consequences of new product development speed: a meta-analysis
TRANSCRIPT
Consequences of New Product Development Speed:A Meta-AnalysisPinar Cankurtaran, Fred Langerak, and Abbie Griffin
Five meta-analyses previously have been published on the topic of new product development involving the concept ofnew product development speed. Three of these studies have investigated antecedents to new product developmentsuccess, of which just one was new product development speed. The other two studies used new product developmentspeed as the dependent variable, and analyzed antecedents to achieving speed. This article extends previous empiricalgeneralizations in this domain by using a meta-analytic methodology to understand the link between new productdevelopment speed and new product success at a more granular level. Specifically, it considers the relationship withdifferent dimensions of success as measured overall or compositely, operationally (i.e., the process measures ofdecreasing development costs and proficiently managing market entry timing and the product measures of technicalproduct performance and product competitive advantage), and relative to external success outcomes (i.e., customerbased and financial success). While the results indicate that, in general, new product development speed is associatedwith improving success outcomes, those relationships may diminish or even disappear depending upon a number ofmethodological design decisions and research contexts. A subsequent meta-analysis of the antecedents of developmentspeed provides a more holistic picture of development speed. These results are broadly consistent with those producedby another recent meta-analytic investigation of the issue. Together, these findings have important implications foracademics pursuing further research in this domain, as well as for managers considering implementing a program toincrease new product development speed.
Introduction
M any firms have adopted time-based strategiesto increase innovation speed (e.g., Stalk,1988). Innovation speed, the ability to move
quickly from ideas to actual products (Kessler and Bierly,2002), is an important intermediate outcome of the newproduct development (NPD) process (Ali, Krapfel, andLaBahn, 1995). Increased speed is achieved by decreas-ing development cycle time, the elapsed time from thebeginning of idea generation to market introduction.Time-to-market (Tatikonda and Montoya-Weiss, 2001),product development time (Lilien and Yoon, 1989),innovation time (Mansfield, 1988), lead time (Ulrich,Sartorius, Pearson, and Jakiela, 1993), project comple-tion time (Terwiesch and Loch, 1999), and total time(Griffin, 1993) also denote the same concept.
A large stream of scholarly research has been devotedto identifying the drivers of faster product development,resulting in a lengthy list. In the first quantitative synthe-
sis of this research, Gerwin and Barrowman (2002)carried out a meta-analysis of how integrated NPD prac-tices bear on development time. They found that theextent of overlap and interaction between NPD activities,employment of technical tools and formal methods, andthe team leader’s organizational influence significantlyimpact development time. In their more recent meta-analysis, Chen, Damanpour, and Reilly (2010) includeda larger number of antecedents. They found that clearproject goals; process concurrency; number, and fre-quency of design iterations; effective leadership; teamexperience and dedication; and internal integration havethe greatest effect on speed. Although an overarchingtheory of development speed determinants is far fromarticulated, these meta-analyses have reconciled some ofthe previously inconsistent empirical findings.
Another stream of research has focused on the perfor-mance implications of development speed. While someaccounts portray speed as a key ingredient for creatingsuccessful new products, empirical evidence for thespeed–success relationship is, at best, mixed (Griffin,2002; Ittner and Larcker, 1997; Kessler and Bierly,2002). Specifically, there exists insufficient—and oftenconflicting—evidence regarding how speed relates to thedifferent dimensions of new product success, such as
Address correspondence to: Pinar Cankurtaran, Faculty of IndustrialDesign Engineering, Delft University of Technology, Landbergstraat 15,2628CE Delft, the Netherlands. E-mail: [email protected]. Tel: +3115 2781015.
J PROD INNOV MANAG 2013;30(3):465–486© 2013 Product Development & Management AssociationDOI: 10.1111/jpim.12011
development cost, product quality, market share, andprofitability. Since speed is not an end in itself, but ameans to ultimate product success, it is necessary toestablish first the character and magnitude of speed’srelationship with success prior to focusing attention on itsantecedents. The question in need of a definitive answeris not how faster product development can be achieved,but whether faster development contributes to NPDsuccess in the first place.
Three meta-analytic reviews of new product perfor-mance antecedents, which also include speed as an ante-cedent, have been published (Henard and Szymanski,2001; Montoya-Weiss and Calantone, 1994; Pattikawa,Verwaal, and Commandeur, 2006). Treating speed as oneaspect of the development process, all three reveal asmall-to-medium positive link between speed and perfor-mance. While these studies point to a speed-performancerelationship, speed is not the focal variable, and they donot provide detailed insight into how this relationship ismanifest for different dimensions of NPD performance.While Gerwin and Barrowman (2002) and Chen et al.(2010) adopt speed as the focal variable, it is the depen-dent variable, and thus a detailed investigation
of the speed–product success relationship remainsoutside their scope.
This study advances scholarly knowledge of the rela-tionship between development speed and new productsuccess using meta-analytic statistical techniques. Itdiffers from the above meta-analyses in that it (1) treatsdevelopment speed as the focal variable and (2) investi-gates in detail its antecedent relationships to the variousdifferent dimensions of new product performance. Toprovide a more holistic picture of development speed, italso presents a comprehensive meta-analysis of its ante-cedents using a larger database and a similar, yet morefine-grained, variable classification scheme than that usedby Chen et al. (2010). Table 1 illustrates the key differ-ences between this research and previous meta-analyticinvestigations involving development speed.
A detailed speed–success meta-analysis closes a majorgap in the literature because NPD success is a multifac-eted construct that can be defined in a variety of ways andmeasured with a diversity of indicators (Griffin and Page,1993, 1996). At least some of the conflicting speed–success relationship findings likely are due to employingdifferent success measures and variations in the theoreti-cal underpinnings offered to explain their relationshipswith speed. Interrelationships between the varioussuccess measures (Brown and Eisenhardt, 1995), asmanifest in the synergies and trade-offs between them(Cohen, Eliashberg, and Ho, 1996), further confound therelationship between speed and new product success.Hence, it is important to understand more precisely theway in which development speed relates to differentsuccess dimensions.
The next section presents the opposing viewpoints onthe speed–success relationship, details the ways in whichthis relationship is manifest for different measures ofnew product performance, and overviews the existingempirical work on this topic. The subsequent sectionexplains the data collection protocol and database forma-tion. This is followed by a description of the meta-analytic estimation procedures, which use the Pearsonproduct–moment correlation for effect size as it emergedas the most commonly reported effect size, and a presen-tation of main-effects results. The subsequent sectiondescribes the moderator analysis procedure and presentsthose results. Then the article proceeds with the maineffects and heterogeneity analysis results of the anteced-ents meta-analysis, albeit in less detail since a thoroughdiscussion of development speed antecedents is beyondthe scope of this study. It closes with a discussion of themain findings, limitations of the study, and futureresearch directions.
BIOGRAPHICAL SKETCHES
Ms. Pinar Cankurtaran is an assistant professor of new product market-ing at the Faculty of Industrial Design Engineering, Delft University ofTechnology in the Netherlands. She received a BSc degree in BusinessAdministration from Bilkent University in Turkey and MSc degrees inInternational Development and Management Research from the Univer-sity of Bath and the University of Oxford, respectively, in the UnitedKingdom. She received an MPhil in Business Research from the Rot-terdam School of Management at Erasmus University in the Nether-lands, and is currently working toward her PhD degree. Her researchinterests include new product development speed, organizational learn-ing, and strategic new product development. Her work has been pre-sented at PDMA 2008, Marketing Science 2010, and IPDMC 2011.
Dr. Fred Langerak is Professor of Product Development and Manage-ment in the Innovation, Technology Entrepreneurship & Marketinggroup in the School of Industrial Engineering at Eindhoven Universityof Technology in the Netherlands. He has an MSc and a PhD from theErasmus School of Economics. His research focuses on market-drivennew product development. He has published on these topics in suchjournals as International Journal of Research in Marketing, Journal ofProduct Innovation Management, Journal of Retailing, IEEE Transac-tions on Engineering Management, R&D Management, and IndustrialMarketing Management.
Dr. Abbie Griffin holds the Royal L. Garff Presidential Chair in Mar-keting at the University of Utah’s David Eccles School of Business. Shehas a BSChE from Purdue University, an MBA from Harvard, and a PhDfrom the Massachusetts Institute of Technology. Her research investi-gates means for measuring and improving the process of new productdevelopment. She was Editor of the Journal of Product InnovationManagement from 1998 to 2003 and is an avid quilter.
466 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
Tabl
e1.
Pre
viou
sM
eta-
Ana
lyti
cSt
udie
sIn
volv
ing
Dev
elop
men
tSp
eed
Stud
yM
onto
ya-W
eiss
and
Cal
anto
ne(1
994)
Hen
ard
and
Szym
ansk
i(2
001)
Patti
kaw
aet
al.(
2006
)G
erw
inan
dB
arro
wm
an(2
002)
Che
net
al.(
2010
)Pr
esen
tSt
udy
(201
0)
Foca
lva
riab
le(s
)N
ewpr
oduc
tpe
rfor
man
ceN
ewpr
oduc
tpe
rfor
man
ceN
ewpr
oduc
tpr
ojec
tpe
rfor
man
ceD
evel
opm
ent
time
and
goal
failu
reN
ewpr
oduc
tde
velo
pmen
tsp
eed
Dev
elop
men
tsp
eed
NPD
outc
ome
vari
able
san
alyz
edin
rela
tion
tode
velo
pmen
tsp
eed
New
prod
uct
perf
orm
ance
(ove
rall)
New
prod
uct
perf
orm
ance
(ove
rall)
New
prod
uct
proj
ect
perf
orm
ance
(ove
rall)
Non
e:•
Focu
son
ante
cede
nts
ofde
velo
pmen
ttim
ean
dgo
alfa
ilure
Non
e:•
Focu
son
ante
cede
nts
ofne
wpr
oduc
tde
velo
pmen
tsp
eed
7N
PDou
tcom
es:
•D
evel
opm
ent
cost
s•
Tech
nica
lqu
ality
•Pr
oduc
tad
vant
age
•M
arke
ten
try
timin
g•
Mar
ketp
lace
succ
ess
•Fi
nanc
ial
succ
ess
•N
ewpr
oduc
tsu
cces
s(o
vera
ll)N
umbe
rof
stud
ies
inda
taba
sere
port
ing
deve
lopm
ent
spee
d–N
PDou
tcom
ere
latio
nshi
pa
1(4
7)6
(41)
7(4
7)N
one
Non
e56
(56)
Ant
eced
ents
anal
yzed
inre
latio
nto
NPD
spee
dbN
one
Non
eN
one
6an
tece
dent
s(4
sets
):•
NPD
proc
ess
(2)
•Pr
oduc
tde
finiti
on(1
)•
Org
aniz
atio
nal
cont
ext
(1)
•Te
amin
g(2
)
17an
tece
dent
s(4
sets
):•
Stra
tegy
(3)
•Pr
ojec
t(2
)•
Proc
ess
(4)
•Te
am(8
)
42an
tece
dent
s(6
sets
):•
Prod
uct
(6)
•Pr
oces
s(9
)•
Team
(10)
•N
PDco
mpe
tenc
ies
(5)
•Fi
rm(8
)•
Env
iron
men
t(4
)N
umbe
rof
stud
ies
inda
taba
sere
port
ing
ante
cede
nt–d
evel
opm
ent
spee
dre
latio
nshi
pa
Non
eN
one
Non
e26
(26)
70(7
0)75
(75)
Eff
ects
corr
ecte
dfo
rar
tifac
ts?
No
Yes
Yes
Yes
Yes
Yes
Eff
ects
test
edfo
rm
oder
ator
s?N
oY
esN
oN
oY
esY
es
aTo
tal
num
ber
ofst
udie
sin
data
base
indi
cate
din
pare
nthe
ses;
bnu
mbe
rof
met
a-fa
ctor
san
alyz
edin
dica
ted
inpa
rent
hese
s.
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4672013;30(3):465–486
Development Speed and NewProduct Success
The increasing emphasis on speed is driven primarily bythe contention that faster product development bringsfirst-mover advantages (Dröge, Jayaram, and Vickery,2000; Lieberman and Montgomery, 1988; Stalk andHout, 1990). As first movers, firms can establish technol-ogy and industry standards (Meyer, 1993), preemptscarce resources and suppliers (Lee, Smith, Grimm, andSchomburg, 2000), gain a competitive edge over laterentrants (Chen, Reilly, and Lynn, 2005), and secure favor-able market positions (Smith and Reinertsen, 1991).Speed also allows firms to adapt more readily to dynamicbusiness environments and quickly address changes inconsumer demand (Eisenhardt and Tabrizi, 1995). Thesebenefits of faster product development are argued to setthe stage for new product success.
