modularity-as-property, …fixson and park, 2008). modularity-as-property and...

MODULARITY-AS-PROPERTY,MODULARIZATION-AS-PROCESS, AND‘MODULARITY’-AS-FRAME: LESSONS FROMPRODUCT ARCHITECTURE INITIATIVES IN THEGLOBAL AUTOMOTIVE INDUSTRY

JOHN PAUL MACDUFFIE*Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania,U.S.A.

Modularity is a design property of the architecture of products, organizations, and interfirmnetworks; modularization is a process that affects those designs while also shaping firmboundaries and industry landscapes; and ‘modularity’ is a cognitive frame that guidescategorization and interpretation of a wide array of economic phenomena. Modularity-as-property and modularization-as-process are deeply intertwined; while modularization pro-cesses are ubiquitous and perpetual as engineers and managers seek to understandinterdependencies across the boundaries of product and organizational architecture, the extentto which modularity-as-property is achieved must be assessed empirically. The framing of‘modularity’ affects strategy by prompting a particular dynamic—and directionality—in theinterplay between modularity-as-property and modularization-as-process. I analyze productarchitecture initiatives in the global automotive industry, examining first the industry-levelantecedents of the emergent production-based definition of modules and then two firm-levelmodularity initiatives that both were based on this common definition, but framed theirstrategies differently. In the first case, a ‘modularity’ frame based on a computer industryanalogy resulted in overemphasis on achieving modularity-as-property that created barriers tolearning about cross-module interdependencies. In the second case, early emphasis onmodularization-as-process yielded quasi-integrated organizational arrangements that facili-tated long-term design improvements. Overall, this single-industry case study demonstrates theimportance of examining the context-specific antecedents of module definition; the multiplicityof potential barriers to modularity that can lead to persistent integrality; the need for longi-tudinal inquiry into the ‘mirroring’ hypothesis that pays as much attention to process as toproperty; and the power of modularity as a cognitive frame, which helps explain divergentfindings in modularity research. Copyright © 2013 Strategic Management Society.

INTRODUCTION

Modularity is a design property of the architecture ofproducts, organizations, and interfirm networks;modularization is a process that affects those designswhile also shaping firm boundaries and industrylandscapes; and ‘modularity’ is a cognitive framethat guides categorization and interpretation ofa wide array of global economic phenomena.Modularity-as-property and modularization-as-

Keywords: modularity; modularization; product architecture;organizational architecture; cognitive frame; framing; analogy;interdependence; automotive industry; industry studies*Correspondence to: John Paul MacDuffie, Wharton School,University of Pennsylvania, 3105 Steinberg Hall-DietrichHall, Philadelphia, PA 19104, U.S.A. E-mail: [email protected]

Global Strategy JournalGlobal Strat. J., 3: 8–40 (2013)

Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1111/j.2042-5805.2012.01048.x

Copyright © 2013 Strategic Management Society

process are deeply intertwined; while modulariza-tion processes are ubiquitous and perpetual asengineers and managers seek to understand interde-pendencies across the boundaries of product andorganizational architecture, the extent to whichmodularity-as-property is achieved is subject to con-tingencies and must be assessed empirically. Theframing of ‘modularity’ affects strategy by prompt-ing a particular dynamic—and directionality—inthe interplay between modularity-as-property andmodularization-as-process.

These three aspects of modularity—property,process, and frame—are frequently entangled ana-lytically, leading to multiple and divergent predic-tions, contradictory observations, and ambiguousconclusions. For example, is Apple’s iPod an exem-plar of the cost-saving and innovation-speedingbenefits of modular product architecture (e.g.,Mudambi, 2008)? Or do its innovations arise fromthe integration of its elegant hardware design withthe iTunes software under a new business model foracquiring music? Understanding how either state-ment, or both, could be true requires a more multi-faceted and context-anchored analysis of these threeaspects of modularity.

