You are on page 1of 21

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/281174629

Product Modularity, Process Modularity, and New Product Introduction


Performance: Does Complexity Matter?

Article in Production and Operations Management · August 2015


DOI: 10.1111/poms.12495

CITATIONS READS

108 3,348

4 authors, including:

Xenophon Koufteros Roger Calantone


Texas A&M University Michigan State University
72 PUBLICATIONS 6,991 CITATIONS 273 PUBLICATIONS 31,789 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Xenophon Koufteros on 14 October 2018.

The user has requested enhancement of the downloaded file.


Vol. 0, No. 0, xxxx–xxxx 2015, pp. 1–20 DOI 10.1111/poms.12495
ISSN 1059-1478|EISSN 1937-5956|15|00|0001 © 2015 Production and Operations Management Society

Product Modularity, Process Modularity, and New


Product Introduction Performance: Does Complexity
Matter?
Shawnee K. Vickery
The Eli Broad College of Business, Michigan State University, East Lansing, Michigan 48824, USA, vickery@msu.edu

Xenophon Koufteros
Mays Business School, Texas A & M University, College Station, Texas 77843, USA, xkoufteros@mays.tamu.edu

Cornelia Dröge, Roger Calantone


The Eli Broad College of Business, Michigan State University, East Lansing, Michigan 48824, USA,
droge@msu.edu, rogercal@broad.msu.edu

odularity has the potential to impact various facets of new product introduction performance including product
M development lead time, frequency of new product introduction, on time introduction and product innovation. The
impact of modularity on new product introduction performance, however, may vary for different levels of product and
process complexity. This study empirically investigates relationships between perceptual measures of product modularity,
process modularity, and new product introduction performance and explores whether an objective product/process com-
plexity measure moderates these relationships. Using survey-based methodology we probe both manufacturers of techni-
cally simple products and technically complex products. Hierarchical regression models are used to test hypotheses
concerning the main effects of product and process modularity and the effects of their interactions with complexity on
new product introduction performance. The results show that the main effect of product modularity was positive and its
interaction with complexity was disordinal and negative, suggesting that the positive effect of product modularity on new
product introduction performance is dampened when complexity is high. For process modularity, only the interaction
effect (positive) was statistically significant and it was also disordinal in nature. Thus, the effect of process modularity on
new product introduction performance is heightened when complexity is high. The implications of these findings are dis-
cussed and more specific theoretical and managerial implications are delineated by examining the impacts of these main
and interaction effects on individual measures of new product introduction performance (frequency of new product intro-
duction, product development lead times, product innovation, and on-time product launch).
Key words: product modularity; process modularity; complexity; new product introduction; empirical
History: Received: August 2013; Accepted: July 2015 by Stelios Kavadias, after 3 revisions.

modular systems can handle environmental perturba-


1. Introduction tions more effectively than tightly coupled elements.
Product and/or process modularity are possible Modular structures are also potentially useful when
approaches for addressing the challenges associated systems become so large and interdependence
with rapidly changing technologies, clockspeed pro- between elements of the system so vast that inte-
duct development cycles, and sophisticated customer grated design efforts become difficult. Ethiraj and
demands. Modularity refers to the structure of a pro- Levinthal (2004) specifically note that “modular
duct or process that is comprised of smaller subsys- designs are a useful means of managing complexity”
tems that can be designed independently yet function (p. 159). While much of the literature on modularity
together holistically (Baldwin and Clark 1997, Ro has focused on product design, similar arguments can
et al. 2007, Schilling 2000, Ulrich 1995, Ulrich and be made concerning the potential beneficial effects of
Eppinger 2004). Pil and Cohen (2006) suggest that process modularity.
modular systems interdependently perform distinc- Core aspects of new product introduction perfor-
tive functions while modular elements can evolve mance include development lead time, on-time intro-
autonomously without effectively amending the over- duction, product innovation, and frequency of new
all structure of the system. Therefore, they argue that product introduction. The emphasis on the timing

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
2 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

aspects is especially salient because accelerated and as they are described by a limited set of components,
on-time introduction are fast becoming a new front in they have few and not very intricate interfaces
global competition (Williamson and Yin 2014). Modu- between components, and they boast few secondary
lar structures have been touted to enhance new pro- functions that combine (or interact) to produce the
duct introduction performance overall and facilitate product’s primary function. Likewise, the production
rapid introduction in particular (Ethiraj and Levinthal process for toasters is far less complicated than that of
2004). According to Pil and Cohen (2006), modular an automobile manufacturer. If both companies
product architectures have the potential to improve espouse similar aspects of modularity (such as
new product introduction performance since they designing the products where components can easily
reduce the size of the design problem, diminish inter- be swapped, specifying standardized interfaces,
dependence among the product’s elements, and designing products that can be decomposed into dif-
lessen the sensitivity of functional requirements and ferent modules, and designing processes that can be
performance to variations in design parameters rapidly disassembled and reconfigured), should we
(El-Haik and Yang 1999). Ethiraj and Levinthal (2004) expect the same relative effects on new product intro-
posit that modular product designs can improve new duction performance? Also, it is not clear whether
product introduction because of autonomous (within companies should invest in both types of modularity
component) and modular (mix and match of mod- (product and process) given a specific level of com-
ules) innovation. On the process side of things, modu- plexity. Applying the same level of modularity for
larity embodies the notion of a dynamic network of products or processes characterized by different
relatively autonomous production modules that can levels of complexity may not be warranted. It is an
easily be reconfigured to support and accelerate the empirical question whether there are boundary condi-
launch of a new product. The ability to rapidly recon- tions for the effectiveness of each type of modularity.
figure a production network should engender more This study addresses a core research question: Does
frequent introductions of new products into the mar- complexity moderate the relationship between modu-
ketplace. Although the benefits of modularity have larity and new product introduction performance?
been widely broadcasted, relatively few studies have Alternatively, are the effects of modularity on new
empirically examined its performance effects and product introduction performance equally salient
none have explicitly considered its effect on new pro- across levels of complexity? Complexity is often mea-
duct introduction performance. Thus, the efficacy of sured as the number of components comprising the
modularity as an explanatory variable of performance whole. Blau and Schoenherr (1971) suggest that com-
is still questioned. Over a wide range of performance plexity is “the number of structural components that
measures, Jacobs et al. (2011) found no statistically are formally distinguished” (p. 302). Similarly, La
significant effects for process modularity while the Porte (1975) states that a complex system has many
results for product modularity were mixed (e.g., parts, while a simple system only has few parts. This
Ahmad et al. 2010, Antonio et al. 2007, Jacobs et al. study defines complexity to encompass both com-
2011, Lin 2003, Worren et al. 2002). plexity in product as well as in process; it acknowl-
Modular designs are becoming omnipresent (see edges that the effects of modularity on new product
for instance the discussion and examination of several introduction performance are subject to the level of
industries in Worren et al. 2002, Ro et al. 2007, Jacobs complexity of the product and the requisite processes
et al. 2007, Lau et al. 2011, and Cabigiosu and that are needed to put the product together. We fash-
Camuffo 2012) and are credited with an increase in ion complexity through an NK-metaphor (Kauffman
product variety and innovation through the swapping 1993, Loch et al. 2003) which is a general and con-
of variant modules, the shortening of product devel- tent-free model for complexity (Frenken 2000). N rep-
opment times due to the decoupling of development resents the number of component parts which can be
tasks, and thus an increase in the frequency of pro- put together via K distinct manufacturing processes.
duct introduction, and efficient and timely product The higher the number of component parts and pro-
launch due to an ability to tolerate a higher risk of cesses, the higher the number of interrelationships/
design changes (avoiding the need for entire product connections that must be recognized and resolved in
changes), and due to the rather accommodating pro- product and process design. Furthermore, more inter-
cess architecture that can be readily adapted to a new dependence is engendered between product and pro-
product. These effects however may be tempered by cess design decisions as the number of parts and
the complexity (Simon 1962) of both the product and processes increases, further exacerbating complexity.
the respective processes tasked to manufacture it. Thus we posit that, relatively, product modularity
Consider for instance a company that produces and process modularity will be more efficacious
toaster ovens and another company that produces in environments characterized by high levels of
cars. Relative to cars, toaster ovens are fairly simple complexity (Ethiraj et al. 2008). Because complex

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 3

environments are characterized by a high number of the effects of their interactions with complexity on
component parts in the bill of materials and a high new product introduction performance.
number of distinct manufacturing processes, they This study is organized as follows. After defining
engender more interdependencies and thus present key research constructs, the theoretical model and
more opportunities for improvement as opposed to hypotheses are introduced. The methodology is then
an environment with few parts in the product and delineated, followed by the results. Next, the findings
process tally. As Ethiraj and Levinthal (2004) note, are provided and their managerial implications are
designing large-scale interacting systems has been explored. Finally, directions for future research are
found to be NP complete; modularization offers the identified.
potential to decompose or partition the problem, mak-
ing it easier to obtain a “good” solution. The opera-
tional load or effort to manage the introduction of a
2. Definitions
simple system (Choi and Krause 2006) would be triv- The research model is shown in Figure 1. It depicts
ial even without product or process modularization. two explanatory variables (i.e., product modularity
On the other hand, modularity offers great potential and process modularity) and their main effects on
to reduce the operational load when the system is new product introduction performance, while speci-
complex and thus more demanding. fying product/process complexity as the moderator
In summary, the research empirically examines (via interaction effects). Model variables are defined
relationships among product modularity, process below.
modularity, and new product introduction perfor-
mance and whether a combined product and process 2.1. Product and Process Modularity (Explanatory
complexity measure moderates these relationships. Variables)
The population of manufacturers targeted for study Modularity (whether product or process) refers to a
using survey methodology includes both companies hierarchical system structure comprised of smaller
that produce technically simple products and techni- subsystems that can be designed independently and
cally complex products. All study measures are per- yet function as a holistic system (Baldwin and Clark
ceptual in nature except for complexity which is an 1997, Ulrich 1995, Ulrich and Eppinger 2004). Lan-
objective measure. Research hypotheses are tested glois (2002) suggests that modules are discrete chunks
using hierarchical regression models that consider the that communicate with each other via standardized
main effects of product and process modularity and interfaces or connections. Product modularity corre-

Figure 1 Conceptual Research Model

Product
Modularity

H1

H3

New Product
Complexity Introduction
Performance

H3

H2
Process
Controlled for Firm Size, Industry, Innovation Magnitude
Modularity
(incremental<>radical), & Quality (Reliability).