While several empirical studies have documented apositive effect of innovation speed on new productsuccess (Carbonell and Rodriguez-Escudero, 2006; Chenet al., 2005; González and Palacios, 2002; Johnson,Piccolotto, and Filippini, 2009; Kessler and Bierly, 2002;Lynn, Abel, Valentine, and Wright, 1999), it has notreceived unanimous empirical support. For example,neither Griffin (2002) nor Meyer and Utterback (1995)find any relationship between development time and per-formance. There also is evidence that the speed–successrelationship, when it does exist, may not be straightfor-ward due to contextual differences. Ittner and Larcker(1997), for instance, document a positive relationshipbetween average firm-level cycle time and perceivedoverall success for the computer industry but find no suchassociation for the automobile industry. Another sourceof disagreement pertains to the role of environmentaluncertainty on the speed–success link. While Kessler andBierly (2002) document a stronger association under con-ditions of low market uncertainty, Chen et al. (2005) findthe opposite. Given this lack of consensus, there is a clearneed for a systematic integration of findings in the formof a meta-analysis.
Dimensions of New Product Performance
New product success is a multidimensional construct thatcan be defined and measured at the firm, program, andproject levels (Griffin and Page, 1993). Indeed, whilemany studies adopt a “global” success measure suchas “overall performance” (Hoegl, Weinkauf, andGemünden, 2004) and “NPD effectiveness” (Rusinko,1999), others investigate the NPD speed relationship with
narrower subdimensions of NPD performance, revealingmarkedly different patterns of association. This literaturesuggests that, as with research on other new productsuccess antecedents (Hart, 1993), the definition andoperationalization of product success will influence thefindings for its relationship with development speed.
Managerially useful and/or empirically significantsuccess dimensions vary with respect to stakeholderinterests (Lipovetsky, Tishler, Dvir, and Shenhar, 1997),organizational goals (Venkatraman and Ramanujam,1986), managerial relevance, and even when after launchsuccess is measured (Hultink and Robben, 1995). Fur-thermore, despite the multidisciplinary character ofNPD projects, most empirical research on the speed–performance link uses only a single functional perspec-tive of success (i.e., marketing or finance), frequentlyproducing results incompatible with those from otherfunctional perspectives (Tatikonda and Montoya-Weiss,2001). Consequently, understanding the mechanismsunderlying the speed–success link warrants a finergrained meta-analytic approach that considers the perfor-mance implications of development speed in relationshipto the different dimensions of new product success.
The NPD literature is rife with categorizations ofproduct success dimensions (Cooper and Kleinschmidt,1987; Griffin and Page, 1993, 1996; Hart, 1993;Tatikonda and Montoya-Weiss, 2001). This researchuses the project-level classification by Tatikonda andMontoya-Weiss (2001) who distinguish between opera-tional and external outcome indicators (see Figure 1).Each of the other categorizations’ project-level dimen-sions fit into this typology.
Operational success measures assess product develop-ment from an internal perspective, including both productand process aspects. Product aspects include the achieve-ment of goals set for technical product performance andproduct competitive advantage. Process aspects measuredevelopment costs and proficiency in market entry timing.External measures reflect the commercial success of newproducts. Since new products are designed to addresscustomer needs, one way in which success is manifest is incustomer-based measures such as market share, salesvolume, revenue, customer satisfaction, and acceptance.
External outcomes:
• Customer-based outcomes
• Financial outcomes
Operational outcomes:
• Development costs
• Market entry timing
• Technical product quality
• Product competitive advantage
Process
Product
Figure 1. Consequences of New Product Development Speed
468 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
Financial measures, reflecting the extent to which eco-nomic goals are fulfilled, include profitability, margin,return on assets and investment, and break-even time(Venkatraman and Ramanujam, 1986). While both exter-nal dimensions capture commercial success outcomes,they nevertheless represent different aspects (Griffin andPage, 1993). Success in one dimension does not necessar-ily imply success in the other. For instance, a productcapturing a large market share may be unprofitable due toits cost structure (Mallick and Schroeder, 2005).
Implied in this multidimensional typology of NPDsuccess is a hierarchical structure: operational outcomescan be viewed as fundamental elements of success; finan-cial measures are the ultimate goal (Huang, Soutar, andBrown, 2004); and customer-based measures link theinternal, operational outcomes of the NPD process totheir financial implications. The typology also reflects atemporal measurement sequence (Hultink and Robben,1995). While operational outcomes are manifest in theshort term, external outcomes become visible only in thelonger term, with customer success manifesting itselffaster than financial success. The following sectionspresent the conceptual arguments and empirical evidencerelating each group of indicators to speed.
Operational Outcomes of Development Speed
Speed is both a competitive capability (Jayaram andNarasimhan, 2007) and an intermediate outcome of theNPD process (Ali et al., 1995). However, developmenttime is not the only intermediate NPD outcome. Devel-opment costs, market-introduction timing, productquality, and product superiority also have crucial impli-cations for achieving favorable customer-based andfinancial outcomes. Development cost and market-entrytiming proficiency, for instance, are associated with lowerand higher levels of profitability, respectively (Langerak,Hultink, and Griffin, 2008). Technical product qualityand product competitive advantage contribute signifi-cantly to both market and financial success (Li andCalantone, 1998; Zirger and Maidique, 1990).
Together, these five performance indicators (speed,cost, market entry, technical quality, and product advan-tage) comprise the NPD process’ operational outcomes.As these outcomes are interrelated (Meyer, 1993; Smithand Reinertsen, 1991), whether they can be simulta-neously achieved remains a heated debate. Some studiesportray a synergistic relationship and suggest that proce-dures and programs targeting faster development speedalso limit overhead costs, promote productivity (Davis,Dibrell, and Janz, 2002), and facilitate developing high-
quality, competitively superior products (Kessler andBierly, 2002). Adherents of the trade-off school, on theother hand, maintain that success of one occurs at theexpense of others (Jayaram and Narasimhan, 2007),arguing that time-based strategies direct attention awayfrom controlling costs and achieving quality (Crawford,1992; Eisenhardt and Tabrizi, 1995).
Development costs. Development speed affects prof-itability through its impact on development cost (Gupta,Brockhoff, and Weisenfeld, 1992), which includes allmonetary and human resources needed to develop a newproduct (Kessler and Bierly, 2002). Conflicting view-points exist regarding the precise link. On the one hand,faster development has been argued to promote efficientresource use by limiting the time allocated to peripheralactivities and providing a cap on man-hours (Eisenhardtand Tabrizi, 1995; Kessler and Chakrabarti, 1996;Rosenthal, 1992). Increased activity paralleling topromote speed also fosters higher team communication,cohesion, and learning (Brown and Eisenhardt, 1995).These, in turn, not only reduce errors and thus rework butalso help eliminate redundancies and overlaps whichelevate development costs with little return in productiv-ity (Clark and Fujimoto, 1991; Meyer, 1993). Shortercycle times also require lower levels of inventory andworking capital, reducing costs, and boosting productiv-ity (Davis et al., 2002). These arguments find empiricalsupport from Meyer and Utterback (1995).
In contrast, others maintain that faster product devel-opment is costly, with potentially serious consequences.Fast-paced development may require developing highlycomplex networks, which can be costly to coordinate andmanage (Calantone and Di Benedetto, 2000; Kessler andChakrabarti, 1999). With teams pushed to the limit oftheir capabilities, firms may incur expenses for addingpersonnel, materials, and equipment to meet aggressiveschedules (Murmann, 1994). Finally, rather than increas-ing productivity by reducing costly mistakes, accelera-tion efforts can in fact increase their occurrence througheliminating critical process steps (Murmann, 1994).
Recent empirical work suggests that there may be aU-shaped association between development time and cost(Langerak et al., 2008), consistent also with the results ofother experiments (Gupta et al., 1992) and models(Bayus, 1997). While these results somewhat reconcilethe two opposing camps, there still is little consensusregarding the true nature of the relationship. The issuebecomes even less clear when one considers studies thatfind no evidence of a significant link between NPD speedand cost, such as Kessler and Bierly (2002).
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4692013;30(3):465–486
Proficiency in market entry timing. Underlying manycompanies’ acceleration efforts are the first-mover advan-tages attributed to shorter NPD cycles. However, firmsentering a market prematurely or with an underdevelopedproduct also may not be able to exploit a product’s stra-tegic window (Lilien and Yoon, 1989). Proficient market-entry timing thus refers to the firm’s ability to get themarket-entry timing “right”: neither too early nor too late(Langerak and Hultink, 2006).
Development speed and market-entry timing areclosely related, yet distinct, intermediate NPD processoutcomes. Speed precedes market-entry timing becauseonly when the product is developed can a firm decidewhen to enter the market and—potentially—exploit astrategic product window (Ali, 2000). With increasedstrategic flexibility brought about by faster NPD(Eisenhardt and Tabrizi, 1995), firms have more freedomto determine the time of market entry and make betterentry decisions. Empirical studies that examine thespeed–market entry timing link are relatively rare. Anotable exception is Langerak et al. (2008), which foundthat longer development time was significantly negativelyassociated with market entry-timing proficiency.
Technical product quality. Technical quality is theproduct’s ability to perform its primary function (Mallickand Schroeder, 2005). This definition is consistent withconformance quality, a product’s conformance to design,and operating specifications (Jayaram and Narasimhan,2007). Also characterized by opposing perspectives is therelationship between speed and technical product quality.
Exclusively focusing on speed may compromiseadherence to product specifications (Eisenhardt andTabrizi, 1995). Further, substantially improving productperformance may run directly counter to accelerationgoals (Cohen et al., 1996). Emphasis on compressingtime-to-market prevents broadly considering alternativesand leaves little time to explore ways to improve productspecifications (Sethi, 2000). Prematurely freezing speci-fications to compress development time can prevent theincorporation of newer technologies that could increaseperformance (Rosenthal, 1992). Finally, eliminatingprocess steps in pursuit of speed also can increase productdefects and manufacturing problems (Crawford, 1992).These arguments are supported by one qualitative studyshowing that a majority of the case projects had to tradeoff speed to obtain the desired technical performance(Rosenthal and Tatikonda, 1993).
However, there is considerable empirical evidence thatfaster development does not necessarily lead to poor tech-nical product quality. For example, both Calantone and
Di Benedetto (2000) and Kessler and Bierly (2002) foundthat greater speed is positively correlated with productperformance. Similarly, Harter, Krishnan, and Slaughter(2000) show that shorter cycle time in software productsis associated with higher quality. On the other hand, Clarkand Fujimoto (1991) report no association between speedto market and product performance. In sum, the nature ofthe speed–technical quality link remains unresolved.
Product competitive advantage. Closely related toproduct quality is the extent to which a new product isperceived by consumers as being superior with respect tobenefit, innovativeness, or function (Montoya-Weiss andCalantone, 1994). The speed/competitive advantage rela-tionship is a topic of debate. Some scholars maintain thatproducts that reach the market more quickly are morelikely to be viewed by consumers as containing cutting-edge technologies (Atuahene-Gima, 2003; Carbonell andRodriguez-Escudero, 2006). Market and technology fore-casts are more accurate with faster development (Smithand Reinertsen, 1991), contributing further to perceivedcustomer fit (Kessler and Bierly, 2002; Tatikonda andMontoya-Weiss, 2001). However, fast development alsocan adversely affect competitive advantage. The absenceof adequate time to study customer needs and experimentwith alternative concepts can prevent developing prod-ucts offering superior solutions (Mallick and Schroeder,2005). Accelerated projects may thus fail to adequatelyalign with market demand, or be limited to incrementalinnovations (Crawford, 1992).
Empirical studies of the speed–competitive advantagerelationship also fail to provide unanimous support foreither viewpoint. Both Ali et al. (1995) and Carbonell andRodriguez-Escudero (2006) find that faster NPD is lin-early associated with increased competitive advantage.However, Lukas and Menon (2004) found evidence of aninverted U-shaped effect of NPD speed on advantage:faster product development results in more highly advan-taged products, up to some level of acceleration, withdetrimental effects after that.
External Outcomes of Development Speed
Customer-based outcomes. Favorable customer-based outcomes depend on the firm’s ability to transformcustomer needs into products that are perceived by cus-tomers as delivering value (Mallick and Schroeder,2005). The prevailing position points to a positive rela-tionship between speed and marketplace success. In thislogic, speed achieves market share objectives by allowingfirms to capture unchallenged markets and establish cus-
470 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
tomer loyalty early (Stalk and Hout, 1990). It alsoextends product life in the marketplace (Smith andReinertsen, 1991), increasing sales by reducing a firm’sreliance on obsolete products (Cordero, 1991). Empiricalstudies support this logic (Calantone, Vickery, and Dröge,1995; Garcia, Sanzo, and Trespalacios, 2008). Jayaramand Narasimhan (2007) even find that time-to-marketsurpasses both cost and product quality in their positiveassociation with market share.