The confusion about modularity is frequentlyevident in discussions of its role in globalizationprocesses. When modularity is cited as a driver of theoutsourcing that can lead to ‘hollowing out’ for firmsand nations (Herrigel, 2004), what is emphasizedare the properties of standardized, commoditizedcomponents whose outsourcing allows knowledge tomove out of the focal firm or country via a possiblyirreversible division of labor. Accordingly, any signof product modularity may be taken as evidence thata technologically determined path to a disintegratedsupply chain (and offshore contract manufacturerswho eventually become design rivals of developedcountry incumbents) is inevitable. Yet the ongoingprocess of coordination at the organizational inter-face for outsourced tasks can reveal unexpectedinterdependencies and lead to (partial) reintegrationof component design and/or production into thefocal firm or nation (Berger, 2006). Disintegration ofproduct or organizational architectures from global-ization is rarely a final destination; fluidity, revers-ibility, and shifts in strategic logic are the norm.

The interaction between globalization and modu-larization processes is at once heavily influenced byantecedents—what initiates these processes, whatgoals are defined, what strategy is being pursued—and highly variable as to outcome, because of the

contingent ways in which modularization processesaffect the extent of modularity-at-property that isachieved. If globalization initiatives by multinationalfirms are viewed as quasi-experimental tests of thefirm’s ability to extend its capabilities into othermarkets and institutional contexts (or to move tasksoutside the firm’s boundary to a global supplier viaoutsourcing) and modularity initiatives are similarlyunderstood as explorations of how best to manageinterdependencies across product and organizationalboundaries, it becomes clear why the interaction ofthese emergent learning processes is unlikely to leadto deterministic consequences.

This article proposes grounded attention to thecontext-specific particulars of modularity, emphasiz-ing these intertwined aspects of property, process,and frame, as remedy to the risks already mentioned.For example, innovation viewed from themodularity-as-property perspective emphasizeshigher interdependencies within modules andreduced requirements for coordination acrossmodules that allows focused and autonomous atten-tion to component design by specialized suppliersand leads to more rapid and less constrained innova-tion (Langlois and Robertson, 1992; Ulrich, 1995;Baldwin and Clark, 2000). Yet a more process-basedview of modularity is needed to understand the now-commonplace observation that coordination-and-communication intensive activities can precedeemergence of modular properties and standardizedinterfaces. When modular properties are viewed lon-gitudinally, i.e., as emergent from a modularizationprocess, they are often revealed to be an ex postconsequence of integrative forms of organizingrather than the ex ante basis for independent orga-nizing (Cabigiosu and Camuffo, 2011; Stauden-mayer, Tripsas, and Tucci, 2005; Brusoni, Principe,and Pavitt, 2001).

This challenges the causal assumption that‘product designs organization’ (Sanchez andMahoney, 1996) and means that even when productand organization evolve to mirror each other, thestarting point may be an organizational change asreadily as a technical or product change (Campag-nolo and Camuffo, 2009). Furthermore, whileproduct architectures are often portrayed as movinginexorably from integral to modular as early idiosyn-crasies give way to a dominant design (Argyres andBigelow, 2010), counterexamples can be found inwhich firms move deliberately from modular to inte-gral product architecture through differentiatinginnovations that provide advantage over competitors

Modularity-as-Property, Modularization-as-Process, and ‘Modularity’-as-Frame 9

Copyright © 2013 Strategic Management Society Global Strat. J., 3: 8–40 (2013)DOI: 10.1111/j.2042-5805.2012.01048.x

(Fixson and Park, 2008). Modularity-as-propertyand modularization-as-process are so intertwined, soembedded in dynamics of reciprocal feedback, thatexamination of one without the other is inevitablyincomplete; unidirectional views of their relation-ship can lead to incomplete and potentially distortedconclusions.

‘Modularity’ as a cognitive frame enters thisaccount as a crucial influence on how the intertwin-ing of modularity-as-property and modularization-as-process unfolds. For example, when a firm frames‘modularity’ as a strategy for allowing a supplier towork autonomously on design innovations, by estab-lishing certain product properties, i.e., clearlydefined module boundaries and standardized inter-face specifications, the consequence may be amutual reduction in coordination efforts that inter-feres with the learning essential to modularizationprocesses. Managers may develop ‘modularity’-as-frame based on the example of another industry’ssuccess with modular product and organizationalarchitecture, neglecting to recognize how thesearchitectures differ across industries. The frequentreliance on the personal computer as the exemplar ofmodularity and harbinger of the future is a case inpoint. The PC, in its precise mapping of functions tocomponents, open industry-level standard interfaces,and clean correspondence between module and firmboundaries, may be more the exception than the rule(see Jacobides, MacDuffie, and Tae, 2012). Used asthe analogy that anticipates these characteristics, inboth product and organizational architecture, thecomputer industry has provided a cognitive framethat overemphasizes achieving modularity-as-property at the expense of attention to what can belearned from modularization processes. Such ananalogy conveys the deterministic idea of an inexo-rable and self-reinforcing logic pointing toward oneset of inevitable outcomes—technological, organiza-tional, or both (due to mirroring).