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
4 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

sponds to “modularity-in-design” and process modu- reconfigured with little loss of function. Fine et al.
larity corresponds to “modularity-in-production,” a (2005) note that process modules comprise standard-
distinction articulated by Kamrad et al. (2013). Extant ized groups having few strong ties, enabling rapid
studies frequently fail to differentiate product modu- decoupling and resequencing of processes and tool-
larity from related and/or complementary concepts ing because each process module is relatively autono-
such as parts commonality, process modularity, mous. If each production operation is relatively
dynamic teaming (Jacobs et al. 2007, Tu et al. 2004). independent of the prior and/or subsequent opera-
The literature identifies three core modularity tion, the process can be viewed as modular (Feitzin-
aspects, each of which is discussed below: functional ger and Lee 1997, Voordijk et al. 2006). Thus, the
binding, interface standardization, and decompos- process module can be activated or de-activated
ability (Parker 2010). We conceptualize and subse- quickly as demand or product specifications change
quently measure product and process modularity as new products come on line.
encompassing these three core aspects. Functional
binding (also called functional mapping or module 2.2. New Product Introduction Performance
partitioning) is the compartmentalization of the (Dependent Variable)
designed functions of a system (Baldwin and Clark Many aspects of innovation performance are
1997, Fixson 2005, Henderson and Clark 1990, Sal- described in extant literature. These range from the
vador 2007, Ulrich 1995, Ulrich and Eppinger 2004). ability of a company to incrementally innovate (minor
The central tenet is the one-to-one mapping of func- design changes) to its ability to radically innovate
tions to components (or sub-processes for process (new-to-the-world products) (Calantone et al. 2010,
modularity) which allows functions to be added (or Chao and Kavadias 2008). Measures related to inno-
subtracted) from a product (or process) by adding (or vation performance include innovation capacity
deleting) components (or sub-processes). Functional (Aydin et al. 2007, Browne et al. 1998), product
binding serves to reduce the interdependence of com- design capability (Aydin et al. 2007, Li 2000), new
ponents, enabling their autonomous design. product development cycle time (Lynn et al. 1999),
Interface standardization refers to the common, rapid new product introduction, product innovation
agreed upon mechanisms for interaction among com- (Eisenhardt and Tabrizi 1995, Lukas and Ferrell 2000)
plementary product or process components of a sys- and improvement of existing products (Droge et al.
tem. These interfaces endure over time, have widely 2000, 2004, Griffin 2002). In this study, we focus on a
known specifications across stakeholders (often few crucial and interrelated measures of new product
referred to as the “visible” rules of design, Baldwin introduction performance which include develop-
and Clark (1997)), and provide for the open inter- ment lead time (Bayus 1997, Cohen et al. 1996, Droge
change of complementary system components. It is et al. 2004, Tatikonda and Montoya-Weiss 2001),
recognized as a key aspect of modularity (Baldwin timeliness (i.e., on-time launch) (Cooper and Klein-
and Clark 1997, Rosenberg 1982, Sanchez and Maho- schmidt 1994, Eisenhardt and Tabrizi 1995), product
ney 1996, Ulrich 1995, Ulrich and Eppinger 2004). San- innovation (ability to introduce new products) (Eisen-
chez and Mahoney (1996) emphasize that hardt and Tabrizi 1995, Koufteros et al. 2007, Roberts
standardized interfaces actually embed coordination 1999), and frequency of new product introduction
into the product development processes. (Smith et al. 2005). We use an aggregate/summated
Decomposability captures how easily a system can be measure of new product introduction performance
separated into its various components, making com- because the main interest was on the impact of modu-
ponent swapping practical (Ahmad et al. 2010, Anto- larity on new product introduction performance over-
nio et al. 2007, Lin 2003, Tu et al. 2004, Worren et al. all. Such performance is key to reaping first-mover
2002). Since decomposability makes component advantages, higher profitability due to premium pric-
swapping practical (Brabazon et al. 2010, Schilling ing (Koufteros et al. 2002), and organic growth (e.g.,
2000), it is this facet of modularity that is most often Droge et al. 2000).
associated with its flexibility advantage. Decompos-
ability allows a system to be readily reconfigured 2.3. Complexity: The Proposed Moderator
using the same, similar, or supplementary compo- Complexity describes the presence of many different
nents without deleterious effects on performance. parts, elements or patterns, connected in ways that
While the literature on process modularity is less make an object or process difficult to fully under-
developed than the product modularity literature stand. The number of components of a system (pro-
(Jacobs et al. 2011), these three core modularity duct or process) is a primary contributor to
aspects apply conceptually to both kinds of modular- complexity (Kauffman 1995, Simon 1962) and com-
ity. Schilling and Steensma (2001) indicate that pro- plex products have the intrinsic advantage of being
cesses are modular when process components can be more difficult to imitate. However, as complexity and

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 5

the inherent amount of information increases, it can management (Ulrich and Eppinger 2004). Ethiraj et al.
approach or surpass the decision maker’s processing (2008) highlight the two mechanisms by which pro-
limits. This consequence of the increasing complexity duct modularity accelerates innovation: autonomous
of products and processes is often explained using the (within component) innovation and modular (mix
concept of bounded rationality (Simon 1957); that is, and match of modules) innovation (see also Baldwin
there are limits to the capabilities of decision makers and Clark 1997, 2000, Ulrich 1995, Ulrich and Eppin-
to process information, their ability to accurately per- ger 2004). A modular product system furnishes possi-
ceive a system is hindered as the number of compo- bilities for autonomous and concurrent innovation
nents increases, and their behavioral response to driven by the division of labor and/or tasks and rapid
receiving additional information is markedly less trial-and-error learning (Langlois and Robertson 1992,
rational (Parker 2010). Ulrich 1995, Ulrich and Eppinger 2004). Product mod-
Complexity is often described as the number of ele- ularity also grants companies the ability to open their
ments in a given system (Flynn and Flynn 1999, Sawh- technology to an external network of firms, stimulat-
ney and Piper 2002) where it is assumed that a higher ing the acquisition and integration of new knowledge
count of elements is associated with higher levels of and engendering faster technological progress (Lan-
complexity; a higher count of elements also implies a glois and Robertson 1992, Ro et al. 2007, Ulk€€ u and
higher number of possible interdependencies, an arte- Schmidt 2011). Thus, diverse suppliers can work on
fact of complexity. We operationalize complexity by their modules in parallel, which are then integrated
quantifying (1) the number of items in the bill of into the firm’s platform. Embedding coordination into
materials (BOM) for a typical finished product, and the standardized interfaces can reduce the need for
(2) the number of distinct manufacturing processes (human) managerial coordination across internal or
required to produce the product. Fashioning com- external units, thereby reducing the intensity of the
plexity using an NK-type metaphor (Kauffman 1993), product development task (Pil and Cohen 2006). This
we posit that N represents the number of component can reduce time-to-market, secure on-time launch,
parts which can be put together via K distinct manu- and allow simultaneous pursuit of a larger number
facturing processes. A variety of different products and/or a greater variety of new projects (Sanchez and
can be produced by a variety of different combina- Mahoney 1996).
tions of distinct manufacturing processes, creating a In addition, product modularity generates options
sizable number of options and interdependencies for product variations from feasible combinations of
amongst product and process components. We posit modular components; this is what Kogut and Zander
that a high number of items in the BOM coupled with (1992) called “combinative capabilities” in new pro-
a high number of distinct manufacturing processes duct creation (Sanchez and Mahoney 1996). Thus, a
signify higher levels of complexity and create a number of new product variations can be developed
“rugged landscape” (Mihm et al. 2003). Mihm et al.’s simultaneously, possibly with little interaction among
(2003) results suggest that “the overall system design teams who may engage in parallel experimen-
becomes complex and nonlinear as it grows in size, tation and development (Baldwin and Clark 1997,
even if all components and their interactions are very 2000, Loch et al. 2001). This increases the number of
simple” (p. 747). new products; and although the variations may not
be radically different, they may appeal to different
target markets. Concurrently, it can help the organiza-
3. Hypotheses Development tion meet target launch dates since a variety of osten-
3.1. The Main Effects of Product Modularity and sibly different products can be handled efficiently
Process Modularity and effectively. Thus:
Most firms seek ways to enhance new product intro-
duction performance, a fact widely recognized in the H1. Product modularity positively affects new product
innovation literature (Calantone et al. 2010). This introduction performance.
research focuses on modularity, both product and
process, and specifies them as variables that can Similar arguments can be advanced regarding the
explain variation in new product introduction perfor- effect of process modularity on new product intro-
mance (Sanchez and Mahoney 1996, Staudenmayer duction performance. Process modularity embodies
et al. 2005, Stremersch et al. 2003, Ulrich and Eppin- the notion of a dynamic network of relatively autono-
ger 2004). mous operating units (or production modules), some
Scholars suggest that product modularity will posi- of which may belong to the firm’s network partners.
tively impact new product introduction performance, Such a network can easily be reconfigured to support
primarily because of modularity’s effects on product and accelerate the launch of a new product. Once the
change, product variety, and product development “visible” design rules (i.e., the interfaces defining

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
6 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

how the separate production entities will synchronize can all be modularized, process demands may vary
or harmonize their activities) are settled, these pro- significantly.
duction modules are completely in control of their For instance, manufacturing a specific Open Runway
manufacturing tasks and the design of their pro- handbag may first require the selection of a base
cesses, potentially allowing parallel experimentation model followed by a choice of material and individual
that might otherwise not occur. A firm can easily components such as trims, zippers, and straps. The
assemble, mobilize, and integrate the requisite pro- process does not necessarily change or deviate across
cess modules (Sirmon et al., 2007) to produce a new “new” products and thus highly modularizing the
product due to the functional binding, interface stan- process may not enhance new product introduction
dardization, and decomposability characteristics performance substantively. On the other hand, pro-
which are native to modular processes. Organizations ducing a Kuka robot may call for modular processes
may deploy a variety of approaches to gain process because they can significantly reduce the ramp-up
modularity; the key is reconfigurability. For instance, time to introduce and manufacture “new” robots.
the use of flexible and general purpose equipment, Kuka uses six basic types of robots which can be min-
mobile robotic equipment, and/or cellular configura- gled with a large number of mechanical and electrical
tions can engender higher levels of process modular- feeding and loading options to generate custom tai-
ity because they can promote high levels of functional lored systems that meet specific customer needs. Pro-
binding (lower levels of interdependence), interface ducing a standard loading/unloading robot as
standardization (open and tolerant interchange), and compared to a shelf-mounted robot or a press-to-
decomposability (swapping of processes is relatively press robot or a gantry robot with different features
painless). A greater variety of products can be intro- and accessories demands a level of process flexibility
duced as the process can accommodate them while that enables the current product assortment to be pro-
the launch time can be settled more accurately. The cessed efficiently and new products to be added with
autonomous process modules would facilitate con- minor process recombinations. In this way, product
current design of processes (Pil and Cohen 2006), fur- development lead times are reduced and the capabil-
ther reducing time-to-market. Thus: ity to innovate more frequently is enhanced. While all
the loading/unloading robots may belong to the same
H2. Process modularity positively affects new product product family, different combinations of sub-pro-
introduction performance. cesses may be needed to introduce and assemble each
variant. A company with process modularization can
It should be noted that few empirical studies have add or remove sub-processes without the need to
examined the consequences of modularity on new change other sub-processes.
product development performance (as opposed to Consider also the total effect of product modularity
firm performance) and even fewer have examined on new product introduction performance. If product
concurrently the explanatory efficacy of both product modularity implies:
and process modularity. Furthermore, results have
(1) the rationalization of components (i.e., fewer
been inconclusive: no effects were found for process
components, and fewer or clearer interdepen-
modularity in some studies (Jacobs et al. 2011) while
dencies among components), thus clarifying con-
results for product modularity were mixed (Ahmad
figuration possibilities and speeding new
et al. 2010, Antonio et al. 2007, Jacobs et al. 2011, Lin
product design, development, and launch;
2003, Worren et al. 2002).
(2) a clearer link between product configuration and
€ u and
functionality due to decomposability (Ulk€
3.2. The Moderating Role of Complexity
Schmidt 2011), thus increasing the timely poten-
Does complexity moderate the effects of modularity
tial for autonomous, concurrent and supply
expressed as H1 and H2 and represented in Figure 1?
chain sourced innovations; and
In other words, are the hypothesized relationships
(3) an increase in adaptability, thus enabling rapid
equally salient across complexity levels? The com-
response in new product development; and
plexity of a Nike shoe or of an Open Runway handbag
operations. . .
is less than the complexity of an eight-axis Kuka robot
used in the automotive industry. The large number of then a firm that succeeds at the relatively more diffi-
components embedded in a robot dictates that they cult task of modularizing a complex product (vis-a-vis
be connected. As the number of components a simple product) should derive stronger effects on
increases, more modules may be used and thus more speed/timeliness/innovation/frequency and overall
interfaces become necessary; furthermore, more con- new product introduction performance. For example,
nections between components within each module the marginal impact of the rationalization of compo-
are needed. Also, while shoes, handbags, and robots nents for a simple product (e.g., shoes, handbag) that