However, speed may not consistently contribute tomarketplace success, nor does speed necessarily impactall indices of customer-based success in the same way.For example, Ali (2000) empirically finds that shorterdevelopment times positively influence achieving initialrevenue goals for minor innovations, but not for majorones. Tatikonda and Montoya-Weiss (2001) report astrong association between shorter development timesand increased customer satisfaction, but none for achiev-ing sales objectives. Langerak, Rijsdijk, and Dittrich(2009) find that development time is not a significantpredictor of overall sales volume.
Financial outcomes. Profitability and return on assetsor investment reflect the bottom-line implications ofNPD, as does break-even time, the time it takes for a firmto start making a profit from a new product (Jayaram andNarasimhan, 2007). The prevailing view on the speed–financial success link is a synergistic relationship: fasterspeed increases financial success. Because fast develop-ment increases the likelihood that a firm enters a marketwith fewer competitors, they can follow a skimmingpricing strategy, which positively impacts profits(Vandenbosch and Clift, 2002). Larger market shares andloyal customer bases achieved through early market pen-etration enhances profitability (Dröge et al., 2000). Thesearguments find empirical support from Jayaram andNarasimhan (2007), with respect to profitability,Calantone et al. (1995) and Langerak and Hultink (2005)for return on investment (ROI) and return on assets(ROA), and Ali et al. (1995) for shorter break-even times.
The above research notwithstanding, an increasingnumber of studies reveal that this relationship also is farfrom straightforward. At least one study reports no linkbetween project length and ROI (Swink and Song, 2007).Ali (2000) finds that the speed–profitability relationshipis moderated by product innovativeness. Langerak andHultink (2006) document an inverted U-shaped relation-ship, suggesting there is an “optimal” developmentspeed with the greatest financial returns. Finally,Adams-Bigelow and Griffin (2005), report a negativerelationship between speed and profitability. These con-
flicting results illustrate the potential hidden costs ofaccelerating development (Crawford, 1992) and highlightthe fact that cycle time reduction may not necessarilyhave favorable financial outcomes.
Speed–success relationship summary. In sum, neitherconceptual arguments nor empirical research on howdevelopment speed relates to operational and externalmeasures of NPD performance has converged to a con-sensus. This study addresses these inconsistencies andadvances scholarly knowledge of the speed–success rela-tionship using meta-analytic techniques. This allows us tointegrate systematically empirical findings from existingresearch with the methodological and contextual charac-teristics of the primary studies responsible for the diver-gent findings.
Methodology
Database Development
The search procedures employed in this investigation tocreate a comprehensive list of relevant studies were con-sistent with those utilized by previous meta-analyses inmarketing and management (Brown and Peterson, 1993;Joshi and Roh, 2009; Kirca, Jayachandran, and Bearden,2005). It involved first conducting computerized databasesearches (ABI/INFORM Global, EBSCO, EconLit,JSTOR, ScienceDirect, and Web of Science) using key-words and phrases such as “product development,” “cycletime,” “innovation speed,” “time-to-market,” and “leadtime.” The Social Science Citation Index for studies ref-erencing the most-cited articles in NPD speed researchwere also consulted (Griffin, 1997; Kessler andChakrabarti, 1996). These steps were supplementedwith a manual search of the following innovation, mar-keting, and management journals: Academy of Manage-ment Journal, IEEE Transactions on EngineeringManagement, Journal of the Academy of MarketingScience, Journal of Engineering-Technology Manage-ment, Journal of Marketing, Journal of MarketingResearch, Journal of Operations Management, Journal ofProduct Innovation Management, Management Science,Marketing Science, R&D Management, and StrategicManagement Journal. References from previous meta-analyses (Chen et al., 2010; Gerwin and Barrowman,2002; Henard and Szymanski, 2001; Montoya-Weiss andCalantone, 1994; Pattikawa et al., 2006) also werereviewed. Finally, to access new or unpublished work,forthcoming articles were requested from scholars knownto be engaged in cycle time research.
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4712013;30(3):465–486
Two criteria determined whether to include a study inthe meta-analytic database. First, it had to be published (orforthcoming) in a peer-reviewed, Institute for ScientificInformation (ISI)-rated journal. This criterion carries therisk of publication bias, which arises from the tendency ofpublished work to report greater effect sizes than unpub-lished work, thus producing inflated meta-analytic esti-mates. However, the extent to which the meta-analyticestimates suffered from this bias was assessed using filedrawer analysis and the findings reported in the resultssection. Second, it had to be an original quantitativeempirical study of NPD speed. Articles presentingtheoretical models (e.g., Emmanuelides, 1993; Hitt,Hosskison, and Nixon, 1993; Zirger and Hartley, 1994),methodology development (e.g., Boer and Logendran,1999), analytical models lacking empirical testing (e.g.,Bayus, Jain, and Rao, 1997; Cohen et al., 1996; Loch andTerwiesch, 1998; Morgan, Morgan, and Moore, 2001),and purely qualitative studies (e.g., Flint, 2002; Millson,Raj, and Wilemon, 1992; Rabino and Wright, 1993;Vesey, 1992; Von Braun, 1990) were discarded. Articlesthat operationalized development speed as predictions(e.g., Moorman and Miner, 1998; Rindfleisch andMoorman, 2001) also were excluded from the database.
Each study also had to provide the following pieces ofinformation: (1) a correlation-based effect size statisticfor the relationship between development speed and atleast one NPD performance dimension; (2) the number ofobservations (sample size) on which the correlation wasbased; and (3) reliabilities for both constructs (whereapplicable). Articles in which development speed wastreated as part of a composite NPD performance measurewere discarded (e.g., Ancona and Caldwell, 1992;Kusunoki, Nonaka, and Nagata, 1998; Rochford andRudelius, 1992; Salomo, Weise, and Gemünden, 2007;Schulze and Hoegl, 2006; Song, Montoya-Weiss, andSchmidt, 1997; Tatikonda and Rosenthal, 2000). If anotherwise eligible study did not report one or more ofthese pieces of information, authors were requested toprovide them. Of the 17 authors contacted, 9 suppliedmeta-analyzable data.
Finally, the following rules were employed to ensurean acceptable level of independence for the correla-tions in the database: (1) If a publication reportedresults from multiple independent samples separately,their results were included as independent samples(Geyskens, Steenkamp, and Kumar, 2006); (2) If mul-tiple publications were based on the same or on par-tially overlapping data sets, they were treated as a singlestudy, and correlations entered between two identicalvariables only once (Franke and Park, 2006; Geyskens
et al., 2006).1 These efforts yielded a database of 52independent samples obtained from 56 articles pub-lished between 1989 and 2009, three of which wereforthcoming when data collection was finalized in April2009 (See Appendix S1).
Variable Classification and Coding
A preliminary coding protocol that specified the informa-tion to be extracted from each study was prepared toreduce errors and ensure coding consistency (Lipsey andWilson, 2001; Stock, 1994). One third of the studies werefirst coded independently by two of the authors based onthis coding protocol, yielding an overall coding consis-tency of 91%.2 Discussion resolved disagreements andambiguities in the coding scheme. The remaining studieswere coded by the first author on the basis of the revisedscheme. The final version of the database was verified bythe second author.
All of the success indicators were classified into eitheroperational or external NPD outcomes based onTatikonda and Montoya-Weiss (2001), and then furtherclassified into the six subcategories of Figure 1. Follow-ing Damanpour (1991) and Geyskens, Steenkamp, andKumar (1998), effect sizes were assigned to performancecategories based on how constructs were operationalized.In the majority of instances, there was no discrepancybetween construct name and specific variable operation-alization. However, due to the limited number of studiesinvestigating certain constructs (e.g., customer satisfac-tion, ROI/ROA, break-even time), all specific variableoperationalizations were collapsed under the umbrellaconstruct for that category (e.g., ROI/ROA and break-even time both became externally determined financialoutcomes).
Because a number of studies either used a single itemmeasure without explicitly specifying the NPD perfor-mance dimension examined (Larson and Gobeli, 1989),or collapsed multiple different items into a single generalmeasure (e.g., Aronson, Reilly, and Lynn [2006] “NPDproject performance”; Rusinko’s [1999] “NPD effective-ness”), an “overall NPD performance” construct was
1 As also noted by an anonymous reviewer, Gleser and Olkin’s (1994)generalized least squares approach offers a theoretically and statisticallyadvanced way of addressing dependent effect sizes. However, this approachrequired knowing the covariance or correlation between dependent effectsizes, which, in a large number of instances were not available. Therefore,in the interest of consistency, the more conservative approach using depen-dent effect sizes was performed.
2 The inter-coder agreement per meta-factor was as follows: New productsuccess 91%, development costs 90%, market-entry timing 100%, technicalproduct quality 95%, product competitive advantage 86%, customer-basedsuccess measures 93%, and financial success measures 86%.
472 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
created to accommodate for these studies. Effect sizeswere reverse coded for studies that operationalized speedas development time. Finally, information on a number ofpotential contextual and methodological study character-istics was coded for moderator analysis.
The following actions were taken to minimize bias thatmay occur due to multiple counts of dependent effect sizeestimates. First, when an article reported correlations forsubsamples but not for the total sample, each pair ofcorrelations involving the same variables was averaged(Gerwin and Barrowman, 2002). Second, for articlesreporting multiple variables for the same relationship ormeasuring the same variable at different times, theaverage correlation was entered in the meta-analysis(Brown and Peterson, 1993; Crosno and Dahlstrom,2008). Coding yielded 84 harvested effect sizes. Samplesizes ranged between 29 and 692. The average and totalsample sizes were 163 and 13,163, respectively.
Meta-Analytic Calculations
As with other recently published meta-analyses (Chenet al., 2010; Gerwin and Barrowman, 2002), this studyadhered to Hunter and Schmidt’s (1990) analyticapproach. This approach provides clear guidelines tocorrect for statistical artifacts that cause effect sizes todeviate from the true population effect size and avoids thestatistical power difficulties associated with testing whenthe number of studies is small (Gerwin and Barrowman,2002). Finally, it allows investigation of the moderatingrole of study characteristics and the determination ofconditions under which certain effects are manifest (Chenet al., 2010).
The biasing effect of both sampling and measurementerror were corrected for prior to analysis. To correct forsampling error, the weighted-mean correlation ( r ) wascalculated by weighing each correlation with its corre-sponding sample size and its standard deviation. Tocorrect for measurement error, each reported correlationwas divided by the square root of the reliabilities of thetwo constructs. For averaged correlations, an averagereliability coefficient also was computed (Gerwin andBarrowman, 2002). When reliability information was notprovided, the mean reliability from the other articlesinvestigating the same relationship was used as the bestestimate of the missing reliability coefficient (Balkundiand Harrison, 2006). The true score correlation (r) wasreached by computing the weighted average of thereliability-corrected correlations.
In calculating the variability in effect size estimatesonce they have been averaged across studies, the fixed-
effects (FE) model attributes variability in findings tosampling error variance only, implicitly assuming thatstudies are otherwise homogeneous. The random effects(RE) model (Franke and Park, 2006; Rodriguez-Cano,Carrillat, and Jaramillo, 2004), on the other hand, treatsthe variability as arising from two sources: (1) studysampling (between-studies variance), and (2) the sam-pling of individuals within studies (sampling error vari-ance). This article thus adopted the RE approach, since itis more conservative and is less susceptible to Type Ierrors (Hunter and Schmidt, 2000).
Results
Main Effects
Table 2 reports the main effects results, giving thenumber of correlations (k), combined sample size (N),sample-weighted uncorrected correlation ( r ), estimatedtrue correlation corrected for sampling error and unreli-ability (r) for each effect size. Ninety-five percent con-fidence intervals were constructed around r for eachrelationship using the random effects standard errorformula (Hunter and Schmidt, 1990). If this confidenceinterval does not include zero, the r estimates are notsignificantly different from zero. As some of these con-fidence intervals are fairly large due to the use of therandom effects model, the resulting conclusions are fairlyconservative.
Also, since published studies tend to report greatereffect sizes than unpublished ones (the “file drawerproblem” or “availability bias”), the exclusion of unpub-lished work may lead to inflated meta-analytic estimatesand erroneous conclusions (Lipsey and Wilson, 2001).The last column thus reports the failsafe k, or number ofadditional unpublished studies that would be needed toreduce the effect size to below statistical significance(p < .05), to address potential concerns of this bias(Orlitzky, Schmidt, and Rynes, 2003). Larger failsafe k’s,such as these, indicate the absence of a serious threat tothe validity and reliability of meta-analytic findings.