In contrast, the best modularization researchavoids determinism by being grounded in closeobservation of the particulars of a given industry,product, firm, country/region, and/or suppliernetwork (the IBM System/360 computer in Baldwinand Clark, 2000; aircraft engines and chemical engi-neering in Brusoni, Principe, and Pavitt, 2001; airconditioners in Cabigiosu and Camuffo, 2011;bicycle drive trains in Fixson and Park, 2008; auto-motive front ends in Fourcade and Midler, 2004; flatpanel displays in notebook computers in Hoetker,2006; mortgage banking in Jacobides, 2005; mobile

handsets in Mudambi, 2008). This research revealsthat the properties of modularity are layered, can beinternally conflicting, and are often incompletelyrealized; that both costs and capabilities are crucialdeterminants of when modularization processes willadvance to modularity-as-property and when theywill not; and that modularization generates dynamicsthat can move architectures away from, but also backtoward, integration.

Here I examine the global automotive industry.While certain industries allow clear characteriza-tions about their technological and organizationalarchitecture (cf. the personal computer industry, asnoted earlier), the global auto industry is a complexcase that is more difficult to classify. It is exactlythese classification difficulties that provide thepotential for insight and theoretical advance.

Numerous modularity initiatives in the globalautomotive industry in recent years have faced unex-pected technological and organizational barriers. Mygoal is to identify what we can learn about modular-ity and modularization from studying the gapbetween hoped-for goals versus the reality of whathas been achieved in this particular context. I startwith a theoretical review to disentangle these con-cepts of modularity-as-property, modularization-as-process, and ‘modularity’-as-frame that are oftenconflated, so that we can see each one distinctly andin relation to the others. I then apply these conceptsas lenses to three different sets of events: first, under-standing the idiosyncratic definition of ‘module’ thathas emerged in the auto industry; and second andthird, comparing cases of modularity initiatives atFord Motor Company and Hyundai Motor Company,selected for theoretical contrast. In the discussionand conclusion, I draw out the implications from thissingle-industry study for future research on modu-larization and for the global architecture of the mul-tinational firm. I use the computer industry as a foilthroughout, not for a full sectoral comparison butrather to highlight the broader theoretical signifi-cance of this study of the automotive industry.

THEORETICAL OVERVIEW

The literature on modularity is diverse, large, anddifficult to absorb. In this overview, I cover keyreference points to orient the reader for the caseanalyses, emphasizing the distinction (and inter-relationship) between modularity-as-property andmodularization-as-process. The term ‘modularity’ is

10 J. P. MacDuffie


often used, in academic literature as well as mana-gerial discourse, to refer to both aspects, a source ofmuch confusion. I distinguish technical property andorganizational process, while also arguing that bothaspects must be examined simultaneously to under-stand the antecedents and consequences of modular-ity and the dynamics of modularization. I treat‘modularity as frame’ separately, at the end of thissection.

Modularity-as-property

In the classic ideal-typical definition by Ulrich(1995), modularity is a design property of productarchitecture such that functions have a one-to-onemapping onto physical components and interfacesare specified for ready decoupling. Its conceptualopposite, integrality, is characterized by the absenceof one-to-one mapping, i.e. ‘one-to-many,’ in whichmultiple physical components fulfill a function, or‘many-to-one’, in which multiple functions are ful-filled in a single physical component, in both caseswith tightly coupled interfaces. Full modularitydecomposes the complexity of a product into fullyseparable components, such that each componentcan be developed independently without necessarycoordination with other components (Langlois,2002). In the ideal type, the combination of fullmodular separability and fully specified interfacesallowing easy interoperability facilitates the reduc-tion of coordination costs and rapid, autonomousinnovation.

