Академический Документы
Профессиональный Документы
Культура Документы
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Marketing Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Marketing Research.
http://www.jstor.org
The past ten years have seen an explosion of experiments especially interestedin discovery and creation processes, or
and insights into new product development approaches. generative learning (Dougherty 1992; Imai, Nonaka, and
Many of the new viewpoints argue that knowledge assets Takeuchi 1985). Therefore,considerable work has focused
(Winter 1987) can be leveraged to achieve competitive ad- on ways organizationscan acquire better informationas a
vantage (Barabba and Zaltman 1991; Day 1994; Garvin means to discover new knowledge. We argue that an equal-
1993; Glazer 1991; Prahalad and Hamel 1990; Sinkula ly importantissue is the role of stored knowledge, or orga-
1994). Even more important,because it requiresthe use of nizationalmemory,in new productdevelopmentactivities.1
knowledge assets in a dynamic setting, scholars have in- As Starbuck(1992, p. 176) suggests, "A knowledge-inten-
creasingly envisioned productdevelopment as a process of sive firm may not be informationintensive ... knowledge is
organizationallearning involving the acquisition, dissemi- a stock of expertise, not a flow of information."
nation, and utilization of information(Day 1994; Dickson The traditionalbusiness strategy and marketingstrategy
1992; Imai, Nonaka, and Takeuchi 1985; Leonard-Barton literaturehas long emphasizedthe role of organizationalex-
1992; Moorman 1995; Nonaka 1991). Understandably,re- perience or familiaritywith productsand markets.This lit-
searchersexploring the creation of new productshave been eraturehas suggested that firms are likely to be more suc-
cessful if they stick to developing productsand marketsthat
reflect their core competencies (Ansoff 1988; Montoya-
*ChristineMoorman is Associate Professor of Marketing,and Anne S.
Miner is Associate Professor of Management,GraduateSchool of Busi- Weiss and Calatone 1994; Rumelt 1974; Varadarajan1983).
ness, University of Wisconsin-Madison.This researchhas benefited from One resulting framework of this view, termed the prod-
National Science FoundationGrant#SBR-9410419. The authorsthankthe uct/market matrix, is limited to examining the effect of
editor,and appreciatethe helpful comments of Jan Heide;Aric Rindfleisch; memory level on the financial performance of marketing
Thekla Rura;participantsof the WhartonConferenceon Innovationin New
ProductDevelopment;seminarparticipantsat DartmouthCollege, Harvard
University, IndianaUniversity,Ohio State University,Universityof South IThere are, as we discuss throughoutthis article, models of organiza-
Carolina, and University of Cincinnati; and three anonymous JMR tional informationprocessing that include the influence of organizational
reviewers. memory (Burgelman 1983; Cohen and Levinthal 1990, 1994; Cyert and
March 1963; Leonard-Barton1992).
Journal of MarketingResearch
91 Vol. XXXIV(February 1997), 91-106
(InternationalOrganizationfor Standardization)9000 rou- tional buyers, the less likely they were to engage in infor-
tines for documentationbecome standardprocedures.Imai, mation search activities.
Nonaka, and Takeuchi (1985), for example, describe the Organizationalmemoryalso varies in the degree to which
"rugbyapproach"used by Hondateams, which involves the it is dispersed, or shared, throughoutthe organization.As
entire team runningthe full length of the new productdevel- Walshand Ungson (1991, p. 62) note, "organizationalmem-
opment process, in contrast to a "relay approach,"which ory is not centrally stored, but distributedacross different
involves functions handingoff the productat distinct times. retention facilities." Organizationalmemory by its nature
Likewise, Orr(1990) observes informalinformationsharing involves some degree of dispersionthroughoutthe organiza-
routines among Xerox service representativeswho repeat- tion. However,there may still exist variancein the degree to
edly gather around a communal coffee machine to share which organizationalmembers adopt firm knowledge and
their field experience. skills, which is determined,in part, by how firm activities
Third, memory is found in an organization'sphysical ar- are designed and structuredto facilitate diffusion across the
tifacts, which embody, to varyingdegrees, the resultsof pri- organization(Nonaka and Nicosia 1979; Websterand Wind
or learning (Epple, Argote, and Devadas 1991; Garud and 1972). The presence of distinct organizationalsubcultures
Rappa 1994; Leonard-Barton1992). For example, Epple, suggests that memory is not necessarily sharedby all mem-
Argote, and Devadas (1991) provide evidence from an in- bers (Cohen and Levinthal 1990; Deshpande and Webster
vestigationof a truckplantthatknowledge may become em- 1989; Martinand Siehl 1983; Smircich 1983).2
bodied in tooling, programming,and assembly line layout. Organizationalmemoryalso varies in accessibility, or the
Othershave suggested that memory is reflected in organiza- extent to which it can be retrievedfor use (Day 1994; Garud
tional structureand ecology (Argote 1995; Leonard-Barton and Nayyar 1994; Walshand Ungson 1991). As Day (1991,
1992; Levitt and March 1988; March 1991; Walshand Ung- p. 8) notes, "Organizationswithoutpracticalmechanismsto
son 1991). Furthermore,in new productdevelopment,Imai, 'remember'what worked and why have to repeat their fail-
Nonaka,and Takeuchi(1985, p. 354-58) describe"a special ures and rediscover their success formulas over and over
corner within the factory where workerscould experiment," again. Memorymechanismsare needed to ensurethat useful
"holdingmeetings in a large room with glass walls,"and the lessons are captured,conserved,and can be readilyretrieved
use of a system in which "all the team membersare located when needed."
in one large room." Featuresof productsand productlines Finally,the contentof organizationalmemoryrefersto the
(such as product design, materials, packaging, and logos) meaningof collectively storedinformation(Walshand Ung-
are also importantphysical artifactsassociated with organi- son 1991). Increasingevidence points to memoryas consist-
zational memory. ing of two types of knowledge: proceduraland declarative
Roles. In all three forms, organizationalmemory is likely (Cohen 1991; El Sawy, Gomes, and Gonzalez 1986; Sinku-
to performtwo fundamentalroles: interpretationand action la 1994). Proceduralmemory refers to process memory or
guidance. Organizationalmemory performs an interpreta- memory of underlying skills for performingtasks (Nelson
tion role by filtering the way in which informationand ex- 1982). An organizationmay know, for example, how to de-
perience are categorized and sorted (Cohen and Levinthal velop prototypes.Declarativememoryrefers to the memory
1990; Daft and Weick 1984; Day 1994; Day and Nedungadi of concepts, facts, or events. Memory here might consist of
1994; Dutton and Jackson 1987; Jackson and Dutton 1988; knowledge aboutcustomerpreferences,or the technical fea-
Sinkula 1994; Walsh and Ungson 1991). Organizational tures of a firm's product line (Day and Nedungadi 1994).