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 7

already has few components and few connections may produce simple parts/products. Second, we
should be lower than the marginal impact of the targeted firms with an executive affiliation with the
rationalization for a complex product (e.g., robot) university supporting the study, and are known to
with numerous components, processes, and connec- produce complex products. Again, it is possible
tions. Similarly, the new product introduction capa- that some of these firms may produce simple
bilities of the firm and the supply chain should parts/products as well. For each prospective firm,
receive a greater boost when the difficult task of mod- a subjective judgment was made regarding contact
ularizing a complex product is accomplished because inclusion based on characteristics of their firm’s
the configuration-functionality link was less clear for primary products as identified by information
complex products to begin with, and thus the marginal available via the firms’ websites. Third, we sam-
potential for innovative insights due to decomposabil- pled firms in NAICS industry classifications – 334
ity is greater for complex products. In other words, (Computer and Electronic Product Manufacturing)
the marginal return on new product introduction per- and 336 (Transportation Equipment Manufactur-
formance of successfully modularizing a complex ing). We have no reason to believe that these firms
product should be greater than that of modularizing a produce primarily simple or complex parts/prod-
simple product. ucts. For the latter, electronic mailing lists from the
Similar logic can be used to argue that the task of Society of Manufacturing Engineers and the Institute
modularizing a complex process should lead to stron- of Industrial Engineers were used.
ger positive effects on new product introduction per- The survey (which included other aspects not
formance than modularizing a more simple process examined here and was split between distinct man-
since it would clarify previously, possible murky agerial and technical parts) was administered to
reconfiguration possibilities, identify opportunities participants online. Questions were randomized
for concurrent processing, and increase the overall within categories to reduce common method bias
adaptability of the system to product changes. These and yea-saying behaviors. For the first two cate-
effects should enable a company to more rapidly and gories of contacts described above, e-mail invita-
more frequently introduce new products while meet- tions to participate were sent directly to most of
ing introduction due dates. Thus: the prospective participants. But for a small portion
of them, invitation letters were sent by U.S. mail,
H3. Complexity moderates model relationships (i.e., H1, and for a smaller portion, invitations were sent via
H2) such that the effects are more salient for complex e-mail to Tier 1 suppliers directly from the focal
products than for simple products. firm. These Tier 1 supplier contacts received invita-
tion letters which contained logos and authorized
signatures of the focal firm. For the third category,
contact information for invitees was not disclosed
4. Research Methodology directly to the research team. Instead, all communi-
4.1. Research Design and Sample Characteristics cation to invitees was conducted via email through
The unit of analysis is a manufacturing unit repre- a third party service provider hired by the respec-
sented by a firm or a Strategic Business Unit tive professional society supplying the list of poten-
(SBU).1 The population of firms targeted includes tial respondents.
both companies that produce technically simple All participants were directed to the same Web-
products and technically complex products, that is, based survey. Two reminder emails were sent, spaced
products composed of a large number of compo- 3–5 business days apart. The 13.5% response rate is
nents that interact in non-simple ways (Simon comparable to similar studies (e.g., Das and Narasim-
1962) and have a large proportion of moving parts han 2000, Narasimhan et al. 2001). Non-response bias
(Novak and Eppinger 2001). Responses were was evaluated using a method similar to Armstrong
sought using a three-pronged category strategy. and Overton (1977). There were no apparent differ-
First, we targeted six focal firms in industries ences between early and late respondents in terms of
known for their complex products such as aircraft, demographics or mean scores based on t-tests. Out of
automobiles, large household appliances and heavy 169 responses, 112 are used for data analyses pur-
commercial equipment (Mihm et al. 2003). Three of poses as 57 responses had substantive missing values
those companies agreed to participate in the study on some constructs of our interest. Sample descriptors
and provided contact information for select Tier 1 appear in Table 1 (e.g., sales, employees).
suppliers. We assumed that companies participat-
ing in these industries would be more prone to 4.2. Survey Instrument and Measurement
produce complex products. Nevertheless, it is pos- A panel of four manufacturing professionals and three
sible that some firms within the specific industries academics pretested an initial research questionnaire

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
8 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

Table 1 Description of the Sample (n = 112) demand a high number of distinct manufacturing
Industry Frequency Percentage
processes can be labeled as more complex. Complex-
ity was also measured perceptually in our study using
1. Continuous Processes & Plastics 11 9.8
Manufacturing
items from the literature that captured both product
2. Primary & Fabricated Metal Manufacturing 28 25.0 and process complexity (see Table 2): our purpose
3. Machinery, Computer, & Electrical 35 31.3 was to correlate the NK measure of complexity with
Component Manufacturing this perceptual measure to underscore its credibility.
4. Transportation Equipment & 26 23.2 Innovation magnitude, a construct used as a control
Aerospace Manufacturing
5. Miscellaneous Manufacturing 12 10.7
variable, captures the degree of innovation, and was
measured using perceptual indicators (see Table 2).
Respondent title Frequency Percentage Another control variable, product quality (reliability)
President/CEO 3 2.7 was measured by a single item on a 7-point scale
Vice president/COO 9 8.0 where 1 = worst industry and 7 = best in industry.
Director 11 9.8
Manager 38 33.9 4.3. Research Methodology
Engineer 26 23.2
Other 25 22.3
We first evaluate the measurement models and
common method bias. Using LISREL 8.51, each
Firm descriptors Mean Median measurement model (i.e., exogenous variables and
Sales $8.80 billion $80 million endogenous variable, and control variables) was
Employees 7747 350 assessed via CFA (Anderson and Gerbing 1988), by
Market share (%) 31.38 25.00 specifying Maximum Likelihood (ML) estimation
with a covariance matrix as input. The fit of the mea-
surement models was evaluated using a variety of fit
to ensure face validity. Their feedback resulted in indices including v2/df, CFI, IFI, NNFI, SRMR,
several refinements to the survey instrument. RMSEA and the p-value for Test of Close Fit
Table 2 lists the constructs and their respective indi- (RMSEA < 0.05). Convergent validity was examined
cators, the types of scales; item-to-factor loadings by probing the statistical and substantive significance
based on confirmatory factor analysis (CFA), the of individual item CFA loadings. Discriminant valid-
means and the composite reliabilities, and measure- ity was appraised by comparing the average vari-
ment model fit indices. Responses were primarily ance extracted (AVE) to the squared correlation
obtained from engineers, managers, directors, and between constructs (Fornell and Larcker 1981). Com-
other executives affiliated with engineering or man- posite reliability and AVE were used to assess
ufacturing; these individuals should have intimate construct reliability.
knowledge pertaining to the variables of interest. To test the hypotheses, we used hierarchical
The indicators for product or process modularity regression models via SPSS 22.0. In Model 1, we
shown in Table 2 were selected to ensure that the core entered several control variables which may par-
aspects of modularity (functional binding, interface tially explain variability in new product introduction
standardization, and decomposability) were properly performance. Control variables include (1) firm size,
represented and thus cover the constructs’ respective because larger firms possess more resources which
content domains. The new product introduction per- can be committed towards product and process
formance indicators are measured relative to industry development activities; (2) industry affiliation,
(i.e., worst to best) and tap into vital aspects of new because some industries have a reputation for high
product introduction performance such as product new product introduction performance including
development lead time, on-time product launch, pro- frequent, on-time, and rapid innovations vis-a-vis
duct innovation, and frequency of new product intro- other industries; (3) quality (measured via product
duction. reliability) as high quality products demand more
Complexity, the moderator, was measured using attention, resources, and effort, and (4) innovation
two measures which were subsequently combined in magnitude, tapping whether the overall product
a multiplicative fashion (i.e., NK). The first measure N innovation strategy is incremental vs. radical. The
represents the number of items in the bill of materials latter is important because differences in new pro-
(BOM) for a typical finished product (median = 44.5) duct introduction performance could be attributed
at each participating firm/SBU. The second measure to the magnitude of innovation instead of the focal
K captures the number of distinct internal manufac- modularity variables. Model 2 adds the variables of
turing processes used in the production of a typical interest (product modularity, process modularity,
major product (median = 6.5); this measure reflects and complexity) as main effects while Model 3
complexity and intricacy where products that further includes the hypothesized interactions (i.e.,