While conceptually valuable, this ideal-type ofmodularity fits relatively few products. (That it fitsthe personal computer quite well is evident fromhow often this product is used as the prototypicalexample of modularity; yet as noted earlier, the PCmay be more the exception than the rule.) In fact,most products exhibit a combination of modular andintegral characteristics. While full decompositionmay be essential for complete modularity, it is rarelyachieved; instead, efforts to segment a complexsystem yield near decomposition (Simon, 1981), inwhich some interdependencies remain acrossmodules, yet many fewer than the interdependenciescontained within module boundaries.

Baldwin and Clark define modules in preciselythis way, i.e., with interdependencies that are greaterwithin modules than across modules. Baldwin(2008) goes on to describe module boundaries as‘thin crossing places,’ a metaphor for the reduced(yet not eliminated) coordination ‘traffic’ between

and among modules. Schilling (2000: 315) notes thatall systems are modular to some degree ‘since [they]are characterized by some degree of coupling(whether loose or tight) between components, andvery few systems have components that are com-pletely inseparable and cannot be recombined.’ (Bythe same logic, all systems are also, to some degree,integral.) For this reason, Fixson (2005) argues thatthe mix of modular and integral characteristics in aproduct can be best understood by assessing thedegree to which interfaces are coupled or decoupledseparately from the extent of function-to-componentmapping.

Furthermore, since each component of a productcan also be analyzed as a product with its own com-ponents, assessing any characteristic of modularity(e.g., the degree of function-to-component match-ing; whether elements are tightly or loosely coupled)depends entirely on the level of the product hierar-chy that is the focus of inquiry. The result, as Fixson(2005: 351) argues, is that ‘a label for the entireproduct is essentially creating an average assess-ment of the product architecture.’ For more preci-sion, the level of the product hierarchy must bespecified before assessing the extent of modularity.This is not only a choice for the external analyst.Those agents actively involved in choosing moduleboundaries and specifying interfaces will evaluatemodularity-as-property differently depending on thelevel where they fix their gaze—whether they aregazing at a component, product, or organization.

I now shift attention from modularity as a noun to‘modularizing’ as a verb, in the spirit of Karl Weick’s(1969) appeal to scholars to shift their attention fromorganizations to ‘organizing.’

Modularization as process

Viewed through the lens of Simon’s idea of neardecomposability, modularization is an evolutionaryprocess that is pursued and gradually advanced, butnever fully completed, rather than an architecturalproperty that is set as a design goal, achieved, andstabilized. As a process, modularization involvesmapping functions to components to create ‘thincrossing places’ at the module boundary and thensetting out to learn and master the remaining inter-dependencies across modules in order to make surethat interfaces can accommodate them. Thus, thenormal process by which managers and engineersaccumulate knowledge about products advancesmodularization by learning how to manage the



impact of interdependencies, whether or not thatlearning advances modularity-as-property.

‘Modularization as process’ also highlights therole of the ‘architect’ who sets the design rules thatmark module boundaries and establish interfacerequirements. Baldwin and Clark (2000) emphasizethat, depending on the interests or objectives pursuedby the architect, different module boundaries maybe chosen; they differentiate ‘modularity-in-production’ (MIP) from ‘modularity-in-design’(MID) and ‘modularity-in-use’ (MIU). A givenmodular boundary and interface specification mayachieve both MID and MIP; but often, these twoobjectives will point toward different boundaries/specifications, presenting trade-offs and requiringprioritization (see Fourcade and Midler, 2004, for anexample of a mismatch between MID and MIP thatcaused a product’s failure.) These conflicts are exac-erbated if there is not a unitary architect but rathermultiple individuals, departments, or organizationsdetermining design rules. Sako and Murray (1999)discuss MID, MIP, and MIU as the focus of optimi-zation, respectively, for product development, manu-facturing, and end users. According to Fixson(2006), designers often emphasize low functionalinteraction, producers easy installation, and userseasy disconnection. Thus, modules may not be opti-mizing modularity-as-property as a unitary technicalcharacteristic, but rather satisficing based on trade-offs among these desired optima or negotiationsamong various interests.