memory also performsan action guidance role by dictating Anothertype of memory content could be found in the dis-
or influencing individual and group action (Amburgeyand tinct culturalcharacteristicsof an organization.Forexample,
Miner 1992; Cyert and March 1963; March and Simon memory content may reflect more clan, market,bureaucra-
1968; Suchman 1994; Walsh and Ungson 1991). Memory, cy, or adhocracy characteristics (Deshpande, Farley, and
for example, may contain a protocol for a new productde- Webster1993). Memorycontent is thereforelikely to be re-
velopment stage that guides team members'actions.The ac- vealed, in part,in an organization'sculture.3
tion guidance role representsone of the most powerful fea-
turesof organizationalmemoryin much traditionalresearch. A Definition of OrganizationalMemory
Cyert and March(1963), for example, emphasize the power Using this review as a basis, we define organizational
of standardoperating proceduresin driving organizational memoryas collective beliefs, behavioralroutines, or physi-
action, whereas Nelson and Winter(1982) stress the overar- cal artifactsthat vary in their content, level, dispersion, and
ching impact of organizationalroutines. accessibility.This view of memory is consistent with thatof
Characteristics. In addition to memory forms and roles,
organizational memory can be viewed as having several 2Althoughour approachfocuses on the degree of dispersion, dispersion
dimensions or characteristics:amount, dispersion, accessi- could be furtherconceptualized as a multidimensionalconstruct that also
bility, and content. The level, or amount, of organizational reflects the structureof that sharing.The structureof informationdistribu-
tion may include, for example, one-way or two-way transmissionsand hor-
memory refers to the amountof stored informationan orga- izontalor verticalstructures.Hence, we includea variablein our model that
nization has about a particularphenomenon. High levels of may partiallycontrol for the effects of the structureof information-sharing
experience in a product category or the accumulation of activities within the firms we studied.
knowledge or skills indicate higher levels of memory. An 3Following Deshpand6 and Webster (1989), culture is reflected in an
abundance of memory has been theorized to influence a organization'svalues and norms.Therefore,if culture is the source, mem-
firm's demand for new marketinformation(Dickson 1992; ory content will have a value or normcomponent.However,as conceptual-
ized, organizationalmemory is much broaderthan organizationalculture,
Sinkula 1994). Regarding this point, Weiss and Heide because, in additionto values and norms, memory includes behavioralrou-
(1993) find that the greaterthe priorexperience of organiza- tines and physical artifactsthat reflect priorlearning.
CONCEPTUALFRAMEWORK
Ourconceptualframeworkfocuses on how two organiza-
tional memory dimensions, level and dispersion, influence
the success of new products. These dimensions were
selected because, as we subsequently suggest, both have
been describedin the literatureas having a positive effect on
new product development activities. However, our frame-
work suggests that the effect of these memory factors
depends on what type of new product outcomes is being
examined and on whetherthe firm operates within a turbu-
lent environment.
In discussing this framework,we first introducethe two uct is partof the firm's longstandingrepertoire,the length of
focal new productoutcomes investigatedhere. Second, we team members'service is high, or a particularnew product
present the effect of organizationalmemory level and dis- development phase-such as prototypedevelopment-is a
persionon each new productoutcome. Third,we discuss the well-developed competency.
moderatingeffects of environmentalturbulenceon the orga- Previous research has observed that change becomes
nizationalmemory-newproductoutcome relationships. more difficult as memory in a particulardomain increases.
This effect has been referredto as a competencytrap (Levitt
Focal New ProductOutcomes and March 1988; March 1991), a core rigidity (Leonard-
The two new product outcomes that we investigate are Barton 1992), routinerigidity,orfunctionalfixedness (Dick-
new productcreativityand new productshort-termfinancial son 1992) for the organization.In the area of new product
performance.New productcreativityrefers to the degree to development,observers also have reportedfrom qualitative
which a new product is novel and has generative capacity studies thathigher levels of memory inhibitany actions out-
side preexisting action patterns (Ghemawat 1991; McDo-
(i.e., the potential to change thinking and practice)
(Andrews and Smith 1996; Wilton and Myers 1986; Zalt- nough 1993). Likewise, both Leonard-Barton(1992) and
man, Heffring, and LeMasters 1983). New product short- Dougherty (1992) describe instances in which groups with
termfinancial performance is defined as the level of new strongmemoriesare least able or likely to deviate from prior
action patternsduring new product development (see also
productprofitabilityand sales thatoccur within the first year
of introduction(Griffinand Page 1993; Montoya-Weissand March 1979).
Calantone 1994). These outcomes were selected because An importantalternativepossibility is that memory actu-
others have suggested that they are influenced by organiza- ally could enhance creativity.For example, researchon re-
tional-level information processes (Day 1991; Dickson lated topics suggests that some forms of creativitythrive in
the presenceof memory.For example, organizationalimpro-
1992; Glazer 1991; Jaworskiand Kohli 1993; Imai, Nonaka,
and Takeuchi 1985; Moorman 1995; Narver and Slater visation, which involves firms acting extemporaneously
withouta plan, has been describedas involving the recombi-
1990; Sinkula 1994). Moreover, there is often a tension
between the creativity and the short-termfinancial perfor- nationof routinesto producenovel outcomes (Weick 1993a,
mance of new products, because highly creative products b). More generally, researchon adaptationhas stressed the
recombinationof priorroutinesas a crucial source of novel
may have greaterpotentialfor short-termperformanceprob-
lems due to the difficulty of changing consumer or retailer activities (Holland 1975). Likewise, Cohen and Levinthal
acceptance of the product,while offering the possibility of (1990) find thatorganizationalmemory-as reflected in pri-
or research investments-can enhance an organization's
greaterlong-termfinancialgain given the possibility of their
revolutionizing a product category (Adams and Lacugna ability to assess and importnew outside information,which
could promotecreativity.Specifically, Cohen and Levinthal
1994; Andrews and Smith 1996; Kleinschmidtand Cooper
1991). These tensions are played out in the hypotheses that (1994, p. 237) suggest that "fortune favors the prepared
are depicted in Figure 1. firm"(see also Feldman 1989). Takentogether,these ideas
and data imply that high organizationalmemory could actu-
The Effects of OrganizationalMemoryon the Performance ally enhance creativityin new productdevelopment.