The results indicate a significant positive correlationbetween the development speed–overall new productsuccess constructs (r = .309, p < .05), with a failsafek sufficiently large to alleviate potential file drawerconcerns.
Of the relationships between development speed andthe four operational NPD outcomes, all but one is statis-tically significant. First, shorter development times aresignificantly associated with lower development costsand increased proficiency in market entry timing, the two
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4732013;30(3):465–486
process-related success constructs. With effect estimatesof r = -.468 and r = .646, respectively (both p < .05), theassociations can comfortably be classified as strong(Cohen, 1992). The speed and technical product qualityrelationship fails to reach significance. Finally, develop-ment speed is positively and significantly associatedwith increased product competitive advantage (r = .298,p < .05). Failsafe k-values (k = 36–159, 6–12 times thenumber of correlations included in the meta-analyses)for the three significant relationships suggest that, ifincluded, new or unpublished studies are unlikely toinfluence the results.
For the two external outcomes, financial and customer-based measures, the estimated true correlation coefficientfor both constructs in relation to speed is positive, sig-nificant, and moderate in magnitude (r = .293 andr = .282, p < .05, respectively). The failsafe k-values(4–5 times the number of correlations included in themeta-analysis) also suggest that these relationshipsappear resistant to unpublished null effects.
Together, these results generally support the conten-tion that NPD speed promotes success. The findings fordevelopment cost and proficiency in market entry timing,the two process-related aspects of operational success,are in line with the synergistic accounts (faster is better)of their association with speed as is the relationshipbetween speed and the product-related operational con-struct of product competitive advantage—faster develop-ment is associated with more competitively advantagedproducts. However, the results suggest no significantassociation between speed and technical product quality,and thus are unable to establish the superiority of either asynergistic (faster is better) or trade-off perspective(faster is worse) for that relationship.
Development speed also is significantly and moder-ately strongly linked both to increased customer-basedand financial new product outcomes, the external mea-sures of new product success. If success measures arehierarchical in nature, with customer-based successacting as one possible link between operational processand product measures of success and the ultimate goal ofcreating financially successful products, as suggested byLangerak et al. (2008), then these findings suggest thatNPD speed must be considered as at least one of thefactors that managers will want to manage to achievethese external success measures.
Heterogeneity Analysis
Although the main effect results indicate a generally syn-ergistic relationship between development speed andTa
ble
2.M
ain
Eff
ect
Res
ults
kN
Nav
r min
r max
r obs
rr
SDE
rL
BU
BVa
r obs
Var s
amp
Var r
elVa
r exp
lQ
-sta
tdf
Fai
lsaf
ek
New
prod
uct
succ
ess:
Ove
rall/
com
posi
te22
3907
178
-.56
0.7
90.3
36.2
51.3
09*
.076
.102
.401
.130
.005
.002
5.27
5%57
9.32
5*21
114
Ope
ratio
nal
outc
omes
:Pr
oces
sm
easu
res
Dev
elop
men
tco
sts
1930
3216
0-.
630
.550
-.34
0-.
423
-.46
8*.0
55-.
530
-.31
6.0
63.0
04.0
0616
.522
%28
3.52
7*18
159
Mar
ket
entr
ytim
ing
321
772
.410
.450
.427
.421
.646
*.0
56.3
11.5
31.0
00.0
09.0
0910
0.00
%.0
502
36Pr
oduc
tm
easu
res
Tech
nica
lpr
oduc
tqu
ality
1721
1412
4-.
342
.650
.220
.131
-.15
1.0
75-.
017
.278
.097
.008
.000
8.50
1%21
2.63
1*16
NA
Prod
uct
com
petit
ive
adva
ntag
e15
2157
144
-.65
9.4
86.2
11.2
47.2
98*
.048
.153
.341
.036
.006
.002
21.6
30%
88.0
29*
1475
Ext
erna
lou
tcom
es:
Cus
tom
erba
sed
1317
6813
6-.
120
.600
.207
.242
.282
*.0
52.1
40.3
45.0
38.0
07.0
0222
.254
%74
.937
*12
60Fi
nanc
ial
817
8922
4.0
15.3
77.2
29.2
20.2
93*
.042
.138
.302
.016
.004
.002
37.6
10%
31.5
47*
739
*p
<.0
5.k,
num
ber
ofco
rrel
atio
nsan
alyz
ed;
N,c
ombi
ned
sam
ple
size
;N
av,a
vera
gesa
mpl
esi
ze(N
/k);
r min,m
inim
umre
port
edco
rrel
atio
n;r m
ax,m
axim
umre
port
edco
rrel
atio
n;r o
bs,s
impl
eav
erag
eof
repo
rted
(unc
orre
cted
)co
rrel
atio
ns;
r ,sa
mpl
e-w
eigh
ted
aver
age
ofre
port
ed(u
ncor
rect
ed)
corr
elat
ions
;r,e
stim
ated
true
corr
elat
ion
corr
ecte
dfo
rsa
mpl
ing
erro
ran
dun
relia
bilit
y;SD
Er ,
sam
plin
ger
ror
vari
ance
ofth
em
ean
obse
rved
corr
elat
ion
(ran
dom
effe
cts
form
ula)
;LB
,low
erbo
und
of95
%co
nfide
nce
inte
rval
arou
ndth
er ;
UB
,upp
erbo
und
of95
%co
nfide
nce
inte
rval
arou
ndth
er ;
var o
bs,o
bser
ved
vari
ance
inef
fect
size
s;va
r sam
p,va
rian
cedu
eto
sam
plin
ger
ror;
var r
el,
vari
ance
due
todi
ffer
ence
sin
relia
bilit
yof
cons
truc
tm
easu
rem
ent;
var e
xpl,
perc
enta
geof
obse
rved
vari
ance
expl
aine
dby
sam
plin
gan
dm
easu
rem
ente
rror
s(w
hen
<75%
,sea
rch
for
mod
erat
ors)
;Q-s
tat,
Qst
atis
ticfo
rhe
tero
gene
ity(w
hen
sign
ifica
nt,s
earc
hfo
rm
oder
ator
s);d
f,de
gree
sof
free
dom
for
Q-s
tatis
tic(k
-1);
Fai
lsaf
ek,
num
ber
ofad
ditio
nal
unpu
blis
hed
orov
erlo
oked
stud
ies
that
wou
ldne
edto
bein
clud
edin
the
data
base
tore
duce
the
effe
ctsi
zeto
r=
.05.
474 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
NPD outcomes, these relationships may not necessarilybe manifest similarly across all 52 independent samples.That is, substantial variance across independent samplesmay have remained even after correcting for artifacts(i.e., sample size and reliability).
The generalizability of the main effects findings wasassessed in two ways. The Q statistics (Lipsey andWilson, 2001) assesses the homogeneity of the effect sizedistribution (Hedges and Olkin, 1985). A significant Qsuggests a likely presence of effect size heterogeneity,warranting a search for moderators to explain it. Thesecond method used was the 75% rule (Hunter andSchmidt, 1990). According to this rule, when statisticalartifacts (sampling and measurement errors) explain 75%or more of the observed variance, moderators are unlikelyto have caused variation, and the effect size can be con-sidered homogeneous across studies. The methodsyielded consistent results for all seven relationships.
Only the speed-market entry timing proficiency rela-tionship proves to be homogeneous: its Q statistic is notstatistically significant, and virtually 100% of observedvariance is accounted for by sampling and measurementerrors. The other six speed/success relationships displayheterogeneity. It was thus necessary to search for moder-ating variables for these success dimensions.
Moderator Effects
Potential moderators were chosen based on previousNPD research (Griffin, 2002; Kessler and Chakrabarti,1996; Mallick and Schroeder, 2005) and meta-analyses(Chen et al., 2010; Henard and Szymanski, 2001). Mod-erators were coded only when information was unam-biguous in the text (Damanpour, 1991). Moderation wastested only when at least three observations were present
for each category to ensure estimate stability (Geyskenset al., 2006).3 Table 3 lists the characteristics consideredin the moderator analysis.
For moderators coded as categorical variables, theanalysis of variance (ANOVA)-analogue test was used toassess their influence on effect size (Lipsey and Wilson,2001). This method splits the effect sizes into the mod-erator categories and tests for between- and within-groupcategory differences, yielding two test statistics: (1) thebetween-group goodness-of-fit statistic Qb, and (2) thewithin-group goodness-of-fit statistic Qw (Aguinis andPierce, 1998). When Qb is statistically significant, themean effect size differs for the two groups, and the mod-erator is significant (Joshi and Roh, 2009). To compareestimated true population correlations of moderator cat-egories, separate meta-analyses were conducted for eachcategory. Ninety-five percent confidence intervals wereconstructed around the uncorrected weighted mean cor-relation for each category to see whether it includes zero.Since year of data collection was coded as a continuousvariable, its moderating role was assessed by treating it asthe independent variable in a simple regression for pre-dicting effect sizes for each meta-factor. Since they aremore accurate representations of true effect size thanreported correlations (Hunter and Schmidt, 1990), thisstudy adhered to the approach used by Henard andSzymanski (2001) and Troy, Hirunyawipada, and Paswan(2008) and conducted the moderator analysis onreliability-corrected correlations. Results from these pro-
3 Consequently, investigating the role of several of the initially codedmoderating variables was not possible, including: (1) key informant versussenior management response data; (2) primary versus secondary data;(3) survey versus other forms of data collection; (4) cross-sectional versuslongitudinal research design; (5) industrial versus consumer markets;(6) single versus multiple organization response data; and (7) U.S. versusnon-U.S. sample.
Table 3. Potential Moderators of the Speed–Performance Relationship
Study Design Characteristics1. Organizational sampling procedure Random vs. nonrandom sampling2. Scope of measurement Project-level vs. program-level scope3. Number of informants Single vs. multiple informants4. Year of data collection The year in which the data used in the study was collected (“publication year-3” if not
specified)Speed Measurement Characteristics
5. Number of items Single-item vs. multi-item scales6. Absolute vs. relative measures Actual development speed vs. speed compared to initial plans or similar past projects7. Objective vs. subjective measures Development speed obtained from company database vs. respondents’ perceptual
assessment of development speedContextual Characteristics
8. Type of innovation Products only vs. products and services9. Number of industries represented in sample Single industry vs. multiple industries
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4752013;30(3):465–486
cedures are presented in Table 4 (research design mod-erators), 5 (speed measurement moderators), and 6(contextual moderators).
As shown in Table 4 by the significant Qb statistics,organizational sampling criteria significantly moderatedall relationships, except for the speed–financial out-comes link. Randomly selected samples had larger effectsizes for customer-based and competitive advantage out-comes than did those not reporting such a selection pro-cedure. Nonrandom samples had higher effect sizes forcomposite success, development costs, and technicalproduct quality outcomes. Sampling procedure wasparticularly influential on the development speed–composite success link, with a very high Qb (319.095)and markedly different mean effect sizes: rrandom = -.099(n.s.) versus rnonrandom = .510 (p < .05).
Since the overwhelming majority of studies engagedin project level as opposed to program-level data collec-tion (46 vs. 6), the only relationship for which a sufficientnumber of observations per moderator category wasavailable was speed–composite success. However, a non-significant Qb statistic indicated no between-category dif-ference in mean effect size.
The moderating role of the number of informants wasassessed for three relationships. Qb was significant foronly the speed–cost link (Qb = 23.607, p < .01), wherethe single-respondent mean effect size was markedlylarger than its multiple-respondent category counterpart,and both effects were significant (rsingle = -.500,rmultiple = -.220, p < .05).
Finally, as indicated by their small but statisticallysignificant regression coefficients (.027 and .026, p < .05,respectively), more recent studies report greater effectsizes regarding the speed–success and speed–technicalperformance links.
The role of speed measurement characteristics on themeta-analytic estimates was examined next (Table 5). Itwas possible to test all six success dimensions for themoderating effect of the number of items in the speedmeasure. Only customer-based outcomes showed nomoderation (Qb = 2.400, p > .10). For the remaining fiverelationships, studies using multi-item scales report cor-relations noticeably greater in magnitude than those usingsingle-item scales.
Whether speed was operationalized in absolute or rela-tive terms was a significantly moderating influence forboth external success outcomes. When measured in rela-tive terms, there was a significant relationship betweenspeed and customer and financial-based success.However, these relationships reduced to insignificancewhen studies used absolute speed measures. The same
result occurred for the speed–product competitive advan-tage relationship. The speed-technical product qualitylink showed no moderation by this variable, while theother success dimensions had insufficient numbers ofcases for testing.
A thorough investigation of the objective versus sub-jective development speed operationalization was infea-sible due to an insufficient number of studies usingobjective measures for all relationships except for thetechnical product quality success dimension. Even then,however, the results did not suggest any moderating roleof this variable.