The relationship betweenmodularity-as-property andmodularization-as-processand the influence of a ‘modularity’ frame

Evolutionary perspectives on product and organiza-tional architectures highlight the crucial inter-relationship between modularity as technicalproperty and modularization as learning process.Products are often highly integral when first intro-duced, as early designs work through a set of com-plicated issues via coordination-intensive solutionsto unanticipated problems (Siggelkow, 2002); at thisstage, integral products can often command a pricepremium based on idiosyncratic features providingnew and unique functionality (Klepper, 1996). Asexperience with designing and building those prod-ucts accumulates at the focal firm (modularization-as-process), interdependencies across componentscan be better understood and minimized or antici-

pated in interface design (Adner and Levinthal,2001), advancing modularity-as-property. Eventu-ally, a dominant design emerges, allowing standard-ization and scale (Utterback, 1996) which, in turn,reduces costs, channels innovation efforts, and facili-tates the division of labor with specialized suppliers;then, as interfaces are more fully specified, supplierscan develop modules more independently. Thus,over time, industries often follow a trajectory frominitial integrality to high levels of modularity inproduct designs (Sanchez and Mahoney, 1996).

But product architecture need not evolve inevita-bly toward modularity. Technological change or cus-tomer demands for new functionality can redefinemodule boundaries, may increase interdependenciesacross modules, and can reverse maturation pro-cesses that lead to dominant designs (Abernathy,Clark, and Kantrow, 1984). Thereby modularizationprocesses can cause product designs to swing frommodularity toward integrality—and back again (Fineand Whitney, 1996). Undertaken strategically, as ameans to improve product performance, this reversalof direction can greatly advantage the architecturalinnovator; witness Shimano’s rise to dominance ofthe bicycle industry following its innovation of theindexing method of shifting gears (Fixson and Park,2008).

Modularization understood as evolutionaryprocess is also central to the literature on industryarchitecture (Jacobides and Winter, 2005; Jacobidesand Billinger, 2006; Jacobides, Knudsen, andAugier, 2006). As firms make choices about verticalscope, they are choosing what activities to manage asintegral, via nonroutine coordination of critical inter-dependencies, versus modular, via standardizedcoordination and independent innovation. Capabili-ties both inform and follow from these decisions.Firms are more likely to retain activities internallyfor which they possess superior capabilities, but theymay also decide to outsource an activity due tosuccess in mastering interdependencies and specify-ing interfaces. Subsequently, the firm will strengthencapabilities for retained activities and weaken or losecapabilities for outsourced activities, thereby rein-forcing, or changing, the industry’s architecture.

The ‘mirroring hypothesis’ posits a particularrelationship between modularity-as-property andmodularization-as-process, emphasizing how tech-nical characteristics of a product create a pull towardmatched organizational characteristics. From thisperspective, a module boundary’s ‘thin crossingpoint’ vis-à-vis technical interdependencies also

12 J. P. MacDuffie


provides the logical opportunity for organizationaldisaggregation, i.e., so that design task responsibili-ties can be allocated to suppliers with specializedexpertise (Hoetker, 2006; Cabigiosu and Camuffo,2011). Examples of mirroring include vertical inte-gration for design and production of crucially inter-dependent modules versus outsourcing for moreperipheral, easily separable modules (for a literaturereview, see Colfer and Baldwin, 2010).

The direction of causality in the mirroring of dif-ferent types of architecture is debated. The mostcommon view is that product architecture exerts apowerful force for isomorphism in organizationalarchitectures. But the reverse causality can also bemodeled, e.g., when tasks/activities are reallocatedacross organizational boundaries and patterns ofcommunication/interaction among people involvedin product design can change, with the ultimatepotential to change the product architecture.

While the mirroring hypothesis anticipates iso-morphism in product and organizational architec-ture, misalignment is not only possible, but may be astrategic choice. Task and knowledge boundarieswill not always coincide (Takeishi, 2001). Firms thathave historically integrated the components of acomplex product risk a competency trap if, fromoutsourcing, they lose their systems integrationcapability (Zirpoli and Becker, 2011). Thus, firmsthat no longer produce certain components may stillneed to retain the knowledge of how to make them;in the words of Brusoni et al. (2001), such firms needto ‘know more than they make.’ Indeed, given risksof imitation from modularity (Pil and Cohen, 2006),firms may benefit from preserving the interdepen-dencies of a near decomposable product design—even when more decomposition is possible—to maintain the tacit knowledge associated withmanaging those interdependencies (Ethiraj andLevinthal, 2004).