and Creativityof New Products Empiricalresearchon new productdevelopmentitself, on
balance, has tended to supportthe potentially negative im-
The effect of organizationalmemorylevel. A high level of
pact of memory on new product development creativity,
organizationalmemory would typically be present when a however.This leads us to hypothesize:
new productproject or action phase representsfamiliarter-
ritory,a new productrepresentsa modest change in an old memoryfor a new
Hi:The greaterthe level of organizational
project,the technological or customerbasis for a new prod- productdomain,thelowerthelevelof newproductcreativity.
The next question concerns the impact of organizational addition, some researchershave suggested that much orga-
memorylevel on the financialperformanceof new products. nizational innovation comes from recombiningroutines or
Theory and currentresearchimply that high levels of mem- ideas in new ways or by mixing routines that were previ-
ory, while inhibiting new product creativity, may enhance ously separate (Nelson 1982; Nonaka 1990). This line of
their short-termfinancial performanceby increasing effi- reasoningimplies that high memorydispersioncould inhib-
ciencies and the likelihood that previous successes will be it creativitybecause it would reduceheterogeneityin the or-
repeated(Cyert and March 1963; Duncan and Weiss 1979; ganization, which, in turn, restrictsthe numberof routines,
Walsh and Ungson 1991). Much new productdevelopment ideas, and competencies available for recombining or for
research shows that for many products, strong memory generating new actions. The early marginalized roles of
reduces the chances of poor outcomes by increasing effi- product champions for Post-It notes or Hitachi lasers em-
ciency and decreasing the chances of costly errors(Cooper body this idea (Garudand Nayyar 1994; Peters 1988).
and Kleinschmit 1986). This line of reasoning is consistent One way to reconcile these conflicting perspectives and
with the finding that higher performingnew productstypi- findings is to propose a curvilinear relationship between
cally have higher levels of marketingand technological syn- memory dispersion and new product creativity, in which
ergy between the new productand the firm's existing com- moderatelevels of dispersion promotethe highest levels of
petencies (Montoya-Wiess and Calatone 1994; Varadarajan new productcreativity.Moderatelevels are predictedto pro-
1983; Zirger and Maidique 1990). Therefore, together with mote the greatestcreativitybecause organizationshave both
Hl, this suggests that organizationalmemory level is likely the breadthand cross-fertilizationthat dispersion provides
to reduce the creativity of new products while increasing while maintaining some heterogeneity among members.
their short-termfinancial performance. Under these circumstances,members share a language and
understandingof problems and solutions but retain some
H2:The greater the level of organizationalmemory for a new
distinctive skills and knowledge. This view is supportedby
productdomain, the greaterthe new product short-termfi-
nancial performance. conceptual literature on group performancethat suggests
that too much diversity restricts communication but too
The effect of organizational memory dispersion. The much similaritymay restrictthe rangeof observationsavail-
effect of memory dispersion on new product outcomes is able for recombination(Katz and Allen 1982), and by Fiol
less clear on the basis of a review of the extant literature. (1994), who suggests that team diversity and unity jointly
Recall that memory dispersion refers to the extent to which promotehigher levels of collective learning.Underthis form
organizationalmemberssharean understandingof organiza- of the relationship,moderatelevels of dispersion have ele-
tional beliefs, behavioralroutines,and physical artifacts. ments of both heterogeneityand homogeneity and therefore
One stream of literaturesuggests that greater dispersion maximize new productcreativity.4
leads to more creativeand betterfinanciallyperformingnew
products.Forexample, Hutt,Reingen,and Ronchetto(1988) H3:Thereexists a curvilinearrelationshipbetween dispersionof
find that creative new productinitiatives are more likely to organizationalmemory for a new productdomain and new
product creativity such that moderate levels of dispersion
be characterized by a greater number of communication
produce the highest levels of new product creativity and
links between organizationalfunctions. Others point to the high and low levels of dispersion result in lower levels of
critical role of dispersing informationacross organizational new productcreativity.
functions, such as marketingand researchand development
(R&D), in the success of new product innovations (Gupta, Although a curvilinear relationship is expected for the
Raj, and Wilemon 1986; Moenaert and Souder 1990a, b). effect of memory dispersion on new product creativity,
This research suggests that the dispersion of memory memorydispersionis expected to have a positive linearrela-
enables functions to understandone another,improvestheir tionship with the short-termfinancial performanceof new
ability to cooperate, facilitates cross-fertilization,and may products, because high levels of dispersion increase the
reduce the tendency of individualfunctions to become con- effectiveness and efficiency of decision making and imple-
fined by theirown thought-worlds(Dougherty 1992; Souder mentation.In fact, as dispersion levels increase, the team's
1987). As lmai, Nonaka, and Takeuchi (1985, p. 544, mentalmodel becomes unified, which resultsin timely, cost-
emphasis added) note, "Project members are expected to effective decisions that help realize a firm's new product
interact with each other extensively, to share everything financial goals. Moreover, as was reviewed previously, the
from risk, responsibility, information,to decision making, literatureon information-sharingmechanismsin cross-func-
and to acquirebreadthof knowledgeand skills." How mem- tional efforts, such as total quality managementand quality
ory gets dispersed is not the subject of the presentresearch function deployment, suggests that shared knowledge and
(see Griffin and Hauser 1992, 1993, 1994; Hutt, Reingen, vision improve the short-term financial performance of
and Ronchetto 1988). However,when dispersed,collectively product development activities by enhancing cross-func-
held knowledge appearsto improve both the creativity and tional understandingand cooperation (Day 1994; Griffin
financial performanceof new products. and Hauser 1993; Hauserand Clausing 1988; Imai, Nonaka,
Another streamof researchsuggests that lack of memory and Takeuchi 1985), as well as by improvingteam efficien-
dispersion or heterogeneity within organizations should
have a positive effect on innovationand creativity (Burgel- 4It would also be reasonable to argue that the form of this curvilinear
man 1983; March 1991; Quinn 1986). From this perspec- relationshipshould be a U-shaped curve, in which high and low levels of
tive, groups with similar values, identical information, or dispersion promote the highest levels of creativity.Nevertheless, we chose
to integrateboth of the literaturesand in so doing, believe the best repre-
overlappingcompetencies should be less capable of produc- sentationof the relationshipis an inverted-U.This form suggests that mod-
ing actions thatdeviate from theirprioractivities thanwould erate dispersion has elements of both heterogeneityand homogeneity, and
more heterogeneous groups (Gigone and Hastie 1993). In thereforemaximizes new productcreativity.
