorly understood, as is the extent to which selection itself can yield systems able to adapt more successfully. This book explores these themes. It shows how complex systems, contrary to expectations, can spontaneously exhibit stunning degrees of order, and how this order, in turn, is essential for understanding the emergence and development of life on Earth. Topics include the new biotechnology of applied molecular evolution, with its important implications for developing new drugs and vaccines; the balance between order and chaos observed in many naturally occurring systems; new insights concerning the predictive power of statistical mechanics in biology; and other major issues. Indeed, the approaches investigated here may prove to be the new center around which biological science itself will evolve. The work is written for all those interested in the cutting edge of research in the life sciences.

The architecture of complexity

Article