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 9

Table 2 CFA Measurement Model with Description of Constructs and Item Analysis

Standard loading
Construct (t-value; All sig. at <0.01)
Perceived Product Modularity:
Descriptives: Mean/item = 3.90; SD/item = 1.98, CR = 0.90 (composite reliability from CFA)
Literature sources: Duray et al. (2000); Worren et al. (2002); Lin (2003); Tu et al. (2004); Parker (2010)
Scale: 1 = “Strongly disagree” to 7 = “Strongly agree”
PdM1 For our products, we can make changes in components without changing other components 0.64
PdM2 For our products, functions can be directly added or deleted by adding or removing components 0.75 (6.97)
PdM3 Our products have interchangeable features and options 0.69 (6.48)
PdM4 Our products are designed to enable the swapping of components 0.75 (6.93)
PdM5 Our product components have standardized interfaces 0.50 (4.81)
PdM6 The interfaces of our product components are designed to accept a variety of other components 0.86 (7.68)
PdM7 Our product components are able to accept a wide range of complements (such as modules or peripherals) 0.81 (7.37)
PdM8 Our products can be decomposed into separate modules 0.62 (5.81)
PdM9 Our products are designed to be easily reconfigured 0.76 (7.06)
Perceived Process Modularity:
Descriptives: Mean/item = 4.52; SD/item = 1.76, CR = 0.86 (composite reliability from CFA)
Literature sources: Duray et al. (2000); Lin (2003); Tu et al. (2004); Parker (2010).
Scale: 1 = “Strongly disagree” to 7 = “Strongly agree”
PcM1 Our production processes are composed of process modules (or self-contained processing units) 0.70
PcM2 Our production processes are broken down into sub-processes that can operate independently 0.63 (7.88)
PcM3 Sub-processes can be added to or removed from our production processes without changing other 0.88 (8.64)
sub-processes
PcM4 Our production processes are designed so that sub-processes can be added or removed 0.95 (8.92)
without significant disruptions/changes to other sub-processes
PcM5 Our production processes are designed to be rapidly disassembled and reconfigured 0.50 (5.09)
PcM6 Only minor equipment modifications are required to produce different products 0.52 (5.24)
Perceived New Product Introduction Performance:
Descriptives: Mean/item = 5.00; SD/item = 1.29, CR = 0.85 (composite reliability from CFA)
Literature sources: Aydin et al. (2007); Droge et al. (2000, 2004).
Scale: 1 = “Worst in industry” to 7 = “Best in industry”
NPI1 Product development lead times 0.81
NPI2 On-time product launch 0.74 (7.82)
NPI3 Product innovation 0.71 (7.53)
NPI4 Frequency of new product introductions 0.81 (8.56)
CFA Fit Statistics: v2 = 181.02, df = 144; v2/df = 1.26; CFI = 0.95; NNFI = 0.94; IFI = 0.95; SRMR = 0.07;
RMSEA = 0.05, p-value for Test of Close Fit (RMSEA < 0.05) = 0.54, n = 112
Perceived Innovation Magnitude:
Descriptives: Mean/item = 3.89; SD/Item = 1.38, CR = 0.88 (composite reliability from CFA)
Literature sources: Ettlie et al. (1984); Dewar and Dutton (1986); Sorescu et al. (2003);
Scale: 1 = “Strongly disagree” to 7 = “Strongly agree”
InM1 Our innovations involve fundamentally new concepts or principles for our SBU 0.68
InM2 Our innovations frequently require our SBU to develop many new skills 0.81 (7.42)
InM3 Our innovations require our SBU to learn from completely new knowledge bases 0.78 (7.23)
InM4 Our innovations require skills that our SBU did not previously possess 0.65 (6.13)
InM5 Our innovations require our SBU to carry out a great deal of retraining 0.82 (7.54)
InM6 Our innovations require our SBU to adopt different methods or procedures 0.67 (6.35)
CFA Fit Statistics: v2 = 15.06, df = 9; v2/df = 1.67; CFI = 0.98; NNFI = 0.96; IFI = 0.98; SRMR = 0.03;
RMSEA = 0.08, p-value for Test of Close Fit (RMSEA < 0.05) = 0.22, n = 112
Perceptual Measure of Complexity:
Descriptives: Mean/item = 4.35; SD/item = 1.71, CR = 0.91 (composite reliability from CFA)
Literature sources: Simon (1962); Langlois (2002); Salvador (2007).
Scale: 1 = “Strongly disagree” to 7 = “Strongly agree”
Compl1 Our products are highly complex .89
Compl2 The secondary functions that combine (or interact) to produce our products’ primary functions are 0.51 (5.21)
quite different from one another
Compl3 Our products are very sophisticated in terms of structure 0.90 (14.04)
Compl4 Our products are very sophisticated in terms of function 0.90 (13.91)
Compl5 I would characterize the structure of our typical finished products as: {1 = very simple (i.e., 0.76 (10.21)
few connections among components); 7=very intricate (i.e., many connections among components)}

(continued)

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
10 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

Table 2 Continued

Standard loading
Construct (t-value; All sig. at <0.01)
Compl6 In a given product, the number of secondary functions that combine (or interact) to produce the 0.59 (7.02)
product’s primary function is: (1 = very low (like in a toaster); 7 = very high like in an airplane)
Compl7 Our production system is very complex 0.79 (10.92)
Compl8 Our production system is very sophisticated in terms of structure/layout 0.73 (9.46)
Compl9 Our production system is very sophisticated in terms of operation 0.71 (9.04)
CFA Fit Statistics: v2 = 65.40, df = 24, v2/df = 2.72; CFI = 0.94; IFI = 0.95; SRMR = 0.05; RMSEA = 0.12,
p-value for Test of Close Fit (RMSEA < 0.05) = 0.00, n = 112

product modularity 9 complexity and process modu- (v2 = 615.43, df = 147, v2/df = 4.19, CFI = 0.63,
larity 9 complexity). Interaction plots were employed IFI = 0.63, NNFI = 0.56, SRMR = 0.15, RMSEA = 0.17,
to probe the interaction effects in more depth. p-value for Test of Close Fit [RMSEA < 0.05] = 0.00).
The explanatory variables were centered by sub- A marker variable technique, which is strongly
tracting the mean from each variable entry in order advocated as an effective approach to assess CMB
to mitigate multicollinearity that may emerge from (Malhotra et al. 2006), was also used. According to
applying interaction terms (Frazier et al. 2004, Craighead et al. (2011), this technique calls for the
Hayes 2013). Multicollinerarity was then evaluated inclusion of additional variables which are theoreti-
to assure that the findings can be meaningfully cally unrelated to at least one variable of interest (Har-
interpreted. The highest VIF value was 1.715, the rison et al. 1996). Four marker variables were
lowest tolerance was 0.583, and the highest condi- included and scored on 7-point scales, where
tion number was 16.480, suggesting that multi- 1 = strongly disagree and 7 = strongly agree. Their
collinearity does not present a significant challenge. correlations with new product introduction perfor-
We also examined the residuals from our estimated mance were rather low: (1) Firms in our industry
model in order to identify potential outliers. The advertise heavily compared to other industries
lowest and highest standardized residuals were (r = 0.073), (2) In our industry, firms are at the mercy
2.393 and 2.656, respectively, while the lowest of powerful suppliers (r = 0.017), (3) Large firms have
and highest studentized deleted residuals were definite cost advantages in our industry (r = 0.020),
2.659 and 2.850, respectively. There were no error and (4) We have a serious excess capacity problem in
terms suggesting the presence of outliers. To evalu- our industry (r = 0.015). The low correlations indi-
ate whether the data set includes influential obser- cate that CMB does not present a serious threat.
vations, we examined Cook’s distance which In terms of separation remedies (Craighead et al.
produced a mean of 0.011 and a maximum value 2011), we relied on methodological and time separa-
of 0.137, both significantly less than 1.00. The tion. To minimize CMB, items measuring the depen-
Mahanalobis distance with an average of 11.893 dent construct were interspersed with items
was not statistically significant (p > 0.45). This sug- measuring other variables not associated with this
gests that there were no influential observations. specific study, so presumptive conclusions would not
Multivariate normality was examined via a P-P consistently influence ratings. The independent vari-
plot. The plot was essentially linear on the 45° line, ables, the moderator variable, and the dependent
suggesting multivariate normality. Scatterplots of variable were measured using different scales/
predicted values and residuals suggested anchors (see Table 2). Furthermore, new product
homoscedasticity. introduction performance was measured at two dif-
ferent times, administered 3–5 days apart, using two
4.4. Common Method Bias and Triangulation different distinct parts of the survey (managerial and
For questionnaire-based surveys completed by a sin- technical, respectively); test-retest reliability was 0.90
gle informant, a frequent concern is common methods based on 83 responses. Also, the mean scores across
bias (CMB). As far as CMB is concerned, our research the two administrations do not differ (p > 0.79). This
design included a variety of statistical and separation affords time separation and attests to the integrity of
approaches to minimize risk (Craighead et al. 2011). responses. In addition, we obtained responses per-
First, we used a CFA approach to Harman’s one-fac- taining to the firm’s new product introduction perfor-
tor test (McFarlin and Sweeney 1992), whose rationale mance from two different respondents at 22 firms
is that if CMB exists, a single latent factor will account There is no evidence to suggest that the mean scores
for all manifest variables (Podsakoff and Organ 1986). differ (p > 0.36). Furthermore, the correlation between
Harman’s test failed to support a single factor the two responses was over 0.76 (p < 0.001).

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 11

For further testing of CMB, we estimated correla- were substantive in magnitude (only three items had
tions amongst the focal variables across industries. If coefficients below 0.60 for the constructs used in our
CMB is alleged and “yeah” saying behavior is preva- hierarchical regression models) thus supporting con-
lent, then the correlations would be similar across vergent validity. Potential cross-loadings were exam-
industries. We produce here a sample of those corre- ined via an assessment of completely standardized
lations as evidence demonstrating significant differ- expected change in loadings (kx). None had a change
ences across industries. For instance, the correlations in kx greater than 0.30, suggesting that items load sub-
(r) between innovation magnitude and new product stantively only on their intended latent variables, thus
introduction performance are 0.662, 0.549, 0.289, supporting item-level discriminant validity (Koufteros
0.267, and 0.710 across the five industry groups. Simi- 1999). The composite reliabilities were acceptable at
larly, the correlations (r) between product modularity 0.85 or better levels. AVEs ranged from 0.51 to 0.59
and new product introduction performance are 0.589, (≥ 0.50; Table 3), providing further evidence of con-
0.323, 0.238, 0.258, and 0.611, respectively. Given these struct reliability. The squared correlations were com-
values, we control for industry effects in our analyses. pared with individual AVEs (Table 3): the highest
In order to render more credibility to the measure was 0.31 (0.55 9 0.55), below the smallest AVE. Thus,
of complexity and triangulate its score, we correlated discriminant validity was held at the construct level.
the complexity score with the subjective measure of
complexity in Table 2. The correlation between the 5.2. Model Results and Hypothesis Testing
measure of complexity used here and the subjective Table 4 shows the findings from the hierarchical
measure was highly significant (r = 0.52, p < 0.000). regression models. Model 1 (p < 0.000, R2 adjusted =
0.384) suggests that some of the control variables are
significantly related to new product introduction per-
5. Results formance. These include product quality (reliability)
5.1. Measurement Analysis Results and innovation magnitude (these remain significant
First we evaluated the measurement models starting in Models 2 and 3).
with the model for product and process modularity Model 2 adds the main effects of product modular-
and new product introduction performance. The fit ity, process modularity, and complexity. Model 2 pro-
indices based on CFA results are v2 = 181.02, df = 144, duces an R2 adjusted = 0.396. Model 3 includes the
v2/df = 1.26, CFI = 0.95, NNFI = 0.94, IFI = 0.95, main effects as well as the two interaction terms (i.e.,
SRMR = 0.07, RMSEA = 0.05 and p-value for Test of product modularity 9 complexity and process mod-
Close Fit (RMSEA < 0.05) = 0.54. All indices suggest ularity 9 complexity). R2 adjusted is 0.414. The
an acceptable measurement model (Table 2). A sepa- F change is moderately significant (p < 0.10) indicat-
rate measurement model was also specified for inno- ing that Model 3 explains more of the variance in new
vation magnitude, a variable that serves as a control product introduction performance than Model 2 does.
variable, and for the perceptual measure of complex- Thus Model 3 is discussed below. The observed statis-
ity; both models fit well (see Table 2 for results). tical power for Model 3 was 1.00 based on probability
In all measurement models, all items were loaded level of 0.05 (or 0.10), 12 predictors, observed
significantly on their latent factors (p < 0.001) and R2 = 0.478, and n = 112.