cies in making decisions and taking action. This does not organizationalmemory (a) level and (b) dispersionand new
mean that the team is making the most creative decisions, productshort-termfinancialperformance.
which we believe will happen under moderate dispersion H6:The greaterthe marketturbulenceassociated with the envi-
levels (H3).Therefore,the creativity-dampeningrisks of too ronment,the weakerthe positive relationshipbetween orga-
nizational memory (a) level and (b) dispersion and new
much dispersion will not have the same impact on short-
productshort-termfinancialperformance.
term financial performance,because cross-functional effi-
ciencies are maximized, not compromised, in high disper- Althoughenvironmentalturbulencemay reduce the value
sion level groups. of organizationalmemoryfor performance,there is a poten-
H4:The greater the dispersion of organizationalmemory for a tially positive effect of environmentalturbulence for the
new productdomain,the greaterthe new productshort-term effect of memory levels on the developmentof creativenew
financialperformance. products.Specifically, if high organizationalmemory levels
reduce the potential for creative new products(Hi), a fast-
The ModeratingEffect of EnvironmentalTurbulence changing environment may attenuate this possibility
The turbulenceassociatedwith an organization'senviron- becausehigh levels of environmentalchange may act as trig-
ment is expected to moderate the effect of organizational gers to "unlearn"currentnew productroutines (Cyert and
memory on new productoutcomes. One of the most funda- March 1963; Hedberg 1981; Starbuck1976).
mental tenets in theories of organizationallearning holds Consideringthe relationshipbetween memory dispersion
thatthe value of organizationalmemoryis contingenton the and new productcreativity(H3), we suggest that organiza-
setting in which the organizationoperates(Argote and Epple tions may be betteroff with internalheterogeneityundertur-
1990; Cyert and March 1963; Levitt and March 1988). bulentconditions(Aldrich 1979; Lawrenceand Lorsch 1967;
Memory, after all, reflects learning from experience, and March1991). Underconditionsof internalheterogeneity(low
thatexperienceoccurs at a specific time in a specific setting. dispersion),firms can draw on previouslymarginalideas or
This insight is consistent with contingency theory's argu- competenciesthatmay act as crucialcreativeengines in times
ment thatbureaucraticstructures-which rigidly institution- of high turbulence(Burgelman1983; Feldman 1989; Miner
alize lessons from prior experience-can enhance perfor- 1990). For example, a firm producingvacuumtubes thathas
mance understable conditions,whereasmore organic struc- a well-dispersedtechnicalmemory(aboutvacuumtube tech-
tures are needed for turbulentconditions (Lawrence and nology) would be worse off when transistorsare discovered,
Lorsch 1967; Mintzberg 1979). than anotherfirm that has a less dispersedmemory,but in-
Several importantmechanisms may produce such envi- cludes a small deviantgroupof engineerswho are interested
ronmentaleffects. At a minimum, the value and impact of in transistortechnology. This reasoning implies that under
stored prior learning may deteriorate with environmental conditionsof high turbulence,we might expect high memory
change (Achrol 1991; Glazer 1991). As Weiss and Heide dispersion(high homogeneity) to have a negative effect on
(1993, p. 221) note, "a rapid pace of technological change creativity.Underconditionsof low turbulence,however,high
creates uncertainty that can be competency destroying" dispersionmay have a positive effect on creativity.This posi-
(Tushmanand Nelson 1990; see also Anderson and Tush- tive effect may accruefroma new productteam'sabilityto re-
man 1990; Tushman and Anderson 1986). Even more im- combinesharedknowledgeinto creativenew products(Borko
portant,memory may standin the way of effective action in and Livingston 1989; Dougherty 1990, 1992; Nelson 1982;
a turbulentenvironment,which restrictsthe organizationto Nonaka 1990;Weick 1993a, b). We thereforehypothesize,
inappropriateactions.
H7:The greaterthe technological turbulenceassociatedwith the
We focus on technological turbulence,that is, the degree environment,(a) the weaker the negative relationship be-
of change associated with new producttechnologies (Glazer tween organizationalmemorylevel and new productcreativ-
and Weiss 1993; Jaworskiand Kohli 1993; Weiss and Heide ity and (b) the weaker the positive relationshipbetween or-
1993), and marketturbulence,that is, the rate of change in ganizationalmemorydispersionand new productcreativity.
the composition of customersand theirpreferences(Jawors- H8:The greaterthe marketturbulenceassociated with the envi-
ki and Kohli 1993, p. 57). Both types of turbulence may ronment, (a) the weaker the negative relationshipbetween
have a potentiallydisruptiveeffect on memory'spositive ef- organizationalmemory level and new productcreativityand
fect on new product short-termfinancial performance,be- (b) the weaker the positive relationshipbetween organiza-
cause turbulenceis likely to reduce the value of priorlearn- tional memorydispersion and new productcreativity.