Finally, the roles of two contextual study characteris-tics on the meta-analytic estimates were assessed. Table 6shows that whether or not studies focused on products(physical goods) only or a combination of products andservices significantly moderated only the effect of devel-opment speed on composite success, (Qb = 65.480,p < .01), with the mixed category reporting greater cor-relations (rproducts = .215 and rmixed = .480).
Three development speed relationships assessedwhether studies using single-industry data differed fromthose using data from multiple industries: customer-based outcomes, development costs, and technicalproduct quality. However, number of industries was not asignificant moderator for any of the studied effects(Qb = .185, .153, and 1.525, respectively).
In sum, although a comprehensive moderator analysisof several potential moderators was not feasible due todata limitations, a number of factors were identified thatdo contribute to a lack of cross-situational consistency inspeed/success relationships.
Antecedents of New ProductDevelopment Speed
To provide a holistic view of NPD speed and validate theresults reported by Chen et al. (2010), a meta-analysis ofNPD speed antecedents was also performed following theanalytic procedures described above. The database,which consisted of 75 independent samples from 86 pub-lications yielded 354 effect sizes spread across 42 meta-factors (see Appendix S2).4 Sample sizes ranged from 11to 692. The average and total sample sizes were 136 and46,354, respectively.
Although this study adopted a more fine-grainedapproach to variable classification (for instance, distin-
4 Chen et al. (2010) used a smaller database of 217 effect sizes spreadacross 17 meta-factors, harvested from 74 independent samples whichappeared in 70 publications.
476 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
guishing between firm and market perspectives of inno-vativeness) and considered additional influences (e.g.,NPD competencies, firm characteristics, and environ-mental factors), its meta-factor groupings correspondedbroadly with the Chen et al. (2010) study. The anteced-ents analyzed are as follows. The number of independentsamples, which reported correlations with developmentspeed (k), are indicated in parentheses. Project charac-teristics: complexity (14), firm perspective of innovative-ness (19), market perspective of innovativeness (9),composite/not specified perspective of innovativeness(5), project newness (5), project size (3); Process char-acteristics: standardization (10), supplier involvement(12), customer involvement (10), use of other outsideassistance (11), goal effectiveness (15), formal processuse (9), process concurrency (13), iteration frequency(5), testing (3); NPD team characteristics: cross-functional team use (9), extent of functional diversity(7), organizational integration (12), teamwork quality(13), team size (14), team stability (4), team commit-ment (4), management style (11), team leader strength(10), team colocation (7); NPD competencies: planningproficiency (4), marketing proficiency (11), technicalproficiency (9), problem-solving proficiency (4), teamlearning (6); Firm characteristics: alignment with corecompetencies (7), resource availability (4), organiza-tional support (8), speed emphasis (6), time-based incen-tives (4), project priority (6), company size (13),innovative organizational climate (5); Environmentalcharacteristics: technology turbulence (12), market tur-bulence (12), competitive intensity (4), market attrac-tiveness and ease of entry (5).
The results, particularly those pertaining to the meta-factors with population correlation coefficients greaterthan .30 (“salient” antecedents of development speed)were consistent with Chen et al. (2010).
No project characteristic emerged as a significantantecedent to development speed.
Goal effectiveness was the only salient (r > .3)process characteristic, with a corrected population corre-lation estimate of .377. Also contributing to faster productdevelopment were process concurrency (r = .278), cus-tomer involvement (r = .277), formal process use(r = .246), testing (r = .240), supplier involvement(r = .152), and standardization (r = .135). The linkbetween these seven process-related antecedents anddevelopment speed was cross-situationally consistent.
Of the 10 NPD team characteristics analyzed, teamstability (r = .385) and team leader strength (r = .334)are salient contributors to development speed. Neither thechi-square test nor the 75% rule suggested heterogeneityTa
ble
4.M
oder
ator
Ana
lysi
sR
esul
tsfo
rR
esea
rch
Des
ign
Cha
ract
eris
tics
1.Se
lect
ion
ofO
rgan
izat
ions
2.L
evel
ofA
naly
sis
3.N
umbe
rof
Info
rman
ts4.
Yea
rof
Dat
aC
olle
ctio
n
Ran
dr
(k)
nRan
dr
(k)
Qb
Pro
jr
(k)
Pro
grr
(k)
Qb
Sing
ler
(k)
Mul
tir
(k)
Qb
b(S
E)
New
prod
uct
succ
ess:
Ove
rall/
com
posi
te-.
099
(5)
.510
*(1
0)31
9.09
5*.2
97*
(18)
.367
*(4
)3.
026
.309
*(1
9).3
10*
(3)
.956
.027
*(.
011)
Ope
ratio
nal
outc
omes
:Pr
oces
sm
easu
res
Dev
elop
men
tco
sts
-.39
3*(3
)-.
524*
(14)
12.5
78*
–(1
9)–
(1)
—-.
500*
(13)
-.22
0*(6
)23
.607
*.0
28(.
036)
Prod
uct
mea
sure
sTe
chni
cal
prod
uct
qual
ity.0
88(5
).2
51*
(12)
13.2
05*
–(1
6)–
(1)
—.1
32(1
7).2
18*
(5)
2.75
5.0
26*
(.01
2)Pr
oduc
tco
mpe
titiv
ead
vant
age
.475
*(5
).2
20*
(10)
28.9
53*
–(1
5)–
(2)
—–
(14)
–(1
)—
.032
(.02
3)E
xter
nal
outc
omes
:C
usto
mer
-bas
ed.3
99*
(5)
.151
(8)
26.8
44*
–(1
1)–
(2)
—–
(11)
–(2
)—
.009
(.01
4)Fi
nanc
ial
.251
*(3
).3
18*
(5)
1.78
0–
(6)
–(2
)—
–(7
)–
(1)
—.0
21(.
018)
*p
<.0
5.r,
estim
ated
true
corr
elat
ion
corr
ecte
dfo
rsa
mpl
ing
erro
ran
dun
relia
bilit
yfo
rm
oder
ator
cate
gory
subs
et;
k,nu
mbe
rof
corr
elat
ions
anal
yzed
inm
oder
ator
cate
gory
subs
et;
Ran
d,ra
ndom
sele
ctio
nof
orga
niza
tions
;nR
and,
nonr
ando
mse
lect
ion
ofor
gani
zatio
ns;
Pro
j,pr
ojec
t-le
vel
succ
ess;
Pro
g,pr
ogra
m-l
evel
succ
ess;
Sing
le,
one
resp
onde
nt;
Mul
ti,
mul
tiple
resp
onde
nts;
b,m
ultip
lere
gres
sion
coef
ficie
ntes
timat
efo
rye
arof
publ
icat
ion;
SE,s
tand
ard
erro
rof
coef
ficie
ntes
timat
efo
rye
arof
publ
icat
ion.
Num
bers
inbo
ldin
dica
tew
hich
ofth
etw
osu
bgro
ups
had
asi
gnifi
cant
lyla
rger
effe
ctsi
zew
hen
Qb
issi
gnifi
cant
.
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4772013;30(3):465–486
in their relationship with development speed. Team com-mitment and management style were also significant andcross-situationally consistent antecedents pertaining tothe NPD team. Organizational integration (r = .260),teamwork quality (r = .229), team commitment (r =.210), management style (r = .190), and cross-functionalteam use (r = .180) were also significant, albeit notsalient, features of the NPD team associated with fasterdevelopment.
With regard to NPD competencies, marketingproficiency (r = .374), problem-solving proficiency(r = .326), and team learning (r = .318) were salientdeterminants of development speed. Technical profi-
ciency was also a significant, but not quite salient, con-tributor to speed (r = .293). However, all of thesecompetencies displayed heterogeneity in their relation-ship with development speed.
Innovative firm climate (r = .377), organizationalsupport (r = .341), and availability of resources(r = .314) were the firm characteristics displaying thestrongest association with faster NPD. Emphasis onspeed (r = .254), project priority (r = .142), andcompany size (r = .082) were also significant meta-factors. None of these meta-factors, except for availabil-ity of resources and facilities, was cross-situationallyconsistent.
Table 5. Moderator Analysis Results for Speed Measurement Characteristics
5. Number of Items 6. Absolute vs. Relative 7. Objective vs. Subjective
Single r (k) Multi r (k) Qb Abs r (k) Rel r (k) Qb Obj r (k) Subj r (k) Qb
New product success:Overall/composite -.114 (8) .500 (14) 318.346* – (2) – (20) — – (1) – (14) —
Operational outcomes:Process measures
Development costs -.340* (8) -.611* (7) 54.111* – (2) – (17) — – (2) – (17) —Product measures
Technical product quality .120 (14) .408 (3) 16.656* -.124* (5) .158 (23) .289 .151 (4) .158 (13) .396Product competitive advantage .134 (6) .388* (9) 30.816* -.027 (3) .410 (12) 76.197* – (1) – (14) —
External outcomes:Customer based .260* (7) .322* (6) 2.400* .067 (5) .368* (8) 32.146* – (2) – (11) —Financial .163* (8) .532* (3) 55.158* .071 (4) .464* (4) 67.217* – (1) – (7) —
* p < .05.r, estimated true correlation corrected for sampling error and unreliability for moderator category subset; k, number of correlations analyzed in moderatorcategory subset; Single, single-item scale; Multi, multiple-item scale; Abs, absolute measure of speed (months); Rel, relative measure of speed; Obj,objective measure of speed; Subj, subjective measure of speed. Numbers in bold indicate which of the two subgroups had a significantly larger effect sizewhen Qb is significant.
Table 6. Moderator Analysis Results for Contextual Characteristics
8. Innovation Type 9. Number of Industries
Prod r (k) Serv r (k) Qb Single r (k) Multi r (k) Qb
New product success:Overall/composite .215 (18) .480* (4) 65.480* — (1) — (21) —
Operational outcomes:Process measures
Development costs — (17) — (2) — –.446* (6) –.471* (13) .185Product measures
Technical product quality — (17) — (1) — .125 (5) .155 (12) .153Product competitive advantage — (15) — (2) — — (2) — (13) —
External outcomes:Customer based .271* (10) .314* (3) .561 .335 (4) .267* (9) 1.525Financial — (8) — (0) — — (1) — (7) —
* p < .05.r, estimated true correlation corrected for sampling error and unreliability for moderator category subset; k, number of correlations analyzed in moderatorcategory subset; Prod, physical good or product; Serv, service; Single, one industry context; Multi, multiple industries in survey. Numbers in bold indicatewhich of the two subgroups had a significantly larger effect size when Qb is significant.
478 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
None of the four environmental characteristics ana-lyzed proved significant.
Table 7 summarizes the main effects and heterogene-ity analysis results and compares it to those reported byChen et al. (2010). The results generally validate the pre-vious findings, discrepancies attributable to differences invariable classification and database size.5
Discussion
The main finding of this meta-analysis is that, at firstglance, development speed is associated with increasednew product success. These findings are independent ofwhether success is measured overall, as an operationaloutcome, or as an external outcome.
Table 8 shows how the results of this study compare toprevious meta-analytic studies. As none of the previousmeta-analyses investigated new product success as a mul-tidimensional construct, their main effect findings arecomparable only to those produced by the studies in ourmeta-analytic database that used just an “overall” or“composite” measure of success (Table 2, row 1). Themain effect size estimate of the present research is justlarger than the average of the effect sizes found previ-ously, and thus corroborates previous findings. Interest-ingly, as these data also show, both Henard andSzymanski (2001) and Pattikawa et al. (2006) found thatthe observed variance explained by sampling and mea-surement errors was less than 75%, suggesting that non-random cross-situational heterogeneity exists in the dataand that the corrected population effect size yielded bythe main effects analysis may not be generalizable ontocertain groups of studies.6 Investigating the moderatinginfluences on effect size and the speed relationships forthe different dimensions of success is the unique contri-bution of this research to extant knowledge in the field.
Closer inspection of the meta-analytic estimates forthe speed–performance associations for the differentsuccess dimensions studied reveals that faster productdevelopment is linked to favorable outcomes in all butone dimension. Improved technical product quality is notassociated with how fast the NPD process is completed.On the one hand, this makes sense. Sometimes, achievingtechnical product quality goals is just more difficult thaninitially envisioned, and thus takes longer than originallyexpected (Cohen et al., 1996). And one of the ways in
which firms sometimes meet commercialization datetargets or shortens cycle times is through compromisingproduct performance specifications (Rosenthal andTatikonda, 1993). On the other hand, a number of previ-ous studies have found significant and positive relation-ships between speed and technical quality (Calantone andDi Benedetto, 2000; Harter et al., 2000; Kessler andBierly, 2002). What may thus have occurred in this dataset then are some studies with positive and some studieswith negative correlations, with each canceling the otherout, and the overall general empirical result being nosignificant relationship. These opposing results also mayindicate a nonlinear association between the two vari-ables, which cannot readily be inferred from these bivari-ate correlations.