Table 3 Construct Correlation Matrix and Discriminant Validity

Variables 1 2 3 4 5 6 7
1. Product modularity‡ 0.51†
2. Process modularity‡ 0.383** 0.51†
3. Complexity¶ 0.062 0.042* NA
4. New product introduction performance‡ 0.296** 0.218* 0.094* 0.59†
5. Firm size 0.054 0.147 0.119 0.051 NA
6. Innovation magnitude‡ 0.344** 0.198* 0.100 0.400** 0.066 0.55
7. Product quality (reliability)§ 0.117 0.255** 0.091 0.553** 0.072 0.016 NA

Notes: **; *. Correlation is significant at the 0.01 or 0.05 level (2-tailed).



Average variance extracted.

See Table 2 for measurement.
§
Measure of the reliability of the product (i.e., Probability of performance without failure over a specified time period), measured on a 7-point scale where
1 = “Worst in industry” to 7 = “Best in industry”.

The measure of complexity is a multiplicative term (No. of items in BOM for a typical finished product 9 No. of distinct internal manufacturing
processes). For data analytic purposes, the items in the BOM are grouped as follows 1: less than 10, 2: 10–99, 3: 100–499, 4: 500–999, 5: 1000–4999,
6: 5000–9999, and 7: Greater than 10,000. Similarly, the distinct internal manufacturing processes are grouped as follows: 1: 1, 2: 2 or 3, 3: 4 or 5, 4: 6
or 7, 5: 8 or 9, 6: 10 or 11, and 7: 12 or more.

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
12 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

Table 4 Hierarchical Regression Analyses

Model 1 Model 2 Model 3

Independent variables Coefficients SE Coefficients SE Coefficients SE


Constant 9.8730*** 2.1440 10.0456*** 2.1292 9.8222*** 2.1024
Control variables
Industry 1 1.0397 1.1910 1.1047 1.1991 1.3258 1.1927
Industry 3 0.8127 0.8477 0.9605 0.8633 1.0099 0.8648
Industry 4 0.7459 0.9195 0.6185 0.9462 0.5817 0.9332
Industry 5 0.4031 1.2008 0.5475 1.2001 0.9113 1.1932
Firm size 0.0164 0.1709 0.0304 0.1742 0.0138 0.1728
Innovation magnitude 0.2085*** 0.0500 0.1734*** 0.0535 0.2010*** 0.0541
Product quality (reliability) 1.8532*** 0.2932 1.7776*** 0.3021 1.8325*** 0.2989
Main effects
Product modularity 0.7376** 0.3576 0.8816*** 0.3669
Process modularity 0.0132 0.3615 0.0992 0.3596
Complexity 0.0202 0.0304 0.0052 0.0306
Interaction effects
Product modularity 9 complexity 0.0004** 0.0002
Process modularity 9 complexity 0.0069** 0.0039
Model fit
R//R2 0.650//0.423 0.671//0.451 0.691//0.478
Adjusted R2 0.384 0.396 0.414
F (Significance) 10.896 (0.000) 8.281 (0.000) 7.545 (0.000)
F change (Prob > F change) Model 2- Model 1 Model 3- Model 2
1.680 (0.176) 2.573(0.081)

Notes: *p < 0.10, **p < 0.05, ***p < 0.01, Bold values represent statistically significant regression coefficients.

In terms of control variables, product quality (relia- Figure 2 Product Modularity and New Product Introduction Perfor-
bility) and innovation magnitude were significant in mance: Their Relationship at Low vs. High Complexity
Model 3. Higher levels of innovation magnitude (i.e., 13
more radical as opposed to incremental) were related
New Product Introduction Performance

to better new product introduction performance. Sim- 12


ilarly, higher levels of product quality (reliability)
were associated with higher levels of new product 11
introduction performance. Firm size (as in Jacobs Low Complexity

et al. 2011) and industry do not impact new product 10 High Complexity

introduction performance in our study.


The results show support for hypothesis H1 which 9

proposed that product modularity positively impacts


8
new product introduction performance (0.8816; Low Product Modularity High Product Modularity
p < 0.01). The analysis does not support H2 (0.0992;
p > 0.10); the results suggest that process modularity
is not related to new product introduction perfor-
mance. However, both interaction effects were signifi- Figure 3 Process Modularity and New Product Introduction Perfor-
cant but the coefficient for process modularity 9 mance: Their Relationship at Low vs. High Complexity
complexity was positive (0.0069, p < 0.05) while the 13
one for product modularity 9 complexity was nega-
New Product Introduction Performance

tive ( 0.0004; p < 0.05). Thus, the effects of product 12


modularity and process modularity on new product
introduction performance must be viewed in light of 11
the level of complexity. Plots of these interactions are Low Complexity

provided in Figures 2 and 3. 10 High Complexity


Figure 2 displays a disordinal interaction between
product modularity and complexity. The figure illus- 9
trates that when complexity is high the effect of pro-
duct modularization on new product introduction 8
Low Process Modularity High Process Modularity
performance is dampened. On the other hand,

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 13

Figure 3 demonstrates that the effect of process mod- result in dysfunctional perturbations in module- and
ularization on new product development perfor- system-level performance that constrain evolutions
mance is heightened when complexity is high. to inferior designs. The speed and efficiency gains
Overall, the relationships between modularity con- from modularization will be offset by the increased
structs and new product introduction performance time spent in the testing and integration phase,
are contingent on the level of complexity. where the consequences of ignored dependencies
will come to fore” (p. 172). Thus, there might be a
6. Discussion and Directions for Future “sweet spot” for product modularization in a high
complexity environment. Too much of it is counter-
Research productive, while too little is insufficient to partially
This research examined relationships among product address the pressing issue of “design oscillations”
modularity, process modularity, and new product (Mihm et al. 2003). In essence, a system becomes
introduction performance in light of low vs. high rather complex and nonlinear as it grows with size
complexity levels. Both product and process modular- which makes the identification of an optimal configu-
ity have been previously linked to firm performance ration very unlikely. Instead, design solutions “oscil-
and other constructs (e.g., Salvador 2007, Salvador late” through various alternative solutions until
et al. 2002); however, empirical work that simultane- convergence to a solution. The larger the problem
ously examines their individual roles concurrently is size, the more likely that the design diverges to an
scant. Caution is advised in interpreting these find- unreasonable solution and this probability grows
ings in a causal sense since the data are cross-sec- exponentially. As Mihm et al. (2003) note, even when
tional. We begin by summarizing and discussing the there is convergence, the time to conversion grows
findings followed by managerial implications and exponentially as well. While the problem cannot be
measurement issues. entirely eliminated, firms can apply product modu-
larization to mitigate the phenomenon. The key,
6.1. Discussion of Findings and Managerial however, is to not “overdo it”.
Implications Allworth (2013) notes that modularization may
Based on the data analysis, we determined that only work effectively if a focal company and its supplier
product modularity has a statistically significant base are absolutely positive they know how different
main effect on new product introduction perfor- modules would work optimally together; relatively
mance; there were no main effects of process modu- an easier task when a product and its respective man-
larity or complexity. However, both product and ufacturing processes are not all that complex. If that is
process modularity interacted with complexity to accomplished, then all constituents can effectively
affect new product introduction performance: this focus on their piece of the puzzle. But frequently com-
interaction was negative for product modularity and panies and their suppliers try to modularize products
positive for process modularity. This means that without a clear resolution of functional binding and
when complexity is high the marginal return of pro- interface standardization concerns, two critical
duct modularity is diminishing. In contrast, the mar- aspects of modularization. As Kusiak (2002) notes,
ginal return of process modularity is intensified such concerns may surface later on in the develop-
when complexity is high. Caution is advised in inter- ment process. When unanticipated problems regard-
preting the interaction effects since the respective ing interdependencies do emerge, internal and
effect sizes are a full order of magnitude or more external actors need to adjust their work, possibly
smaller than the associated main effects for the two creating new challenges for other modules and
modularity constructs. respective supply chain partners. Small changes by
The negative interaction between product modu- one constituent may have ripple effects across many
larity and complexity can perhaps be explained by members of the internal and external supply chains,
the concept of “over modularity” (Ethiraj and culminating in further delays and compromising new
Levinthal 2004). In essence, when the situation is product introduction performance. This suggests that
rather complex the number of cross-module interde- the adoption of an intensive product modularity
pendencies grows, hampering and disrupting auton- approach for a complex product, with the accompa-
omous innovation at the module level (Ethiraj and nying organizational and supply chain upheaval and
Levinthal 2004). Any gains from parallelism (Mar- the concomitant challenge of effectively integrating
engo et al. 2000) may be offset by the time consumed new organizational units and suppliers, could nega-
to test and integrate the complex system. Ethiraj and tively affect new product introduction performance.
Levinthal (2004) write specifically: “Excessive modu- Collectively, modularization may be fruitful in a com-
larization may blind the designer to potentially plex environment but too much of it might produce
important interactions between decision choices and adverse effects. Thus, we caution against excessive

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
14 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

levels of product modularization when complexity Overall, even though process modularity is clearly
is high. conducive in complex as opposed to simple environ-
In contrast, “overmodularity” is not a concern for ments, more research is due to examine whether
process modularity in a complex environment. there are boundary conditions especially at the
Indeed, when complexity is high, process modularity upper levels of complexity.
had a heightened impact on new product introduc- The measure of new product introduction perfor-
tion performance. This may be due to the fact that, mance that we use here encompasses four pertinent
relatively, there are fewer interactions or interdepen- aspects: product development time, on-time product
dencies arising from the number of manufacturing launch, product innovation, and frequency of new
processes vis-a-vis the number of component parts. product introduction. While the correlations amongst
Based on our sample, the median number of compo- these aspects are substantial (i.e., minimum r ~ 0.59),
nent parts in the BOM is 44.5 (with a mean of 260) it is possible that modularity and complexity may
while the median number of distinct manufacturing have differential effects on these aspects with poten-
processes is 6.5 (with a mean of 7.6). The number of tial theoretical and managerial implications. To exam-
manufacturing processes appears to be relatively low ine these differential effects, we tested separate
to trigger the amount of cross-module process inter- models for each of the four measures of new product
dependencies that would lead to the phenomenon of introduction performance. Appendix A shows results
over modularity in the realm of processes. In addi- for the final model (i.e., Model 3) for each aspect of
tion, we measured process modularity by asking com- new product introduction performance. Plots of
panies to reflect on their own production processes, respective statistically significant interactions appear
which implies that such processes come under the in Appendices B–E.
direct control of the focal company. Since the focal Appendix A suggests that a little over 29% of the
companies have direct control of their own processes, variance in frequency of product introduction can be
they can manage them more effectively because func- explained by the explanatory variables. Product mod-
tional binding, interface standardization, and decom- ularity has a positive main effect (p < 0.05) on the
posability decisions related to processes do not have frequency of product introduction and process modu-
to be negotiated or renegotiated with external parties larity alone interacts with complexity (p < 0.10). The
if challenges do emerge. If and when challenges do interaction (see Appendix B) between process modu-
emerge, they are typically contained within the larity and complexity suggests a heightened effect
boundaries of the organization and thus decisions can when complexity is high; these results are consistent
be made faster, leading to higher levels of new with the findings based on the aggregate measure of
product introduction performance. new product introduction performance (see Table 4).
The advantage of process modularity for new pro- However, since the interaction between product mod-
duct introduction performance is lacking in an envi- ularity and complexity is non-significant, while the
ronment characterized by low complexity (see main effect is positive, the results suggest that pro-
Figure 3). Modularizing the production process duct modularity has a positive impact on frequency
when the environment is already simple may not be of product introduction irrespective of complexity.
warranted. Partitioning or decomposing the produc- For on-time product launch, Appendix A shows
tion system and trying to address interfaces, or the that the explanatory variables explain roughly 24.5%
visible design rules (Baldwin and Clark 1997), for a of the variance. Only product modularity plays a role
simple environment may amount to overkill. Lan- as its main effect (p < 0.05) and its interaction with
glois (2002) states that a “well-decomposed modular complexity (p < 0.10) are the only significant variables
system must pay a kind of fixed cost that an inter- of interest. Appendix C reveals the interaction is dis-
twined system need not pay. . .” (p. 23). Determining ordinal in nature, so only the interaction can be effec-
the appropriate number of modules, the mapping of tively interpreted. The interaction suggests that
design elements to the modules, the interactions product modularity’s effect on on-time launch is
amongst design elements within each module, and diminished when complexity is high. These results
the interfaces between modules is time consuming are in line with the findings for the aggregate measure
(Baldwin and Clark 1997) and may not be all that of new product introduction performance.
productive when the process is rather simple. Bald- For product development lead time (Appendix A),
win and Clark (1997) and Langlois (2002) note that the model explained 27.7% of the variance with only
modular systems are more difficult to design than two variables of interest found to be related to pro-
comparable interconnected systems partly because duct development lead time – product modularity
modular designers have to first gain an in-depth and complexity. Product modularity is statistically
understanding of the inner workings of the overall significant and positive (p < 0.10), and given that its
system to form proper interfaces among modules. interaction with complexity is nonsignificant, higher