ing, which forces the organizationto search for and process
more information about the environment (Lawrence and METHOD
Lorsch 1967; Sinkula 1994; Weiss and Heide 1993). Rapid Sample and Procedure
environmentalchange also may stimulate metalearning,in The initial sample consisted of 396 firms in the 1992
which the people in organizationslearn to identify patterns
AdvertisingAge list of top 200 advertisers.After eliminating
of environmentalbehavior,butorganizationsin turbulenten- firms for which the questionnairewas inappropriate(i.e., no
vironments generally find it difficult to cope and survive. new productdevelopmentoccurred),the overall sample was
Given this view, we expect the positive effect of organiza- reduced from 396 to 300. Of the eligible sample, 92 firms
tional memory level (H2) and dispersion (H4) on the short-
(31%) responded.In terms of process, three weeks follow-
term financialperformanceof new productsto be weakened
ing the first mailing, nonrespondents were telephoned,
underconditionsof high technologicalor marketturbulence: remindedof the questionnaire,and encouragedto complete
associatedwiththe
turbulence
H5:Thegreaterthetechnological and returnit. Two weeks following the calls, a second mail-
environment,the weaker the positive relationshipbetween ing was sent to nonrespondents.No systematic differences
were found between those who respondedbefore and after variables:the two organizationalmemorymeasures,the two
the second mailing (Armstrongand Overton 1977).5 new productoutcomes, and the two environmentalmodera-
Vice presidentsof marketingwere used as informantsbe- tors. Because of the small sample size, this approachwas
cause of theirorganizationalknowledge and access to strate- chosen over examining all variables in a six-factor model,
gic and financial information (Aguilar 1967). Informant which violates the recommendationsmade by Bentler and
firm tenurelevels averaged 18 years, which is comparableto Cho (1988) to not exceed a five-to-one ratio of sample size
other samples of informants at this level (Larwood et al. to parameterestimates. Results suggest that the three mod-
1995).6When completing the survey,informantswere asked els fit well: the memory variables(X2(26)= 28.74, p = .323,
to focus on the most recentproductdevelopmentprojectthat goodness-of-fit index [GFI] = .938), the two new product
had been in the market for a minimum of 12 months for outcomes (X2(19)= 26.60, p = .114, GFI = .930), and the two
which their division was responsible. If new products did environmentalmoderators(X2(34)= 58.16, p = .006, GFI =
not fall under their purview, informantswere asked to for- .900).
ward the questionnaireto the appropriatevice president in Within these three models, discriminantvalidity was as-
theirdivision. All questions regardingthe organization,then, sessed by constrainingand freeing the phi coefficient. The
focused on the division as the unit of analysis. model with the free coefficient was found to be superiorto
the fixed coefficient for the two organizational memory
Measurement variables AX2(l) = 9.18, the two new product outcomes
Appendix A contains all of the measures and their AX2(1)= 11.34, and the two environmentalmoderatorsAX2(l)
sources. Memory level was operationalizedby measuring = 13.42, suggesting independentconstructs.In addition,be-
the amount of knowledge, experience, and familiarity an cause the memory measures are new, discriminantvalidity
organizationhas in a productcategory. Memory dispersion was assessed and found between the memory variablesand
level was measured by the degree of consensus or shared measures of individualmanageruse of information(Desh-
knowledge among new product participants.The assump- pandeand Zaltman1982), organizationaluse of information
tion underlying this approach is that when organizational (Moorman 1995), and differentforms of organizationalcul-
memory is dispersed, members' beliefs would intersect or ture(Deshpand6,Farley,and Webster1993).7Finally,the re-
converge on a particulartopic. If, for example, informants liability of the measureswas found to exceed standardsfor
noted that there was little consensus among people working acceptance.8In Table 1, we presentpsychometricinforma-
on the product,a reasonableconclusion would be that team tion and a correlationmatrixof all measures.
members' beliefs were based on different assumptions,
experiences, and information. Note that this approach to General TheoryTestingApproach
measuringmemory dispersion reflects the collective under- The hypotheses were examinedin two regressionmodels,
pinningsof organizationalmemory.That is, the defining ele- with the two new productoutcomes as dependentvariables.
ment-degree of convergence-is assessed as a propertyof Following accepted guidelines for examining interactions,
the collective. for each model, the main effects associated with the two
Technologicaland marketturbulencewere measuredwith organizational memory variables and turbulence were
Jaworskiand Kohli's (1993) operationalizations,which fo- enteredin additionto their interactioneffects (the productof
cus on the pace of technological change and customer the memory variables and the moderators).Following the
changes in the industry (see also Glazer and Weiss 1993; conceptual framework,a quadraticform of the memorydis-
Weiss and Heide 1993). Finally, organizationalbureaucrati- persion variable also was entered in the model with new
zation, which will be entered in the model to control for the productcreativityas the dependentvariable.The main effect
structureof informationdispersion, is defined as the degree variables were mean-centered before we constructed the
to which an organizationis managedthroughformalizedre- interactionsand quadraticversions to reduce the potential
lationshipsand centralizedauthority(Johnand Martin1984) effects of collinearity (Cronbach1987). Significant interac-
and is measuredwith Deshpand6's(1982) scales. tions were investigated with the slope analysis procedures
Following the datacollection, measureswere subjectedto specified in Aiken and West (1991) to improve understand-
a purificationprocess involving undimensionality,reliabili- ing of the coefficients. These proceduresenable significant
ty, and discriminant validity assessments (see Anderson relationshipsto be understoodat differentlevels of the con-
1987; Bagozzi and Phillips 1982; Churchill 1979; Gerbing tinuous moderator variables without creating categorical
and Anderson 1988). To assess unidimensionality,the mea-
sures were divided into three subsets of theoreticallyrelated
7Results indicate that the critical value (AX2(i)= 3.84) was exceeded in
all tests: memory level and individual instrumentaluse of information
5The results of these tests are the following (where ER = early respon- (AX2(1)= 32.18); memory dispersion and individual instrumentaluse of
ders and LR = late responders):memory level (ER = 5.29, LR = 5.17, t(91) information(Ax2(i) = 26.52); memory level and organizationalinstrumen-
= .34), memory dispersion (ER = 5.50, LR = 5.37, t91) = .69), technologi- tal use of information(Ax2(I)= 5.06); memory dispersion and organiza-
cal turbulencelevel (ER = 3.98, LR = 4.23, t(91)= -.76), new productper- tional instrumentaluse of information(AX2(I) = 9.79); memory level and
formance (ER = 4.90, LR = 4.65, t(91)= .71), and new productcreativity clan cultures (Ax2(I)= 21.58), marketcultures (AX2(I)= 4.46), adhocracy
(ER = 5.18, LR =4.65, t(91)= .71). cultures(AX2(I)= 11.82), and bureaucracycultures(AX2(}= 14.58); as well
61nformanttenurelevels were not collected duringthe initial administra- as memory dispersion and clan cultures (AX2(i)= 9.65), marketcultures
tion of the questionnaire.However, half of the organizationswere subse- (Ax2(i) = 10.16), adhocracycultures(Ax2(I)= 26.04), and bureaucracycul-
quently telephoned and this information was gathered as a safeguard to tures (AX2(l)= 35.05).