Decreasing NPD cycle time has the largest correlationwith proficiently managing market entry timing. Further-more, two separate tests suggest that this finding (andonly this finding) is robust across research design deci-sions and contexts. Thus, from a managerial perspective,explicitly managing NPD speed is most important forproducts and industries that have very narrow strategicwindows of opportunity, where, if one misses thewindow, the product likely will not return the firm’sinvestment in it. A recent example of this is Motorola inthe smart phone market, which, with a Droid 2009 marketshare of 9.7%, lags significantly behind Research inMotion (49.2%) and Apple (23.1%).7 Even with a numberof quality-differentiating features,8 forecasters predictthat the Droid will not overtake the market pioneers.
The remaining four success indicators—developmentcosts, product competitive advantage, and customer-based and financial success—all show statistically sig-nificant main effect correlations with NPD speed. Fasterdevelopment is associated with favorable outcomes inthese indicators. At first glance, these findings stand instark contrast to the literature that emphasizes the unin-tended consequences of NPD speed on other operationaloutcomes (e.g., Crawford, 1992; Lilien and Yoon, 1989;Smith and Reinertsen, 1991). Instead, they align with thesynergistic view on the link between operational NPDoutcomes. Rather than posing an additional burden onfirm resources (Rosenthal, 1992), developing productsfaster allows firms to limit expenditures such as man-hours. Similarly, products that are moved to marketquickly benefit from more accurate forecasts, such thatthey are better aligned with the rapidly changing needs of
5 Since a development speed antecedent meta-analysis is not a centralaim of this study, the results are not reported in as much detail as thosepertaining to NPD performance. Details can be provided upon request.
6 With only one correlation, Montoya-Weiss and Calantone (1994)could not perform this analysis.
7 http://seekingalpha.com/article/194442-predicting-2010-north-american-smartphone-market-share
8 http://www.pcworld.com/article/174609/verizon_droid_5_standout_features.html
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4792013;30(3):465–486
Table 7. Antecedents of Development Speed—Main Effects and Heterogeneity Analysis Summary Comparison
Present Study Chen et al. (2010)
Antecedent r (k) Htg. Antecedent r (k) Htg
Project characteristicsComplexity -.104 (14) N Complexity -.13 (18) NFirm perspective of innovativeness .013 (19) N
Newness
.06 (27) NMarket perspective of innovativeness .067 (9) NMixed/unspecified perspective of innovativeness .117 (5) NProject newness .008 (5) YProject size -.185 (3) N n/av — —
Process characteristicsStandardization .135* (10) Y
Process formalization.23* (12) Y
Formal process use .246* (9) NSupplier involvement .152* (12) N
External integration.29* (18) N
Customer involvement .277* (10) NUse of other outside assistance/information .084 (11) NGoal effectiveness .377* (15) Y Goal clarity .38* (12) NProcess concurrency .278* (13) Y Process concurrency .34* (12) YIteration/build frequency -.123 (5) N
Iteration.32* (6) Y
Testing .240* (3) YNPD team characteristics
Cross functional team use .180* (9) NInternal integration
.38* (36) NOrganizational integration .260* (12) NTeamwork quality .229* (13) NFunctional diversity .067 (7) N Functional diversity .09 (6) NTeam size -.020 (14) N (n/av) — —Team stability .385* (4) Y
Team dedication.36* (10) N
Team dedication and commitment .210* (4) YManagement style .190* (11) Y Team empowerment .22* (10) NStrength and influence of team leader .334* (10) Y Team leadership .37* (10) NTeam proximity/same site location .037 (7) N Team co-location .04 (8) N
NPD competenciesUp-front planning proficiency .105 (4) N (n/av) — —Marketing proficiency .374* (11) N (n/av) — —Technical proficiency .293* (9) N (n/av) — —Problem solving proficiency .326* (4) N (n/av) — —Team learning .318* (6) N Learning .27* (11)
Firm characteristicsExperience & alignment with core competencies .156 (7) N Team experience .38* (5) YAvailability of resources and facilities .314* (4) Y
Top management support.29* (10) N
Organizational support .341* (8) NProject priority .142* (6) N (n/av) — —Speed emphasis .254* (6) N
Emphasis on speed.19 (7) N
Presence of time-based rewards and incentives .082 (4) YCompany size .082* (13) N (n/av) — —Innovative firm climate .377* (5) N (n/av) — —
Environmental characteristicsTechnological turbulence .068 (12) N (n/av) — —Market/demographic turbulence .014 (12) N (n/av) — —Competitive intensity .017 (4) N (n/av) — —Market attractiveness/ease of entry .086 (5) N (n/av) — —
* p < .05.r, estimated true correlation corrected for sampling error and unreliability; k, number of correlations analyzed; Htg, result of heterogeneity analysis basedon the 75% rule and significant; Qb (Y = cross-situationally consistent; N = heterogeneous). Numbers in bold indicate statistically significant antecedents(p < .05); Numbers in bold and italicized indicate salient antecedents (r > .3).
480 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
consumers (Smith and Reinertsen, 1991). The synergyaccount also enjoys theoretical support from a learningperspective, in that faster product development allows formore learning loops and fosters NPD competencieswhich, in turn, translate into a more efficient developmentprocess (Eisenhardt and Tabrizi, 1995).
It is important to note, however, that these meta-analysis findings should not necessarily be interpreted asrefuting the possibility of trade-offs between successdimensions. As highlighted by Swink, Talluri, andPandejpong (2006), trade-offs are more apparent inhighly efficient projects than in those operating with sub-optimal levels of efficiency. This is because projects oper-ating with low efficiency have more scope to utilizeprocesses, techniques, and resources to achieve improve-ments on multiple fronts. Efficient projects, on the otherhand, have already exhausted these possibilities andcannot increase efficiency further in the absence of addi-tional resources. It is therefore reasonable to suggest thatthe lack of meta-analytic evidence for performance trade-offs can be due to the disproportionate representation ofinefficient projects in the primary studies.
Results of the heterogeneity analyses conducted alsoare indicative of a more complex speed–performancerelationship than indicated by the main effects resultsalone. Proficiency in market-entry timing aside, no otherperformance measure displays heterogeneity in its rela-tionship with development speed. Thus, it is possible forthe trade-off and synergy scenarios to both be appli-cable, depending upon the features of the research andthe setting in which it was conducted. The moderator
analyses conducted for the heterogeneous relationshipsprovide detailed insight on how research design fea-tures, choice of speed measure, and study context bearon the findings. As Table 8 shows, none of the previousdevelopment speed meta-analyses included potentialmoderators.
Worthy of particular attention are the moderatinginfluences on the speed–technical product performancelink, the main effect estimate for which failed to reachsignificance. However, close inspection of Tables 4–6reveals that the relationship is in fact statistically signifi-cant for certain subgroups of studies. In particular,research design and speed variable measurement deci-sions can change this insignificant main effect relation-ship into a significant one through moderation.
First, while random samples generated no significantrelationship between the two variables, correlations forthe purposively drawn (nonrandom) samples were sig-nificant and moderately strong (Cohen and Cohen, 1983).An implication of this finding is that there may be somespecific industries, as Harter et al. (2000) found for soft-ware development, or other specific contexts for whichthe general tendency of the relationship will be positive,and yet other contexts exist in which it is not. Academics,then, in future research need to state explicitly whetherthe sample was randomly chosen (and thus they wouldnot expect a relationship between speed and technicalproduct quality), or whether they are purposively choos-ing a sample for which speed might be associated withperformance for some theoretical or empirical reason.Potential contexts to test further include industry (soft-
Table 8. Results across Meta-analytic Studies
StudyMontoya-Weiss and
Calantone (1994)Henard and Szymanski
(2001) Pattikawa et al. (2006) Present Study (2010)
Speed effect size .18 .22* .39* .31*Heterogeneity Not tested: Low #
of studies;need replication
Variation explained bysampling andmeasurement <75%(test for moderators)
Variation explained bysampling andmeasurement <75%(test for moderators)
Variation explained by sampling andmeasurement <75% and Q statisticsignificant (test for moderators)
Moderators tested • None • Level of respondent• # Success items• Objective/subjective
success measure• Innovation type• Sample geography
• None • Sample selection• Analysis level• # informants• # speed items• Absolute/relative speed measure• Objective/subjective speed measure• Innovation type• # industries
Moderation found • Objective/subjectivesuccess measure
• Sample selection• # informants• # speed items• Absolute/relative speed measure• Innovation type
* p < .05.
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4812013;30(3):465–486
ware has already shown this effect), technology depen-dency, and services versus products.
Second, when speed is measured with just a singleitem, there is no significant speed–technical productquality relationship. However, when speed is measured bymultiple items, the correlation is significant and large(r = .41, p < .05). This pattern aligns with the notion thatmulti-item scales capture the different facets of constructsbetter (Henard and Szymanski, 2001) and yield largereffect sizes than single-item scales (Kirca et al., 2005).
As noted earlier, the other five success dimensions—composite success, development costs, product competi-tive advantage, customer-based success, and financialsuccess—all show statistically significant main effectcorrelations with NPD speed. However, and more impor-tantly, the correlations for all five of these success dimen-sions also are cross-situationally heterogeneous. In otherwords, the magnitude of those correlations can differdepending upon research decisions and contexts. Further-more, it is surprising how many initially statistically sig-nificant main effect correlations reduce to completeinsignificance for different relationship moderators, asshown in Table 9.
The three most important moderating variables aresample randomness, number of items in the speed mea-surement scale, and whether the study used absolute orrelative measures of speed (Tables 4 and 5), all method-ological decisions. Using random samples, multi-itemscales and absolute measures reflect current standardsfor achieving methodological rigor in social scienceresearch. Random samples produce true empirical gener-alizations that hold across all contexts (which meta-analysis strives to uncover). As noted above, multipleitems allow a measure to reflect all of a construct’svarious facets. Finally, using absolute measures ofspeed eliminates comparative judgment influences onoutcomes.
Interestingly, as Table 9 shows, using random samplesand absolute measures of speed, although more
methodologically rigorous, is more likely to produceinsignificant speed/success relationships—while theless methodologically rigorous choices of nonrandomsamples and relative speed measures produce more sta-tistically significant relationships. The only successmeasure that differs from these results is customer-basedsuccess, where the relationship with speed is just theopposite—insignificant for nonrandom samples, andsignificant for random ones. Using multiple items formeasuring speed (the methodologically more rigorousdecision) always produces higher correlations withsuccess than using a single measurement item (statisti-cally significantly higher in six of seven success dimen-sions), and for three dimensions, the single-itemrelationships become insignificant.
Of the previous meta-analyses that have investigatedthe NPD speed/success relationship, none have includedpotential moderators associated with measurement deci-sions for the speed construct. Furthermore, only Henardand Szymanski (2001) investigated any research designor contextual moderators that overlapped with the analy-ses performed from the data available for this research.Both studies investigated differences in the relationshipthat depended upon whether the subsamples includedphysical goods versus services (Henard and Szymanski,2001) or goods versus a mix of goods and services (thisinvestigation). Henard and Szymanski (2001) found nosignificant moderation, while this study’s results show anincrease in the speed/success relationship when bothgoods and services constitute the sample, reinforcing therecommendation above concerning context and samplingrandomness as control variables in future research.
Implications
The moderating findings have important implicationsboth for academics pursuing research in NPD success andcycle time as well as for practitioners developing newproducts. From an academic perspective, the results of
Table 9. Moderators That Reduce Main Effect Correlations to Insignificance
Sampling Procedure Number of Speed Items Speed Measure Product Type
New product success:Overall/composite Random Single Physical products
Operational outcomes:Product measures
Technical product quality Random SingleProduct competitive advantage Single Absolute
External outcomes:Customer Nonrandom AbsoluteFinancial Absolute
482 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
this research (along with moderation findings forPattikawa et al. [2006] and Chen et al. [2010]) show thatmethodological differences are very important in under-standing potential NPD speed–success relationships. Forexample, Pattikawa et al. (2006) concluded from theirmoderation analysis on antecedent–performance relation-ships that the number of items in the speed measure, andwhether it was absolute or relative and objective (versussubjective), all were important differentiators of theirmagnitude. In addition to absolute versus relative and thenumber of indicators in the speed measure, the results ofthis investigation also indicate that sample randomnessmatters. Most notably, these four methodological deci-sions do not always produce results in the same directionfor all speed relationships.