‘Modularity’ also provides a cognitive frame formanagers and engineers that drives a particulardynamic, and directionality, to the interplay ofmodularity-as-property and modularization-as-process. A cognitive frame simplifies informationprocessing (Gavetti and Levinthal, 2000) and pro-vides a stable basis for understanding environmentalchanges (Bingham and Kahl, forthcoming); it affectsattention, influencing both automatic and intentionalinformation processing (Ocasio, 2011); it provides ameans of interpreting something that is completelynew (Tripsas and Gavetti, 2000); it supplies meta-phors, symbols, and cognitive cues, ‘accenting and

highlighting some issues, events, or beliefs as beingmore salient than others’ and suggesting possibleways to respond to them (Benford and Snow, 2000:623); and it is used by organizational actors to givemeaning and legitimacy to their actions (Feldmanand Orlikowski, 2011). When strategies are takingshape or changing, internal debates often take theform of framing contests (Kaplan, 2008) in whichproponents of different positions seek to make theirindividual frame into a collective frame throughinfluence activities. Thus, the concept of modularitycan provide a powerful cognitive frame for managersrethinking which activities the focal firm shouldconduct internally and which can usefully be allo-cated outside the firm to a specialized supplier, andadvocating accordingly.

The cognitive frame of ‘modularity’ will reflectthe goals of the senior leader, or dominant coalition,but will be subject to contestation from intereststhat have different goals or priorities. Furthermore,either modularity-as-property or modularization-as-process can provide the primary or initial orientationfor the frame, with an implied sequence and goal.

Summary

I have argued that modularity-as-property is distinctfrom modularization-as-process. The latter is ubiq-uitous, linked to ongoing learning about how bestto manage interdependencies across and withinmodules. Modularization processes can, either delib-erately or unintentionally, however, lead to greatermodularity-as-property or to greater integrality, andthis can be true for either product or organizationalarchitecture. The cognitive frame of ‘modularity’helps shape the inter-relationship and dynamics ofchange between these two aspects. Next, I describethe data that informs my analysis of how the threeaspects of modularity interacted in the global auto-motive industry.

RESEARCH METHODOLOGY ANDSAMPLING APPROACH

The industry context and case study data I willpresent were gathered under auspices of the Interna-tional Motor Vehicle Program (IMVP) through amulti-year project exploring modularity and out-sourcing trends at automakers and suppliers world-wide. This project provoked the collective curiosityof a globally distributed team of researchers in the



IMVP network (named in the acknowledgementsand cited as appropriate throughout) who wanted tounderstand more about modularity, which was thenfrequently discussed as a factor in the rapid growth,technological innovation, and competitive volatilityof the computer industry. Earlier research on productarchitecture (e.g., Ulrich, 1995) already identifiedthe automobile as relatively ‘integral’ in contrastwith modular products such as bicycles and personalcomputers. While we did not have a precise way tocharacterize the two industries in relation to all otherindustries, we could assert with confidence that theywere separated widely apart along a spectrumbetween modular and integral endpoints.

Thus, these industries could serve as ‘criticalcases’ or ‘polar types’ (Eisenhardt and Graebner,2007), with theoretical sampling allowing predic-tions about the applicability and diffusion of modu-larity (Yin, 1994). Automobiles are exemplars ofcomplex products that can be analyzed at many dif-ferent levels, from basic parts and components tosubassemblies and functional systems. The autoindustry is also global in scope, with multinationalfirms competing in multiple markets and managingelaborate supply chains spanning developed anddeveloping countries; its outsourcing decisionsinvolve choices among geographically diverse loca-tions and, hence, cast a light on the relationshipbetween modularity and decisions on whether pro-duction and design activities need to be proximate orcan be globally dispersed.