ensure that respondentshad enough organizationalexperience to be capa- 8The only exception, memory dispersion, also could be argued to be a
ble of assessing organizationalmemory,though recent evidence has found reflective, ratherthana formative,scale. This statuswould suggest thatcon-
an insignificant relationship between firm tenure levels and executives' ceptual considerationsregardingconstructspace coverage, and not reliabil-
articulationof their corporatevisions (Larwoodet al. 1995). ity assessments, should be the evaluativecriteria.
Table 1
MEASUREMENT
INFORMATION
versions. For both models, variance inflation factors were effect on new productcreativity(b = .204, t = 2.286) but no
estimated to examine collinearity and found to be below effect on new product short-term financial performance.
harmful levels (Mason and Perreault1991). In addition to Marketturbulencehas no main effects. Finally, the control
these predictedeffects, organizationalbureaucratizationalso variable,organizationalbureaucratization,has no effect on
was entered as a control variablein the models for the rea- the short-termfinancial performanceof new productsand a
sons described previously.Table 2 presents model estima- marginalnegative effect on their creativity (b = -.200, t =
tion results. 1.837)
RESULTS The Impactof OrganizationalMemoryLevel
Overview The first two hypotheses examine the effect of organiza-
Results show that, overall, the two models were signifi- tional memory level on new productoutcomes. Hl predicts
cant: new product short-termfinancial performance(R2 = that higher levels of organizational memory reduce new
.310, F(9,83)= 4.135, p = .0001) and new productcreativity productcreativity.Results indicatea nonsignificantrelation-
(R2 = .255, F(10,82)= 2.816, p = .0001). As was noted pre- ship between memory level and new product creativity,
viously, in testing the interaction hypotheses, the main which fails to support Hi, though the relationshipis in the
effects associated with the moderatorvariables(marketand expecteddirection(b = -.089, t = -1.123). H2predictsa pos-
technological turbulence)must be entered into the regres- itive effect for organizationalmemory level on new product
sion model (Pedhazur1982). Therefore,several nonhypoth- performance,which the resultssupport(b = .258, t = 2.787).
esized main effects also are noted. Results indicate that Considering the effect of technological turbulence,H5a
technological turbulence has a significant positive main and H6apredictthat the greaterthe technologicaland market
Table 2
ESTIMATES
STANDARDIZED OF HYPOTHESIZED
RELATIONSHIPS
Dependent Variables
New ProductShort-Term New Product
IndependentVariables Financial Performance Creativity
Prediction Actual Prediction Actual
OrganizationalMemory Level (H2, + ) .258* (.092) (Hi, -) -.089 (.080)
OrganizationalMemory Dispersion (H4, + ) .582* (.196) .418* (.184)a
Memory Dispersionx Memory Dispersion (H3, nc) .086 (.126)
Memory Level x TechnologicalTurbulence (Hsa,-) -.025 (.061) (H7a,-) -.013 (.053)
Memory Level x MarketTurbulence (H6a,-) .007 (.069) (Ha, -) -.017 (.060)
Memory Dispersion x TechnologicalTurbulence (H5b, -) .054 (.150) (H7b,-) -.402* (.135)
Memory Dispersion x MarketTurbulence (H6b,-) -.351* (.165) (H8b,-) -.156 (.145)
TechnologicalTurbulencea -.093 (.104) .204* (.089)
MarketTurbulencea .044 (.125) .062 (.109)
OrganizationalBureaucratizationb -.064 (.125) -.200t (.108)
Note: The degrees of freedom for the new productshort-termfinancialperformancemodel were (9,83), whereasthey were (10,82) for the new productcre-
ativity model. Standarderrorsare in parentheses.
*p < .05.
tp <.10.
aFollowing Pedhazur(1982), the main effects associated with both the interactionsand the quadraticterms must be entered into models examining inter-
action and quadratichypotheses.
bOrganizationalbureaucratizationis a control variablereflecting the structureof organizationalinformationsharingactivities.
cH3predictsan inverted-Urelationship.
while interpretingthese results. For this initial empirical to in-house designers and even vendors. Some firms also
study, we used single informantsat high levels of the orga- seek to institutionalize the new product development
nization to achieve a broad organizationalview. Although process itself throughefforts to achieve ISO 9000 certifica-
the use of multipleinformantdesigns remainsthe exception tion. Otherorganizationsare experimentingwith new orga-
in most marketingstudies, such an approachwould provide nizationalstructuresthataffect the natureand availabilityof
a better test in some respects, though such designs are not organizationalmemory (Womack,Jones, and Roos 1990).
without their methodological concerns. Despite these con- Our results support the importance of identifying which
cerns, future studies might profit from seeking multiple in- productoutcomes the firm seeks to enhance and attempting
formantsto enhance the validity of organizationalmemory to link these activities to memory in order to enhance spe-
measures(Bagozzi and Phillips 1982). cific outcomes over time.