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 15

levels of product modularity are associated with embraces process modularity. Note that process
better product development lead time performance modularity will not only facilitate frequency of new
irrespective of complexity level. However, complexity product introduction when complexity is high, but
is negatively associated with product development its positive impact on product innovation in the
lead time, suggesting that higher levels of complexity same environment should counteract the diminished
are associated with adverse effects on the ability of effect of product modularity on product innovation.
the organization to compete based on product devel- Finally, from a theoretical perspective, our findings
opment lead time. suggest that researchers should be careful advancing
Finally, Appendix A illustrates that the model for and testing hypotheses relating modularity to per-
product innovation has the greatest explanatory formance in general as the effects may be subject to
power of the four with 37.7% of the variance complexity levels and the results may vary by the
explained. Both the main effects for product modular- type of indicators deployed to measure the specific
ity (p < 0.10) and process modularity (p < 0.10) are performance construct.
statistically significant as well as their interactions The managerial implications of the overall findings
with complexity (p < 0.01 and p < 0.01, respectively). are straight forward: when product/process complex-
Consistent with prior results involving statistically ity is low, investments in product modularity will
significant interaction terms, the interaction between engender new product introduction performance, but
product modularity and complexity is negative and investments in process modularity are not called for.
the interaction between process modularity and com- For high levels of complexity, product modularity is
plexity is positive. Appendices D and E show these desirable (as long as it is not “overdone”) to mitigate
interaction plots. design oscillations and facilitate convergence to a
In comparing the results of Model 3 for our aggre- design solution (Mihm et al. 2003); however, an
gate measure of new product introduction perfor- investment in product modularization should be
mance in Table 4 with the results for the four accompanied by a concomitant investment in process
indicators comprising this measure, several insights modularity which proved to be rather conducive
arise. First, while product modularity is shown to when complexity is high. If a company produces both
affect some aspects of new product introduction per- complex and simple products with respective types of
formance (frequency of new product introduction processes, then a spackling strategy should perhaps
and product development lead time) irrespective of be assessed (Cattani et al. 2010). The nuanced find-
the level of complexity, its effects on other perfor- ings for the individual measures of new product
mance aspects (on-time product launch and product introduction performance also serve as a cautionary
innovation) are dampened when complexity is high tale for managers that want to use modularity to tar-
(the latter being consistent with the earlier analysis get specific aspects of new product introduction per-
of the aggregate measure of new product introduc- formance (frequency of new product introduction for
tion performance). We conjecture that product mod- instance). The results demonstrate that care needs to
ularity shortens product development time and be taken in devising a modularity strategy, given the
concurrently improves the frequency of new product differential impacts of the two types of modularity
introduction, even when complexity is high, but it across the various performance indicators for differ-
may breed unexpected problems involving interde- ent complexity levels.
pendencies and integration of modules when com- Consistent with current practice, we controlled for
plexity is high, possibly creating new challenges for effects that potentially can influence new product
other modules and respective supply chain partners, introduction performance, including firm size, indus-
that hamper product innovation and eventually lead try group, innovation magnitude, and product qual-
to delays in product launch. With respect to process ity. The effects of firm size and industry were found
modularity, the results show that it has no effect on to be negligible in the presence of other variables in
on-time product launch and product development the model. However, innovation magnitude and
lead time irrespective of the level of complexity. But quality were found to be positively associated with
process modularity has a positive effect on fre- new product introduction performance across all
quency of new product introduction and product model analyses. At first glance, the finding for inno-
innovation when complexity is high (the latter being vation magnitude seems provocative. In this regard,
consistent with the earlier analysis of the aggregate it must be noted that a similar finding in an earlier
measure of new product introduction). Thus, from a seminal study showing a positive relationship
managerial standpoint, if a competitive environment between product innovativeness and new product
demands frequency of new product introduction, commercial performance was considered surprising in
product modularity is an attractive option when its time (see Kleinschmidt and Cooper 1991, p. 251;
complexity is high, especially if the firm also also Evanschitzky et al. 2012). But this result was

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
16 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

later interpreted as an exemplar of the overall posi- rapid product improvements, the user may either buy
tive relationship between risk and return. Likewise, an all-new product, or upgrade a particular module(s)
the current finding is not as surprising as it first of an existing product to bring the product up to
appears. Consider that companies may allocate par or near par with a new product. A modularly
increased effort and resources to more innovative upgradeable design “may isolate improving compo-
products because they can command premium nents from more stable ones and thereby enable an
prices, particularly if they have early indications that upgrade via replacement of only the improving
the products will be well received. Furthermore, it module(s)” (Kamrad et al. 2013, p. 289). A modu-
should be noted that truly radical (i.e., new-to-the- larly upgradeable product may be pleasing to users
world) industrial products are rather rare (Hultink because the respective product retains some value
et al. 2000), so that even if a new product is highly (Kamrad et al. 2013, Ulk€€ u et al. 2012) and can be
innovative, it may not substantially disrupt existing brought up-to-date without significant cost. But, it is
development and launch processes. Thus, it is not unclear how this will affect for instance the fre-
unexpected that innovation magnitude is positively quency of product introduction. Under ceteris paribus
related to new product introduction performance. conditions, modularly upgradeable products should
The positive association between product quality have longer lives and thus firms may not need to
and new product introduction performance is not introduce products as often in the market; rather,
unexpected either. Koufteros et al. (2002) demon- such firms may focus instead on introducing new
strated that the ability to introduce new products modules. However, which firm (the focal company
and features was highly correlated with the ability to or its suppliers) will benefit may depend on the
compete based on quality. € u and Sch-
specific supply chain configuration (Ulk€
midt 2011). If these modules are outsourced, the
6.2. New Areas for Empirical Research focal firm may be at a disadvantage as this limits
The results should be cautiously interpreted. Accord- profit opportunities and may even lead to disinter-
ing to our findings, for example, product modularity mediation.
leads to higher new product introduction perfor- In addition, it is not known how outsourcing and
mance when complexity is low; but any advantage modularity work in concert to enhance new product
may erode quickly (Ethiraj and Levinthal 2004) introduction performance (Ro et al. 2007). Theoreti-
because it may be relatively easy to imitate simple cally, in the short term, new product introduction
products. However, Ethiraj and Levinthal also sug- performance should improve as the focal firm, such
gested that inimitability is preserved in many cases as Boeing, can capitalize on the expertise of its sup-
by a competitor’s failure to copy inter-modular link- ply base to contribute towards new product intro-
ages in addition to the product modules themselves. duction performance through the outsourcing of
Our study does not address whether the impact of both design and assembly of modules. In other
product modularity on new product introduction words, a focal company can engage in resource
performance for simple products outweighs any orchestration (Sirmon et al. 2007) across its supply
downside from potential imitation. More empirical chain where each major supplier undertakes the
research is needed to explore whether a product responsibility to design (or partially design), source,
modularity – imitation trade-off exists, and if so, the integrate, and deliver a complete module (Ro et al.
exact nature of the trade-off taking into account com- 2007). For instance, Spirit AeroSystems is the inte-
plexity. In addition, even if vistas into product tech- grator for the entire flight deck of the Boeing 787
nology can be obtained through strategies such as Dreamliner. In the long term, however, due to the
reverse engineering, this is not necessarily the case outsourcing of complete modules to suppliers, there
for process technologies. Thus, even if product mod- will be a diminishing desire to retain product
ularity is replicable by a firm’s competitor, advan- designers and process designers on staff, which
tage may still be preserved by the competitor’s may then deplete the firm’s future capacity to inno-
inability to perceive and/or copy the firm’s associ- vate. Without its own cadre of talented product and
ated modular process (if it has developed one) (see process engineers, innovation efforts may be stifled
Teece et al. 1997). More research is needed regarding and the firm may not be able to assure the integrity
these issues. of each module and importantly, safeguard the con-
While we examined the impact of product and pro- nectivity across modules designed and sourced from
cess modularity on new product introduction perfor- various suppliers (Ulk€€ u and Schmidt 2011). This
mance, there are specific aspects of modularity that may have an adverse effect on the ability of a focal
are currently understudied. For instance, Kamrad firm to improve its new product introduction per-
et al. (2013) examined the value of modularly upgrade- formance. Future research can examine the relation-
able products via an economic model. In the face of ship between product modularity, process

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 17

modularity, and new product introduction perfor- the use of these three aspects to define other types of
mance in light of specific supply chain configura- modularity such as “modularity-in-use” (Kamrad
tions. In other words, the effects of modularity on et al. 2013) or supply chain modularity.
new product introduction performance may be sub- In defining modularity, we avoided measurement
ject to a specific type of supply chain configuration cross-contamination with related constructs. Past
(vertical integration vis-a-vis decentralized supply studies have often failed to disentangle product
€ u and Schmidt (2011) examine the role of
chain). Ulk€ modularity (for example) from related and/or com-
supply chain configuration in the realm of modular plementary concepts such as commonality, process
vs. integral designs through an analytical approach modularity, dynamic teaming, etc. For example, Tu
while Ro et al. (2007) study modularity and supply et al.’s (2004) single modularity construct encom-
chain configurations in the context of the US.auto- passing both product and process modularity (as
motive industry via case studies, but large-scale well as dynamic teaming) does not allow for the dis-
empirical evidence is lacking. entanglement of product vs. process modularity
effects. Nevertheless, Tu et al.’s inclusion of dynamic
6.3 Measurement and New Prospects teaming raises the possibility that organizational
After conceptually defining the core aspects of modu- design may impact modularity and its subsequent
larity as functional binding, interface standardization, effects. Future modularity research could also focus
and decomposability, we measured product and pro- on supply chain design. For example, Ulk€ € u and Sch-
cess modularity as separate constructs. By tapping midt (2011) analytically modeled external product
these three key aspects for both modularity con- development to show that modular architectures are
structs, we seized the contrast between product vs. more likely when the parties have adversarial (vs.
process modularity (i.e., modularity in product collaborative) relationships. Future research could
design architecture vs. modularity in production pro- empirically explore these kinds of issues, using the
cess architecture) while retaining the core meanings measures of product and process modularity devel-
underlying modularity. An area for future study is oped herein.