Academics are confronted with a series of method-ological decisions shaped not only by the demands of theresearch question but also by practical constraints. It isnot the aim of this research to delineate what the“correct” courses of action are in relation to these deci-sions. However, the moderating effects of these decisions,as well as the directional inconsistencies in the way theyare manifest in different speed relationships highlightsthe importance of clear communication. In this way, onecan better assess the accuracy, meaningfulness, andcross-sectional generalizability of empirical findingsregarding NPD speed.
Furthermore, given the large percentage of variabilityin effect sizes which remained after sampling and mea-surement errors are corrected for, there is every chancethat there were statistically significant moderators whichcould not be included in the analyses because of datalimitations. This possibility further echoes the need forstudy characteristics to be reported clearly, accurately,and completely for the results to be meaningfully inter-pretable. Therefore, despite the main effects results point-ing overwhelmingly to a positive association betweenspeed and NPD performance, it is possible that theseresults do not necessarily validate synergistic accounts atthe expense of trade-off arguments. On the contrary, thesubsequent moderator analyses suggest the validity ofboth perspectives under different conditions.
For managers, the results of this study suggest thatthere may be some divisions or product categories in theirfirms for which decreasing NPD cycle time may increasesuccess, but there may be some others too for which itmay not. Before setting off on a unilateral corporateprogram to reduce NPD cycle time, therefore, they maywant to analyze archival data for their different divisionsto discern which ones have historical results that supportimplementing such a program.
Additionally, based on the results concerning market-entry timing proficiency, they should consider cycle timereduction programs for product categories with thenarrowest windows of opportunities first. Cycle timereduction may not be as important to hit windows ofopportunity for slowly evolving categories.
Also worthy of emphasis is the possibility of relation-ships being manifest differently under different organiza-tional contexts. Although an extensive analysis ofmoderators pertaining to the organizational context wasinfeasible due to data limitations, the sheer amount ofunexplained variance remaining after artifact correctionmay be indicative of different organizational settingsfavoring different speed–performance associations.
The results also do not afford any inference ofoutcome directionality, as sufficient data to create and testa meta-analytic structural model for these constructs wasunavailable. This last point underlines an importantimplication for managers, as well as one area of criticalneed for future research. From this analysis, indeed fromthe sum of the previous research done on NPD cycle timeand success, it is unclear which construct, speed, orsuccess is the cart, and which is the horse. We cannotconclude whether faster development speed leads toimproved NPD success, or whether companies that aremore successful at developing new products also becomespeedier than others due to their development capabili-ties. There is no evidence to show which capability leads,and which follows, nor does any other study, to the best ofour knowledge, have any evidence to this. This is animportant point for practitioners.
Limitations and Further Research
As with all research, this study has a number of limita-tions which should be taken into account when inter-preting and evaluating its results. As is typical ofmeta-analytic reviews, a construct could be consideredfor synthesis only if the minimum recommended numberof three studies reported it. To include as many effectsizes as possible, some related specific success itemswere collapsed under broader construct headings. Whilethe classification of NPD success measures used in thisstudy has solid grounding in previous conceptual andempirical work and the use of broader meta-factors servesthe interest of parsimony, it is possible that this approachwill have prevented us from uncovering certain additionalnuances in the speed–success relationship.
Similarly, not all of the empirical work initially iden-tified as relevant provided the information necessary forquantitative synthesis. It was possible to obtain the nec-
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4832013;30(3):465–486
essary missing information for only a subset of thestudies from authors. The absence of information onstudy characteristics such as moderator variables and alack of sufficient variability across studies with regard tosome characteristics posed a major constraint on theextent to which a comprehensive assessment of moderat-ing factors could be conducted and limited the number ofrelationships for which the effect of a particular modera-tor could be examined.
A final limitation is that this study used bivariate cor-relations between development speed and new productsuccess, as did Henard and Szymanski (2001) and Chenet al. (2010). Therefore, these results do not reflect anypotentially nonlinear relationships between these vari-ables (e.g., Langerak and Hultink, 2006; Langerak et al.,2008; Lukas and Menon, 2004).
This study has taken the first step toward a systematicreview of development speed which brings together theantecedents and consequences. The complementary useof meta-analysis and structural equation modelingmethods such as path analysis (e.g., Viswesvaran andOnes, 1995), which this study was unable to implementdue to the large number of empty cells in the pooledmeta-analytical correlation matrix, would be invaluablefor comprehensive synthesis provided that its practicalchallenges are overcome.
References
Adams-Bigelow, M., and A. Griffin. 2005. Product development cycle timeand success: New results from PDMA’s comparative performanceassessment study. Proceedings of the Product Development & Man-agement Association International Conference, San Diego, CA,October.
Aguinis, H., and C. A. Pierce. 1998. Testing moderator variable hypothesesmeta-analytically. Journal of Management 24 (5): 577–92.
Ali, A. 2000. The impact of innovativeness and development time on newproduct performance for small firms. Marketing Letters 11 (2): 151–63.
Ali, A., R. Krapfel Jr., and D. Labahn. 1995. Product innovativeness andentry strategy: Impact on cycle time and break-even time. Journal ofProduct Innovation Management 12 (1): 54–69.
Ancona, D. G., and D. F. Caldwell. 1992. Demography and design: Pre-dictors of new product team performance. Organization Science 3 (3):321–41.
Aronson, Z. H., R. R. Reilly, and G. S. Lynn. 2006. The impact of leaderpersonality on new product development teamwork and performance:The moderating role of uncertainty. Journal of Engineering and Tech-nology Management 23 (3): 221–47.
Atuahene-Gima, K. 2003. The effects of centrifugal and centripetal forceson product development speed and quality: How does problem solvingmatter? Academy of Management Journal 46 (3): 359–73.
Balkundi, P., and D. A. Harrison. 2006. Ties, leaders, and time in teams:Strong inference about network structure’s effects on team viability andperformance. Academy of Management Journal 49 (1): 49–68.
Bayus, B. L. 1997. Speed-to-market and new product performance trade-offs. Journal of Product Innovation Management 14 (6): 485–87.
Bayus, B. L., S. Jain, and A. G. Rao. 1997. Too little, too early: Introductiontiming and new product performance in the personal digital assistantindustry. Journal of Marketing Research 34 (1): 50–63.
Boer, M., and R. Logendran. 1999. A methodology for quantifying theeffects of product development on cost and time. IIE Transactions31(4): 365–78.
Brown, S. L., and K. M. Eisenhardt. 1995. Product development: Pastresearch, present findings, and future directions. Academy of Manage-ment Review 20 (2): 343–78.
Brown, S. P., and R. A. Peterson. 1993. Antecedents and consequences ofsalesperson job satisfaction: Meta-analysis and assessment of causaleffects. Journal of Marketing Research 30 (1): 63–77.
Calantone, R. J., and C. A. Di Benedetto. 2000. Performance and time tomarket: Accelerating cycle time with overlapping stages. IEEE Trans-actions on Engineering Management 47 (2): 232–44.
Calantone, R. J., S. K. Vickery, and C. Dröge. 1995. Business performanceand strategic new product development activities: An empirical inves-tigation. Journal of Product Innovation Management 12 (3): 214–23.
Carbonell, P., and A. I. Rodriguez-Escudero. 2006. The impact of marketcharacteristics and innovation speed on perceptions of positionaladvantage and new product performance. International Journal ofResearch in Marketing 23 (1): 1–12.
Chen, J., F. Damanpour, and R. R. Reilly. 2010. Understanding antecedentsof new product development speed: A meta-analysis. Journal of Opera-tions Management 28 (1): 17–33.
Chen, J., R. R. Reilly, and G. S. Lynn. 2005. The impacts of speed-to-market on new product success: The moderating effects of uncertainty.IEEE Transactions on Engineering Management 52 (2): 199–212.
Clark, K., and T. Fujimoto. 1991. Product development performance: Strat-egy organization and management in the world auto industry. Boston:Harvard Business School Press.
Cohen, J. 1992. A power primer. Psychological Bulletin 112 (1): 155–59.
Cohen, J., and P. Cohen. 1983. Applied multiple regression/correlationanalysis for the behavioral sciences (2nd ed.). Hillsdale, NJ: LawrenceErlbaum Associates.
Cohen, M. A., J. Eliashberg, and T. H. Ho. 1996. New product develop-ment: The performance and time-to-market tradeoff. ManagementScience 42 (2): 173–86.
Cooper, R. G., and E. J. Kleinschmidt. 1987. Success factors in productinnovation. Industrial Marketing Management 16 (3): 215–23.
Cordero, R. 1991. Managing for speed to avoid product obsolescence: Asurvey of techniques. Journal of Product Innovation Management 8(4): 283–94.
Crawford, M. 1992. The hidden costs of accelerated product development.Journal of Product Innovation Management 9 (3): 188–99.
Crosno, J. L., and R. Dahlstrom. 2008. A meta-analytic review of oppor-tunism in exchange relationships. Journal of the Academy of MarketingScience 36 (2): 191–201.
Damanpour, F. 1991. Organizational innovation: A meta-analysis of effectsof determinants and moderators. Academy of Management Journal 34(3): 555–90.
Davis, P. S., C. C. Dibrell, and B. D. Janz. 2002. The impact of time on thestrategy–performance relationship implications for managers. Indus-trial Marketing Management 31 (4): 339–47.
Dröge, C., J. Jayaram, and S. K. Vickery. 2000. The ability to minimize thetiming of new product development and introduction: An examinationof antecedent factors in the North American automobile supplier indus-try. Journal of Product Innovation Management 17 (1): 24–40.
Eisenhardt, K., and B. N. Tabrizi. 1995. Accelerating adaptive processes:Product innovation in the global computer industry. AdministrativeScience Quarterly 40 (1): 84–110.
Emmanuelides, P. A. 1993. Towards an integrative framework of perfor-mance in product development projects. Journal of Engineering andTechnology Management 10 (4): 363–92.
Flint, D. J. 2002. Compressing new product success-to-success cycle time:Deep customer value understanding and idea generation. IndustrialMarketing Management 31 (4): 305–15.
484 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486
Franke, G. R., and J. Park. 2006. Salesperson adaptive selling behavior andcustomer orientation: A meta-analysis. Journal of Marketing Research43 (4): 693–702.
Garcia, N., M. J. Sanzo, and J. A. Trespalacios. 2008. New product internalperformance and market performance: Evidence from Spanish firmsregarding the role of trust, interfunctional integration, and innovationtype. Technovation 28 (11): 713–25.
Gerwin, D., and N. J. Barrowman. 2002. An evaluation of research onintegrated product development. Management Science 48 (7): 938–53.
Geyskens, I., J. Steenkamp, and N. Kumar. 1998. Generalizations abouttrust in marketing channel relationships using meta-analysis. Interna-tional Journal of Research in Marketing 15 (3): 223–48.
Geyskens, I., J. E. M. Steenkamp, and N. Kumar. 2006. Make, buy, or ally:A transaction cost theory meta-analysis. Academy of ManagementJournal 49 (3): 519–43.
Gleser, L. J., and I. Olkin. 1994. Stochastically dependent effect sizes. InThe handbook of research synthesis, ed. H. Cooper and L. V. Hedges,339–55. New York: Russell Sage Foundation.
González, F. J. M., and T. M. B. Palacios. 2002. The effect of new productdevelopment techniques on new product success in Spanish firms.Industrial Marketing Management 31 (3): 261–71.
Griffin, A. 1993. Metrics for measuring product development cycles times.Journal of Product Innovation Management 10 (2): 112–25.
Griffin, A. 1997. The effect of project and process characteristics onproduct development cycle time. Journal of Marketing Research 34 (1):24–35.
Griffin, A. 2002. Product development cycle time for business-to-businessproducts. Industrial Marketing Management 31 (4): 291–304.
Griffin, A., and A. L. Page. 1993. An interim report on measuring productdevelopment success and failure. Journal of Product Innovation Man-agement 10 (4): 291–308.
Griffin, A., and A. L. Page. 1996. PDMA success measurement project:Recommended measures for product development success and failure.Journal of Product Innovation Management 13 (6): 478–96.
Gupta, A. K., K. Brockhoff, and U. Weisenfeld. 1992. Making trade-offs inthe new product development process: A German/US comparison.Journal of Product Innovation Management 9 (1): 11–18.
Hart, S. J. 1993. Dimensions of success in new product development: Anexploratory analysis. Journal of Marketing Management 9 (1): 23–41.
Harter, D. E., M. S. Krishnan, and S. A. Slaughter. 2000. Effects of processmaturity on quality, cycle time, and effort in software product devel-opment. Management Science 46 (4): 451–66.
Hedges, L. V., and I. Olkin. 1985. Statistical methods for meta-analysis.New York: Academic Press.
Henard, D. H., and D. M. Szymanski. 2001. Why some new products aremore successful than others. Journal of Marketing Research 38 (3):362–75.