In addition,thatour informantsassessed new productde- The results also supportthe importanceof careful atten-
velopment projects after their completion raises the poten- tion to the multiple dimensionsof new productoutcomes in
tial of a retrospectivejustification bias. This would occur if theoretical research. They are consistent with predictions
informants,knowingthe outcome of projects,tendedto give that memory may have different effects on different out-
responses for the independent variables consistent with comes, which reducesthe likelihood thatwe will find a sim-
their knowledge of the outcome. Our informantsprovided ple formula linking memory to new productoutcomes. For
their assessments of these variables in the context of other example, researchlinking memorydimensions with key ad-
measures, thus making it less likely they would pay atten- ditional outcomes, such as timeliness, long-range financial
tion to the congruenceof their assessments with new prod- outcomes, and whethera productbecomes a dominantde-
uct outcomes. Moreover,survey questions were designed to sign, may offer potentially important frontiers for such
focus informantattentionon the appropriatetime period for work (Foster 1986; Schoonhoven, Eisenhardt,and Lyman
each variable,in partto help avoid this effect. Nonetheless, 1990).
furtherwork could fruitfullyseek to measurememory vari-
ables before project outcomes are known to alleviate such The Effect of OrganizationalMemoryLevel
concerns. Our researchaugmentspreviousworkconcernedwith the
Finally,our datashow meaningfulvariancein termsof or- relationship between new products and a firm's existing
ganizationalmemory and project outcomes, which reduces competencies, which has often viewed knowledge assets as
concerns about limited scope in our sample. However, as- having unconditionally positive effects. For practitioners,
suming our sample of projects is representativeof product our results supporta greatdeal of the marketingstrategylit-
development projects in general, it is likely to contain a eratureand practice by finding that a reliance on memory
large proportionof projects involving the modification of (which representsstored informationand competencies) in
existing products.Therefore,our results may not be gener- new product development increases new product financial
alizable to more radical projects. Hence, an importantav- performance(Ansoff 1988; Montoya-Weiss and Calatone
enue for furtherwork would be to determine if our results 1994; Rumelt 1974; Varadarajan1983). However, we did
are replicated in samples with higher proportionsof more not find active supportfor the predictionthat high memory
radical projects. These issues notwithstanding,the study's levels would detractfrom productcreativity,but we did find
results offer interesting implications for both practitioners that high memory levels failed to enhancecreativity.
and marketingtheorists,which we next consider. Ourfindingthathigh memorylevel enhancedfinancialre-
turn but did not enhance creativity also reinforces practi-
New ProductOutcomes tioner concern about the possible dangers of formalizing
The effects of memory differ between the two new prod- new product development processes. Specifically, many
uct developmentoutcome dimensions studied here. In par- firms are now formalizingnew productdevelopmentproce-
ticular, memory level enhances relatively short-term(one dures,sometimes in pursuitof ISO 9000 certificationor sup-
year) financial performance,but not creativity.Markettur- plier qualificationprograms.Yet, to the degree a firm seeks
bulence moderates the impact of memory dispersion on to enhance new productcreativity,formalproceduresaimed
financial performance, whereas technological turbulence at increasingthe level of memory may have little or no val-
moderates the impact of memory dispersion on creativity, ue. These firms might want to look to other dimensions of
but not on short-termfinancial performance.This pattern their organizationalmemory (beyond just trying to capture
highlights thatmemory may have variedeffects on different the most informationpossible) in seeking to institutionalize
features of productperformance.There are additionalindi- their best practices.
cators of success in new product activities, including cus- Turningto theoreticalissues, thathigh memorylevels nei-
tomer measures and time-to-marketmeasures (Griffin and ther enhances nor detracts from product creativity in our
Page 1993). By implication,the effect of memory level and study leads us to speculate that memory level may be less
dispersion may vary for these other outcome indicators as importantthan how flexibly or inflexibly a firm holds its
well. knowledge (March 1979). Furtherresearchcould consider
One implicationof our work for practitioners,then, is to moderatingfactorsin the firm'scultureor structurethatmay
underscore the importanceof sensitivity to organizational reflect a flexible approachto what has been learned in the
memory's potentiallydistinct impact on differentnew prod- past or that encourage careful reconsiderationof current
uct outcomes. Many firms are in the process of creating in- routinesand knowledge(Barabbaand Zaltman1991; Olson,
creasingly sophisticatedorganizationalmemory systems in Walker,and Ruekert1995). In addition,the relationshipbe-
which they make engineeringdrawings,partsspecifications, tween memory level and creative productoutcomes may be
costs, and otherconcrete featuresof priorproductsavailable influenced by stage of the new product development
Overall, these results extend thinking that has tended to which new, interactiveinformationsystems and new organi-
advocate either homogeneity or heterogeneityin organiza- zational designs affect the balance between heterogeneity
tional knowledge. For manydecades, for example, it was as- and homogeneity in productdevelopment projects. Further
sumed that specialization,which can be seen as fragmented researchshould also probe how these dispersioneffects are
organizationalmemory,was an efficient way to captureand influenced by their occurrenceduring certain stages of the
use knowledge (Scott 1987). Today,popularwisdom has re- new productdevelopmentprocess. For example, it is widely
versed that assumption,often calling for sharedknowledge assumedthatheterogeneityis more useful in the early stages
and redundancyas a simple answer to producenew product of the process, when new knowledge appears to be more
success (e.g., Griffinand Hauser 1994; Nonaka 1990). Both necessary;however,we lack systematic empiricaldata sup-
our theoreticaldevelopmentand empiricalresults point to a portingthis view.
more complex world. Furtherresearchcould explore more Finally, in considering the effect of memory dispersion,
carefully the precise mechanismsthroughwhich consensus furtherresearchwould benefit from accounting for its spe-
and heterogeneityaffect outcomes in differentnew product cific content. We speculate that if a firm's memory is dis-
developmentenvironments(Guzzo and Salas 1995; Watch- persed, but contains primarilyproceduralknowledge about
er 1983). More broadly, researchersmay want to explore how to innovate, it could produce successful productseven
factorsbeyond variationamong persons,such as the ways in under turbulentconditions. However, we speculate that de-
Appendix A
STUDY MEASURES
11.ModeratorVariables IV ControlVariables
TechnologicalTurbulence Jaworskiand Kohli (1993)
OrganizationalBureaucratization AdaptedfromDeshpande(1982)
(Seven-pointscale, where 7 = strongly agree and I = stronglydisagree) (Seven-pointscale, where 7 = strongly agree and I = stronglydisagree)
*Thetechnology in this productarea is changing rapidly. *Wheneveremployees have a problem,they are supposedto go to the same
*Technologicalchanges provide big opportunitiesin this productarea.
person for an answer.
*Itis very difficultto forecast where the technology in this productareawill *Thereis little action taken until a superiorapprovesthe decision.
be in the next five years. *If employees wished to make their own decisions, they would be quickly
*A large numberof new productideas in this area have been made possible
discouraged.
throughtechnological breakthroughs.
*Going through the proper channels in getting a job done is constantly
*Technologicaldevelopmentsin this productarea are ratherminor.* stressed.