Appendix A. Hierarchical Regression Analyses for Each Indicator of Product


Introduction Performance†
Frequency of product On-time Product Development
introduction Product Launch lead time Product innovation

Independent variables Coefficients SE Coefficients SE Coefficients SE Coefficients SE


Constant 1.996*** 0.707 2.351*** 0.743 2.853*** 0.641 2.622*** .679
Control variables
Industry 1 0.322 0.401 0.110 0.421 0.239 0.364 .655* .385
Industry 3 0.256 0.299 0.533** 0.306 0.004 0.264 .216 .279
Industry 4 0.294 0.314 0.012 0.330 0.027 0.285 .302 .301
Industry 5 0.020 0.401 0.225 0.422 0.186 0.364 .480 .385
Firm size 0.041 0.058 0.002 0.061 0.033 0.053 .086* .056
Innovation magnitude 0.034** 0.018 0.029* 0.019 0.053*** 0.016 .085*** .017
Product quality (reliability) 0.564*** 0.101 0.509*** 0.106 0.392*** 0.091 .368*** .096
Main effects
Product modularity 0.303** 0.123 0.232** 0.130 0.159* 0.112 .187* .118
Process modularity 0.081 0.121 0.077 0.127 0.094 0.110 .163* .116
Complexity 0.011 0.010 0.009 0.011 0.012* 0.009 .005 .010
Interaction effects
Product Modularity 9 complexity 0.000 0.000 0.000* 0.000 0.000 0.000 .000*** .000
Process Modularity 9 complexity 0.002* 0.001 0.000 0.001 0.001 0.000 .004*** .001
Model fit
R//R2 0.607//0.369 0.571//0.327 0.596//0.355 0.667//0.445
Adjusted R2 0.292 0.245 0.277 0.377
F (Significance) 4.823 (0.000) 4.001 (0.000) 4.543 (0.000) 6.607 (0.000)
F change (Prob > F change) Model 3 Model 2 Model 3- Model 2 Model 3- Model 2 Model 3- Model 2
0.891 (0.413) 1.254 (0.290) 0.891 (0.412) 6.142 (0.003)

*p < 0.10, **p < 0.05, ***p < 0.01, Bold values represent statistically significant regression coefficients.

Only Model 3 for each regression model is shown.

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
18 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

Appendix B. Process Modularity and Frequency of Product Introduction: Note


Their Relationship at Low Vs. High Complexity
1
See Parker (2010), whose dissertation was part of an over-
5
arching research project on modularity and complexity for
additional details regarding data collection.
Frequency of Product

4
Introduction

3
Low Complexity

2 High Complexity
References
Ahmad, S., R. G. Schroeder, D. N. Mallick. 2010. The relationship
1 among modularity, functional coordination, and mass cus-
tomization, Implications for competitiveness. Eur. J. Innov.
0 Manag. 13(1): 46–61.
Low Process Modularity High Process Modularity
Allworth, J. 2013. The 787’s problems run deeper than out-sour-
cing. Harv. Bus. Rev. Available at https://hbr.org/2013/01/
the-787s-problems-run-deeper-t# (accessed January 30, 2013).
Appendix C. Product Modularity and On Time Product Launch: Their
Anderson, J. C., D. W. Gerbing. 1988. Structural equation model-
Relationship at Low Vs. High Complexity
ing in practice: A review of the two-step approach. Psychol.
Bull. 103(3): 53–66.
5
Antonio, K. W., R. C. Yam, E. Tang. 2007. The impact of product
modularity on competitive capabilities and performance: An
On Time Product Launch

4
empirical study. Int. J. Prod. Econ. 105(1): 1–20.
3 Armstrong, J. S., R. S. Overton. 1977. Estimating nonresponse bias
Low Complexity in mail surveys. J. Mark. Res. 14(3): 396–402.
2 High Complexity Aydin, S., A. T. Cetin, G. Ozer. 2007. The relationship between
marketing and product development process and their effects
1 on firm performance. Acad. Mark. Stud. J. 11(1): 53–68.
Baldwin, C. Y., K. B. Clark. 1997. Managing in an age of modular-
0 ity. Harv. Bus. Rev. 75(5): 84–93.
Low Product Modularity High Product Modularity Baldwin, C. Y., K. B. Clark. 2000. Design Rules: The Power of Modu-
larity (Volume 1). The MIT Press, Cambridge, MA.
Bayus, B. L. 1997. Speed-to-market and new product performance
Appendix D. Product Modularity and Product Innovation: Their trade-offs. J. Prod. Innov. Manage 14(6): 485–497.
Relationship at Low Vs. High Complexity Blau, P. M., R. A. Schoenherr. 1971. The Structure of Organizations.
Basic Books, New York.
5
Brabazon, P. G., B. MacCarthy, A. Woodcock. 2010. Mass cus-
tomization in the automotive industry: Comparing interdealer
4 trading and reconfiguration flexibilities in order fulfillment.
Product Innovation

Prod. Oper. Manag. 19(5): 489–502.


3 Browne, J., J. Devlin, A. Rolstadas, B. Andersen. 1998. Perfor-
Low Complexity mance measurement – the ENAPS approach. Int. J. Bus. Trans-
2 High Complexity
for. 1(2): 73–84.
Cabigiosu, A., A. Camuffo. 2012. Beyond the “mirroring” hypoth-
1 esis: Product modularity and interorganizational relations in
the air conditioning industry. Organ. Sci. 23(3): 686–703.
Calantone, R. J., N. Harmancioglu, C. Droge. 2010. Inconclusive
0
Low Product Modularity High Product Modularity innovation “returns”: A meta-analysis of research on innova-
tion in new product development. J. Prod. Innov. Manage
27(7): 1065–1081.
Appendix E. Process Modularity and Product Innovation: Their Cattani, K. D., E. Dahan, G. Schmidt. 2010. Lowest cost may not
lower total cost: Using “spackling” to smooth mass-cus-
Relationship at Low Vs. High Complexity
tomized production. Prod. Oper. Manag. 19(5): 531–545.
Chao, R., S. Kavadias. 2008. A theoretical framework for manag-
5
ing the new product development portfolio: When and how
to use strategic buckets. Manage. Sci. 54(5): 907–921.
4
Choi, T. Y., D. R. Krause. 2006. The supply base and its complex-
Product Innovation

ity: Implications for transaction costs, risks, responsiveness,


3 and innovation. J. Oper. Manag. 24(5): 637–652.
Low Complexity
Cohen, M. A., J. Eliasberg, T. H. Ho. 1996. New product develop-
2 High Complexity ment: The performance and time-to-market tradeoff. Manage.
Sci. 42(2): 173–186.
1 Cooper, R. G., E. J. Kleinschmidt. 1994. Determinants of timeliness
in product development. J. Prod. Innov. Manage 11(5): 381–396.
0 Craighead, C. W., D. J. Ketchen, K. S. Dunn, G. G. Hult. 2011.
Low Process Modularity High Process Modularity Addressing common method variance: Guidelines for survey