Hitt, M. A., R. E. Hoskisson, and R. D. Nixon. 1993. A mid-range theory ofinterfunctional integration, its antecedents and outcomes. Journal ofEngineering and Technology Management 10 (1-2): 161–85.
Hoegl, M., K. Weinkauf, and H. G. Gemünden. 2004. Interteam coordina-tion, project commitment, and teamwork in multiteam R&D projects: Alongitudinal study. Organization Science 15 (1): 38–55.
Huang, X., G. N. Soutar, and A. Brown. 2004. Measuring new productsuccess: An empirical investigation of Australian SMEs. IndustrialMarketing Management 33 (2): 117–23.
Hultink, E. J., and H. S. J. Robben. 1995. Measuring new product success:The difference that time perspective makes. Journal of Product Inno-vation Management 12 (5): 392–405.
Hunter, J. E., and F. L. Schmidt. 1990. Methods of meta-analysis: Correct-ing error and bias in research findings. New York: Academic Press.
Hunter, J. E., and F. L. Schmidt. 2000. Fixed effects vs. random effectsmeta-analysis models: Implications for cumulative research knowl-edge. International Journal of Selection and Assessment 8 (4): 275–92.
Ittner, C. D. and D. F. Larcker. 1997. Product development cycle time andorganizational performance. Journal of Marketing Research 34 (1):13–23.
Jayaram, J. and R. Narasimhan. 2007. The influence of new product devel-opment competitive capabilities on project performance. IEEE Trans-actions on Engineering Management 54 (2): 241–56.
Johnson, W. H. A., Z. Piccolotto, and R. Filippini. 2009. The impacts oftime performance on market knowledge competence on new productsuccess: An international study. IEEE Transactions on EngineeringManagement 56 (2): 219–28.
Joshi, A., and H. Roh. 2009. The role of context in work team diversityresearch: A meta-analytic review. Academy of Management Journal 52(3): 599–627.
Kessler, E. H. and P. E. Bierly III. 2002. Is faster really better? An empiricaltest of the implications of innovation speed. IEEE Transactions onEngineering Management 49 (1): 2–12.
Kessler, E. H., and A. K. Chakrabarti. 1996. Innovation speed: A conceptualmodel of context, antecedents and outcomes. Academy of ManagementReview 21 (4): 1143–91.
Kessler, E. H., and A. K. Chakrabarti. 1999. Speeding up the pace of newproduct development. Journal of Product Innovation Management 16(3): 231–47.
Kirca, A. H., S. Jayachandran, and W. O. Bearden. 2005. Market orienta-tion: A meta-analytic review and assessment of its antecedents andimpact on performance. Journal of Marketing 69 (2): 24–41.
Kusunoki, K., I. Nonaka, and A. Nagata. 1998. Organizational capabilitiesin product development of Japanese firms: A conceptual framework andempirical findings. Organization Science 9 (6): 699–718.
Langerak, F. and E. J. Hultink. 2005. The impact of new product develop-ment acceleration approaches on speed and profitability: Lessons forpioneers and fast followers. IEEE Transactions on Engineering Man-agement 52 (1): 30–42.
Langerak, F. and E. J. Hultink. 2006. The impact of product innovativenesson the link between development speed and new product profitability.Journal of Product Innovation Management 23 (3): 203–14.
Langerak, F., E. J. Hultink, and A. Griffin. 2008. Exploring mediating andmoderating influences on the links among cycle time, proficiency inentry timing, and new product profitability. Journal of Product Inno-vation Management 25 (4): 370–85.
Langerak, F., S. A. Rijsdijk, and K. Dittrich. 2009. Development time andnew product sales: A contingency analysis of product innovativeness.Marketing Letters 20 (4): 339–413.
Larson, E. W. and D. H. Gobeli. 1989. Significance of project managementstructure on development success. IEEE Transactions on EngineeringManagement 36 (2): 119–25.
Lee, H., K. G. Smith, M. Grimm, and A. Schomburg. 2000. Timing, orderand durability of new product advantages with imitation. StrategicManagement Journal 21 (1): 23–30.
Li, T., and R. J. Calantone. 1998. The impact of market knowledge com-petence on new product advantage: Conceptualization and empiricalexamination. Journal of Marketing 62 (4): 13–29.
Lieberman, M. B., and D. B. Montgomery. 1988. First-mover advantages.Strategic Management Journal 9 (1): 41–58.
Lilien, G. L., and E. Yoon. 1989. Determinants of new industrial productperformance: A strategic reexamination of the empirical literature.IEEE Transactions on Engineering Management 36 (1): 3–10.
Lipovetsky, S., A. Tishler, D. Dvir, and A. Shenhar. 1997. The relativeimportance of project success dimensions. R&D Management 27 (2):97–106.
Lipsey, M. W., and D. B. Wilson. 2001. Practical meta-analysis. ThousandOaks, CA: Sage Publications.
Loch, C. H., and C. Terwiesch. 1998. Communication and uncertainty inconcurrent engineering. Management Science 44 (8): 1032–48.
Lukas, B. A. and A. Menon. 2004. New product quality: Intended andunintended consequences of new product development speed. Journalof Business Research 57 (11): 1258–64.
META-ANALYSIS ON NEW PRODUCT DEVELOPMENT SPEED J PROD INNOV MANAG 4852013;30(3):465–486
Lynn, G. S., K. D. Abel, W. S. Valentine, and R. C. Wright. 1999. Keyfactors in increasing speed to market and improving new productsuccess rates. Industrial Marketing Management 28 (4): 319–26.
Mallick, D. N., and R. G. Schroeder. 2005. An integrated framework formeasuring product development performance in high technologyindustries. Production and Operations Management 14 (2): 142–58.
Mansfield, E. 1988. The speed and cost of industrial innovation in Japanand the United States: External vs. internal technology. ManagementScience 34 (10): 1157–68.
Meyer, C. 1993. Fast cycle time: How to align purpose, strategy, andstructure for SPEED. New York: Free Press.
Meyer, M. H. and J. M. Utterback. 1995. Product development cycle timeand commercial success. IEEE Transactions on Engineering Manage-ment 42 (4): 297–304.
Millson, M., S. P. Raj, and D. Wilemon. 1992. A survey of majorapproaches for accelerating new product development. Journal ofProduct Innovation Management 9 (1): 53–69.
Montoya-Weiss, M. M., and R. Calantone. 1994. Determinants of newproduct performance: A review and meta-analysis. Journal of ProductInnovation Management 11 (5): 397–417.
Moorman, C., and A. S. Miner. 1998. The convergence of planning andexecution: Improvisation in new product development. Journal of Mar-keting 62 (3): 1–20.
Morgan, L. O., R. M. Morgan, and W. L. Moore. 2001. Quality and time-to-market trade-offs when there are multiple product generations.Manufacturing & Service Operations Management 3 (2): 89.
Murmann, P. A. 1994. Expected development time reductions in theGerman mechanical engineering industry. Journal of Product Innova-tion Management 11 (3): 236–52.
Orlitzky, M., F. L. Schmidt, and S. L. Rynes. 2003. Corporate social andfinancial performance: A meta-analysis. Organization Studies 24 (3):403–41.
Pattikawa, L. H., E. Verwaal, and H. R. Commandeur. 2006. Understandingnew product project performance. European Journal of Marketing 40(11/12): 1178–93.
Rabino, S., and A. Wright. 1993. Accelerated product introductions andemerging managerial accounting perspectives: Implications for market-ing managers in the technology sector. Journal of Product InnovationManagement 10 (2): 126–35.
Rindfleisch, A., and C. Moorman. 2001. The acquisition and utilization ofinformation in new product alliances: A strength-of-ties perspective.Journal of Marketing 65 (2): 1–18.
Rochford, L. A., and W. R. Rudelius. 1992. How involving more functionalareas within a firm affects the new product process. Journal of ProductInnovation Management 9 (4): 287–99.
Rodriguez-Cano, C., F. A. Carrillat, and F. Jaramillo. 2004. A meta-analysisof the relationship between market orientation and business perfor-mance: Evidence from five continents. International Journal ofResearch in Marketing 21 (2): 179–200.
Rosenthal, S. R. 1992. Effective product design and development: How tocut lead time and increase customer satisfaction. Homewood, IL: Busi-ness One Irwin.
Rosenthal, S. R., and M. V. Tatikonda. 1993. Time management in newproduct development: Case study findings. IEEE Engineering Manage-ment Review 21 (3): 13–20.
Rusinko, C. A. 1999. Exploring the use of design-manufacturing integra-tion (DMI) to facilitate product development: A test of some practices.IEEE Transactions on Engineering Management 46 (1): 56–71.
Salomo, S., J. Weise, and H. G. Gemünden. 2007. NPD planning activitiesand innovation performance: The mediating role of process manage-ment and the moderating effect of product innovativeness. Journal ofProduct Innovation Management 24 (4): 285–302.
Schulze, A., and M. Hoegl. 2006. Knowledge creation in new productdevelopment projects. Journal of Management 32 (2): 210–36.
Sethi, R. 2000. New product quality and product development teams.Journal of Marketing 64 (2): 1–14.
Smith, P. G., and D. G. Reinertsen. 1991. Developing new products in halfthe time. New York: Van Nostrand Reinhold.
Song, X. M., M. M. Montoya-Weiss, and J. B. Schmidt. 1997. Antecedentsand consequences of cross-functional cooperation: A comparison ofR&D, manufacturing, and marketing perspectives. Journal of ProductInnovation Management 14 (1): 35–47.
Stalk, G. 1988. Time—the next source of competitive advantage. HarvardBusiness Review 66 (4): 41–51.
Stalk, G., and T. M. Hout. 1990. Competing against time: How time-basedcompetition is reshaping global markets. New York: Free Press.
Stock, W. A. 1994. Systematic coding for research synthesis. In The hand-book of research synthesis, ed. H. Cooper and L. V. Hedges, 231–44.New York: Russell Sage Foundation.
Swink, M., and X. M. Song. 2007. Effects of marketing-manufacturingintegration on new product development time and competitive advan-tage. Journal of Operations Management 25 (1): 203–17.
Swink, M., S. Talluri, and T. Pandejpong. 2006. Faster, better, cheaper: Astudy of NPD project efficiency and performance tradeoffs. Journal ofOperations Management 24 (5): 542–62.
Tatikonda, M. V., and M. M. Montoya-Weiss. 2001. Integrating operationsand marketing perspectives of product innovation: The influence oforganizational process factors and capabilities on development perfor-mance. Management Science 47 (1): 151–72.
Tatikonda, M. V., and S. R. Rosenthal. 2000. Technology novelty, projectcomplexity, and product development project execution success: Adeeper look at task uncertainty in product innovation. IEEE Transac-tions on Engineering Management 47 (1): 74–87.
Terwiesch, C., and C. H. Loch. 1999. Measuring the effectiveness of over-lapping development activities. Management Science 45 (4): 455–65.
Troy, L. C., T. Hirunyawipada, and A. K. Paswan. 2008. Cross-functionalintegration and new product success: An empirical investigation of thefindings. Journal of Marketing 72 (6): 132–46.
Ulrich, K., D. Sartorius, S. Pearson, and M. Jakiela. 1993. Including thevalue of time in design-for-manufacturing decision making. Manage-ment Science 39 (4): 429–47.
Vandenbosch, M., and T. Clift. 2002. Dramatically reducing cycle timesthrough flash development. Long Range Planning 35 (6): 567–89.
Venkatraman, N., and V. Ramanujam. 1986. Measurement of businessperformance in strategy research: A comparison of approaches.Academy of Management Review 11 (4): 801–14.
Vesey, J. T. 1992. Time-to-market: Put speed in product development.Industrial Marketing Management 21 (2): 151–58.
Viswesvaran, C., and D. Ones. 1995. Theory testing: Combining psycho-metric meta-analysis and structural equations modeling. PersonnelPsychology 48 (4): 865–85.
Von Braun, C. F. 1990. The acceleration trap. Sloan Management Review32 (1): 49–58.
Zirger, B. J., and J. L. Hartley. 1994. A conceptual model of productdevelopment cycle time. Journal of Engineering and Technology Man-agement 11 (3-4): 229–51.
Zirger, B. J., and M. A. Maidique. 1990. A model of new product develop-ment: An empirical test. Management Science 36 (7): 867–83.
Supporting Information
Additional Supporting Information may be found in the online version ofthis article at the publisher’s web-site:
Appendix S1. Meta-Analytic Database (Consequences).Appendix S2. Meta-Analytic Database (Antecedents).Appendix S3. Studies in the Meta-Analytic Database (otherwise not citedin text).
486 J PROD INNOV MANAG P. CANKURTARAN ET AL.2013;30(3):465–486