MarketTurbulence Jaworskiand Kohli (1993) *Employeeshave to ask their boss before they do almost anything
(Seven-point scale, where 7 = strongly agree and I = stronglydisagree) *Any decision employees make has to have their boss' approval.
*Thereis no specific rule manualdetailing what employees should do.*
*Inour kind of business, customers'productpreferenceschange quite a bit *In this organization,everyone has a specific job to do.
over time.
*Ourcustomers tend to look for new productsall the time.
*We are witnessing demand for our productsand services from customers V DiscriminatingVariables
who never bought them before.
(These variableswere used only in discriminantvalidity exercises, not in
*New customers tend to have product-relatedneeds that are differentfrom the model testing.)
those of our existing customers.
*Wecater to much the same customers that we used to in the past.* OrganizationalCulture D)eshpande,Farley,and Webster(1993)
OrganizationalInstrumentalUse of
MarketInformation Moorman(1995)
IndividualManager Instrumental
Use of MarketInformation Deshpandeand Zaltman(1982)
Strategic Perspective, Mel Horwich, ed. New York: Pergamon and Elaine Romanelli (1985), "OrganizationalEvolution:
Press, 167-83. A MetamorphosisModel of Convergenceand Reorientation,"in
Rumelt, RichardP. (1974), Strategy,Structureand Economic Per- Research in Organizational Behavior, Vol. 7, B. Staw and L.
formance. Cambridge,MA: HarvardUniversityPress. Cummings,eds. Greenwich,CT: JAI Press, 171-222.
Sandelands,Lloyd E. and Ralph E. Stablein(1987), "The Concept Varadarajan,Poondi (1983), "IntensiveGrowthOpportunities:An
of OrganizationMind,"in Researchin the Sociology of Organi- Extended Classification," California Management Review, 25
zations, Vol. 5, Samuel B. Barcharach,ed. Greenwich, CT: JAI (Spring), 118-32.
Press, Inc., 135-61. Walsh,JamesP. andGerardoRiveraUngson(1991), "Organizational
Schoonhoven,Claudia.B., Kathleen.M. Eisenhardt,and Katherine Memory,"Academyof ManagementReview, 16 (January),57-9 1.
Lyman(1990), "SpeedingProductsTo Market:WaitingTime to Watcher,J. (1983), "CreativeProblemSolving as a Result of Ma-
First Product Introductionin New Firms,"AdministrativeSci- jority Versus Minority Influence,"EuropeanJournal of Social
ence Quarterly,35 (March), 177-207. Psychology, 13 (January-March),45-55.
Scott, Richard(1987), Organizations:Rational,Natural,and Open Webster,FredandYoramWind (1972), OrganizationalBuyingBe-
Systems,2d ed. Englewood Cliffs, NJ: Prentice-Hall. havior. Englewood Cliffs, NJ: Prentice-Hall.
Seely Brown,John (1993), "KeynoteAddress:Seeing Differently," Weick, Karl E. (1979), "CognitiveProcesses in Organizations,"in
in Marketing Science Institute Report No. 93-103, Marjorie Researchin OrganizationalBehavior,Vol. 1, BarryN. Staw, ed.
Adams, ed. Cambridge,MA: MarketingScience Institute, 1-5. Greenwich,CT: JAI Press, 41-74.
Sinkula,James M. (1994), "MarketInformationProcessingand Or- (1993a), "OrganizationalRedesign as Improvisation,"in
ganizationalLearning," Journalof Marketing,58 (January),35-45. Organizational Change and Redesign, George P. Huber and
Smircich, Linda (1983), "Conceptsof Cultureand Organizational WilliamH. Glick, eds. Oxford:OxfordUniversityPress,346-79.
Analysis,"AdministrativeScience, 28 (September),339-58. (1993b), "Managingas Improvisation:Lessons from the
Souder, William E. (1987), Managing New Product Innovations. Worldof Jazz,"workingpaper,GraduateSchool of BusinessAd-
Lexington, MA: Lexington Books. ministration,Universityof Michigan.
Starbuck,William H. (1976), "Organizationsand their Environ- Weiss, Allen and Jan B. Heide (1993), "The Nature of Organiza-
ments,"in Handbookof Industrialand OrganizationalPsychol- tional Search in High Technology Markets,"Journal of Market-
ogy, M. D. Dunnette,ed. New York:RandMcNally, 1069-123. ing Research, 30 (May), 220-33.
(1992), "Learningby Knowledge-IntensiveFirms,"Jour- Wilton, Peter C. and John G. Meyers (1986), "Task,Expectancy,
nal of ManagementStudies, 29 (November),713-40. and InformationAssessment Effects in InformationUtilization
Suchman, Mark (1994), "On Advice of Counsel: Law Firms and Processes,"Journalof ConsumerResearch, 12 (March),469-86.
Venture Capital Funds as Information Intermediariesin the Winter,Sidney G. (1987), "Knowledgeand Competenceas Strate-
Structurationof Silicon Valley,"unpublisheddoctoral disserta- gic Assets," in The CompetitiveChallenge:Strategiesfor Indus-
tion, Sociology Department,StanfordUniversity. trial Innovation and Renewal, David J. Teece, ed. New York:
Teece, David J. (1987), "ProfitingFromTechnologicalInnovation: Harper& Row, 159-84.
Implicationsfor Integration,Collaboration,Licensing, and Pub- Womack, J., D. Jones, and D. Roos (1990), The Machine That
lic Policy," in The CompetitiveChallenge, David J. Teece, ed. Changed the World:The Story of Lean Production.New York:
New York:Harper& Row, 185-219. Rawson-Macmillian.
Tushman, Michael L. and Phil Anderson (1986), "Technological Zaltman,Gerald, Michael Heffring, and KarenLeMasters(1983),
Discontinuitiesand OrganizationEnvironments," Administrative TheoryConstructionin Marketing:Some Thoughtson Thinking.
Science Quarterly,31 (September),439-65. New York:John Wiley & Sons.
and Richard Nelson (1990), "Introduction:Technology, Zirger Billie Jo and Modesto A. Maidique (1990), "A Model of
Organizations,and Innovation,"AdministrativeScience Quar- New Product Development: An EmpiricalTest," Management
terly, 35 (March), 1-8. Science, 36 (July), 867-93.