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society 19

research on information technology, operations, and supply Henderson, R. M., K. B. Clark. 1990. Architectural innovation: The
chain management. IEEE Trans. Eng. Manage. 58(3): 578–588. reconfiguration of existing product technologies and the fail-
Das, A., R. Narasimhan. 2000. Purchasing competence and its ure of established firms. Adm. Sci. Q. 35(1): 9–30.
relationship with manufacturing performance. J. Supp. Chain Hultink, E. J., S. Hart, H. S. J. Robben, A. Griffin. 2000. Launch
Manag. 36(2): 17–28. decisions and new product success: An empirical comparison
Dewar, R. D., J. E. Dutton. 1986. The adoption of radical and of consumer and industrial products. J. Prod. Innov. Manage
incremental innovations: An empirical analysis. Manage. Sci. 17(1): 5–23.
32(11): 1422–1433. Jacobs, M., S. K. Vickery, C. Droge. 2007. The effects of product
Droge, C., J. Jayaram, S. K. Vickery. 2000. The ability to minimize modularity on competitive performance - do integration
the timing of new product development and introduction: An strategies mediate the relationship? Int. J. Oper. Prod. Manag.
examination of antecedent factors in the North American 27(9–10): 1046–1068.
automobile supplier industry. J. Prod. Innov. Manage 17(1): Jacobs, M., C. Droge, S. K. Vickery, R. Calantone. 2011. Product and
24–40. process modularity’s effects on manufacturing agility and firm
Droge, C., J. Jayaram, S. K. Vickery. 2004. The effects of internal growth performance. J. Prod. Innov. Manage 28(1): 124–138.
versus external integration practices on time-based perfor- Kamrad, B., G. M. Schmidt, S. Ulk€ € u. 2013. Analyzing product
mance and overall firm performance. J. Oper. Manag. 22(6): architecture under technological change: Modular upgrade-
557–573. ability tradeoffs. IEEE Trans. Eng. Manage. 60(2): 289–300.
Duray, R., P. T. Ward, G. W. Milligan, W. L. Berry. 2000. Ap- Kauffman, S. 1993. The Origins of Order: Seld-Organization and
proaches to mass customization: Configurations and empirical Selection in Evolution. Oxford University, New York.
validation. J. Oper. Manag. 18(6): 605–625. Kauffman, S. 1995. At Home in the Universe. Oxford University
Eisenhardt, K. M., B. N. Tabrizi. 1995. Accelerating adaptive pro- Press, New York.
cesses: Product innovation in the global computer industry. Kleinschmidt, E. J., R. G. Cooper. 1991. The impact of product inno-
Adm. Sci. Q. 40(1): 84–110. vativeness on performance. J. Prod. Innov. Manage 8(4): 240–251.
El-Haik, B., K. Yang. 1999. The components of complexity in engi- Kogut, B., U. Zander. 1992. Knowledge of the firm, combinative
neering design. IIE Trans. 31(10): 925–947. capabilities, and the replication of technology. Organ. Sci. 3(3):
Ethiraj, S. K., D. Levinthal. 2004. Modularity and innovation in 383–397.
complex systems. Manage. Sci. 50(2): 159–173. Koufteros, X. A. 1999. Testing a model of pull production: A para-
Ethiraj, S. K., D. Levinthal, R. R. Roy. 2008. The dual role of digm for manufacturing research using structural equation
modularity: Innovation and imitation. Manage. Sci. 54(5): modeling. J. Oper. Manag. 17(4): 467–488.
939–955. Koufteros, X. A., M. A. Vonderembse, W. J. Doll. 2002. Examining
Ettlie, J. E., W. P. Bridges, R. D. O’keefe. 1984. Organization strat- the competitive capabilities of manufacturing firms. Struct.
egy and structural differences for radical versus incremental Equ. Model. 9(2): 256–282.
innovation. Manage. Sci. 30(6): 682–695. Koufteros, X. A., T. C. E. Cheng, K. H. Lai. 2007. “Black-box” and
Evanschitzky, H., M. Eisend, R. J. Calantone, Y. Jiang. 2012. Suc- “gray-box” supplier integration in product development:
cess factors of product innovation: An updated meta-analysis. Antecedents, consequences and the moderating role of firm
J. Prod. Innov. Manage 29(S1): 21–37. size. J. Oper. Manag. 25(4): 847–870.
Feitzinger, E., H. K. Lee. 1997. Mass customization at Hewlett- Kusiak, A. 2002. Integrated product and process design: A modu-
Packard: The power of postponement. Harv. Bus. Rev. 75(1): larity perspective. J. Eng. Des. 13(3): 223–231.
116–121. La Porte, T. R. 1975. Organized Social Complexity: Challenge to Poli-
Fine, C. H., B. Golany, H. Naseraldin. 2005. Modeling tradeoffs in tics and Policy. Princeton University Press, Princeton, NJ.
three-dimensional concurrent engineering: A goal program- Langlois, R. N. 2002. Modularity in technology and organization.
ming approach. J. Oper. Manag. 23(3, 4): 389–403. J. Econ. Behav. Organ. 49(1): 19–37.
Fixson, S. K. 2005. Product architecture assessment: A tool to link Langlois, R. N., P. L. Robertson. 1992. Networks and innovation in
product, process, and supply chain decisions. J. Oper. Manag. a modular system: Lessons from the microcomputer and
23(3): 345–369. stereo component industries. Res. Policy 21(4): 297–313.
Flynn, B. B., J. E. Flynn. 1999. Information-processing alternatives Lau, A. K. W., R. Yam, E. Tang. 2011. The impact of product
for coping with manufacturing environment complexity. modularity on new product performance: Mediation by pro-
Decis. Sci. 30(4): 1021–1052. duct innovativeness. J. Prod. Innov. Manage 28(2): 270–284.
Fornell, C., D. F. Larcker. 1981. Evaluating structural equations Li, L. X. 2000. An analysis of sources of competitiveness and per-
models with unobservable variables and measurement error. formance of Chinese manufacturers. Int. J. Oper. Prod. Manag.
J. Mark. Res. 18(1): 39–50. 20(3): 299–315.
Frazier, P. A., A. P. Tix, K. E. Barron. 2004. Testing moderator Lin, B. W. 2003. Cooperating for supply chain effectiveness: Man-
and mediator effects in counseling psychology research. J. ufacturing strategy for Chinese OEMs. Int. J. Manuf. Technol.
Couns. Psychol. 51(1): 115–134. Manage. 5(3): 232–245.
Frenken, K. 2000. A complexity approach to innovation networks. Loch, C. H., C. Terwiesch, S. Thomke. 2001. Parallel and sequen-
The case of the aircraft industry (1909–1997). Res. Policy 29(2): tial testing of design alternatives. Manage. Sci. 47(5): 663–678.
257–272.
Loch, C., J. Mihm, A. Huchzermeier. 2003. Concurrent engineer-
Griffin, A. 2002. Product development cycle time for business-to- ing and design oscillations in complex engineering projects.
business products. Ind. Mark. Manage. 31(4): 291–304. Concurr. Eng. 11(3): 187–199.
Harrison, D. A., M. E. McLaughlin, T. M. Coalter. 1996. Context, Lukas, B. A., O. C. Ferrell. 2000. The effect of market orientation
cognition, and common method variance: Psychometric and on product innovation. J. Acad. Mark. Sci. 28(2): 239–247.
verbal protocol evidence. Organ. Behav. Hum. Decis. Process.
Lynn, G., K. D. Abel, W. S. Valentine, R. C. Wright. 1999. Key fac-
68(3): 246–261.
tors in increasing speed to market and improving new product
Hayes, A. F. 2013. Introduction to Mediation, Moderation, and Condi- success rates. Ind. Mark. Manage. 28(4): 319–326.
tional Process Analysis. Guilford Press, New York.

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
Vickery, Koufteros, Droge, and Calantone: Product and Process Modularity & Performance
20 Production and Operations Management 0(0), pp. 1–20, © 2015 Production and Operations Management Society

Malhotra, N. K., S. K. Sung, P. Ashutosh. 2006. Common method Schilling, M. A., H. K. Steensma. 2001. The use of modular orga-
variance in IS research: A comparison of -alternative nizational forms: An industry-level analysis. Acad. Manag. J.
approaches and a reanalysis of past research. Manage. Sci. 44(6): 1149–1168.
52(12): 1865–1883. Simon, H. A. 1957. Models of Man, Social and Rational. John Wiley
Marengo, L., G. Dosi, P. Legrenzi, C. Pasquali. 2000. The structure and Sons Inc, New York, NY.
of problem solving knowledge and the structure of organiza- Simon, H. A. 1962. The architecture of complexity. Proc. Am. Phi-
tions. Ind. Corp. Change 9(4): 757–788. los. Soc. 106: 467–482.
McFarlin, D. B., P. D. Sweeney. 1992. Distributive and procedural Smith, K. G., C. J. Collins, K. D. Clark. 2005. Existing knowledge,
justice as predictors of satisfaction with personal and organi- knowledge creation capability, and the rate of new product
zational outcomes. Acad. Manag. J. 35(3): 626–637. introduction in high-technology firms. Acad. Manag. J. 48(2):
Mihm, J., C. Loch, A. Huchzermeier. 2003. Problem–solving oscilla- 346–357.
tions in complex engineering projects. Manage. Sci. 49(6): 733–750. Sirmon, D. G., M. A. Hitt, D. R. Ireland. 2007. Managing firm
Narasimhan, R., J. Jayaram, J. R. Carter. 2001. An empirical exam- resources in dynamic environments to create value: Looking
ination of the underlying dimensions of purchasing compe- inside the black box. Acad. Manag. Rev. 32(1): 273–292.
tence. Prod. Oper. Manag. 10(1): 1–15. Sorescu, A. B., R. K. Chandy, J. C. Prabhu. 2003. Sources and
Novak, S., S. D. Eppinger. 2001. Sourcing by design: Product com- financial consequences of radical innovation: Insights from
plexity and the supply chain. Manage. Sci. 47(1): 189–204. pharmaceuticals. J. Mark. 67(4): 82–102.
Parker, D. B. 2010. Modularity and complexity: an examination of Staudenmayer, N., M. Tripsas, C. L. Tucci. 2005. Interfirm modu-
the effects of product structure on the intricacy of production larity and its implications for product development. J. Prod.
systems. Unpublished doctoral dissertation, Michigan State Innov. Manage 22(4): 303–321.
University, East Lansing, MI. Stremersch, S., A. M. Weiss, B. G. Dellaert, R. T. Frambach. 2003.
Pil, F. K., S. K. Cohen. 2006. Modularity: Implications for imita- Buying modular systems in technology-intensive markets. J.
tion, innovation, and sustained advantage. Acad. Manag. Rev. Mark. Res. 40(3): 335–350.
31(4): 995–1011. Tatikonda, M. V., M. M. Montoya-Weiss. 2001. Integrating opera-
Podsakoff, P. M., D. W. Organ. 1986. Self-reports in organizational tions and marketing perspectives of product innovation: The
research: Problems and prospects. J. Manage. 12(4): 531–544. influence of organizational process factors and capabilities on
Ro, Y. K., J. K. Liker, S. K. Fixson. 2007. Modularity as a strategy development performance. Manage. Sci. 47(1): 151–172.
for supply chain coordination: The case of US auto. IEEE Teece, D. J., G. Pisano, A. Shuen. 1997. Dynamic capabilities and
Trans. Eng. Manage. 54(1): 172–189. strategic management. Strateg. Manag. J. 18(7): 509–533.
Roberts, P. W. 1999. Product innovation, product-market competi- Tu, Q., M. A. Vonderembse, T. S. Ragu-Nathan, B. Ragu-Nathan.
tion and persistent profitability in the US pharmaceutical 2004. Measuring modularity-based manufacturing practices
industry. Strateg. Manag. J. 20(7): 655–670. and their impact on mass customization capability: A cus-
Rosenberg, N. 1982. Inside the Black Box: Technology and Economics. tomer-driven perspective. Decis. Sci. 35(2): 147–167.
Cambridge University Press, Cambridge, MA. € u, S., G. M. Schmidt. 2011. Matching product architecture and
Ulk€
Salvador, F. 2007. Toward a product system modularity construct: supply chain configuration. Prod. Oper. Manag. 20(1): 16–31.
Literature review and reconceptualization. IEEE Trans. Eng. € u, S., C. V. Dimofte, G. M. Schmidt. 2012. Consumer valuation
Ulk€
Manage. 54(2): 219–240. of modularity upgradeable products. Manage. Sci. 58(9): 1761–
Salvador, F., C. Forza, M. Rungtusanatham. 2002. Modularity, 1776.
product variety, production volume, and component sour- Ulrich, K. T. 1995. The role of product architecture in the manu-
cing: Theorizing beyond generic prescriptions. J. Oper. Manag. facturing firm. Res. Policy 24(3): 419–440.
20(5): 549–575. Ulrich, K., S. D. Eppinger. 2004. Product Design and Development,
Sanchez, R., J. T. Mahoney. 1996. Modularity, flexibility, and 3rd edn. McGraw-Hill, New York.
knowledge management in product and organizational Voordijk, H., B. Meijboom, J. D. Haan. 2006. Modularity in supply
design. Strateg. Manag. J. 17(S2): 63–76. chains: A multiple case study in the construction industry.
Sawhney, R., C. Piper. 2002. Value creation through enriched mar- Int. J. Oper. Prod. Manag. 26(6): 600–618.
keting-operations interfaces: An empirical study in the Williamson, P. J., E. Yin. 2014. Accelerated innovation: The new
printed circuit board industry. J. Oper. Manag. 6(4): 259–272. challenge from China. MIT Sloan Manage. Rev. 55(4): 27–34.
Schilling, M. A. 2000. Toward a general modular systems theory Worren, N., K. Moore, P. Cardona. 2002. Modularity, strategic
and its application to interfirm product modularity. Acad. flexibility, and firm performance: A study of the home appli-
Manag. Rev. 25(2): 312–334. ance industry. Strateg. Manag. J. 23(12): 1123–1140.

Please Cite this article in press as: Vickery, S. K., et al. Product Modularity, Process Modularity, and New Product Introduction Perfor-
mance: Does Complexity Matter? Production and Operations Management (2015), doi 10.1111/poms.12495
View publication stats

You might also like