At the time the data were collected, the computerindustry was covered far more extensively inwritings about modularity than any other industry(e.g., Langlois and Robertson, 1992; Garud andKumaraswamy, 1995; Sanchez and Mahoney, 1996;Baldwin and Clark, 2000). By giving the automotiveindustry an equivalent in-depth investigation, wehoped to challenge facile assumptions that all indus-tries are destined to evolve toward modularity or thatmodular product architecture would inevitably leadto modular organizational and industry architectures.By examining modularity initiatives over time, wesought to uncover the longitudinal, evolutionarydynamics among different aspects of modularity—our focus was on property and process—rather thaninferring causality from cross-sectional data. (Myemphasis here on ‘modularity-as-frame’ emergedonly later.)

The advantage of studying a single industrycontext with multiple firm-level case studies was theability to hold constant factors that are common for

the industry while highlighting factors varying at aregional or firm level. Furthermore, we could seethat modularity initiatives varied with respect to firmstrategy. Thus, we specifically chose to study modu-larity initiatives at firms that had certain similarities(e.g., a common starting point of production-basedmodule definitions, a common strategic goal oflinking the outsourcing of large and complex subas-semblies to the pursuit of modularity), but differed inthe extent to which the firms were able to sustaintheir modularization initiatives over time and accom-plish design as well as production gains. Thus, wecould triangulate among the common industrycontext, the common features of the modularity ini-tiatives, regional differences affecting policies onoutsourcing, and firm-level differences in out-sourcing and modularity strategy to gain analyticalleverage.

The global research team undertook several dif-ferent subprojects, collecting questionnaire data inEurope and Japan and doing interview-based casestudies in the U.S., Europe, Latin America, and Asia.For the case studies, two or more researchers were,most often, present at each interview. Extensive fieldnotes were taken; at the request of the respondents,no recording of interviews was done. Interviewswere often augmented by plant visits to see moduleproduction or by viewings of module prototypedesigns. Researchers also wrote up and exchangedfield notes. Quotes were recorded as close to verba-tim as possible; identities of individual respondentshave been disguised.

MODULARITY INITIATIVES IN THEGLOBAL AUTOMOTIVE INDUSTRY

Idiosyncratic definition of an automotivemodule—historical account

The automobile is a complex product, made up ofmany components and many technologies; as such, itis impossible to characterize as ‘modular’ or ‘inte-gral’ in its entirety. Auto industry managers andengineers began to pay attention to modular con-cepts in the early 1990s, following an earlier logic ofunbundling production activities to be carried out bysuppliers. The industry’s definition of a ‘module’—alarge chunk of physically adjacent components pro-duced as a subassembly by a supplier and theninstalled in a single step in an automaker’s (knownas original equipment manufacturer, or OEM)

14 J. P. MacDuffie


assembly plant—was idiosyncratic from the start;examples are the instrument panel; the front end;seats; and the rolling chassis. By the late 1990s,‘modularity’ had drawn the attention of seniorexecutives and was supported by a much broaderstrategic and financial rationale. The benefits ofmodular design were now sought, but still within theearlier production-based definition. Here I describethe emergence of this definition and its conse-quences, drawing on the IMVP longitudinal researchproject.

Separation of subassemblies

Beginning in the 1970s, the first step in separatingtasks to achieve more modular production was tomove work off the assembly line to subassemblylines in the same plant. Instead of working awk-wardly inside the vehicle body, each task could occurwithout obstruction on a physically separate, opti-mized subassembly line, where subassemblies couldalso be tested before being installed into the vehicle.The benefits were in greater flexibility, improvedergonomics, and better quality; labor cost savingswere minimal since employees on the subassemblyline typically received the same wages as those onthe assembly line.

Seats—the first outsourced ‘module’

Starting in the 1980s, U.S. OEMs began to ask sup-pliers to produce these same subassemblies, leverag-ing the lower wages at (often nonunion) supplierplants. Seats were the first subassembly to be out-sourced. Making seats is a mix of labor intensiveoperations (sewing covers for seat cushions andstuffing them with foam) and capital intensive opera-tions (e.g., frame construction, electronic seat con-trols). Seats are physically bulky with a high unitcost, plus variable content due to customer-chosenoptions, so it is not practical to hold a large inven-tory. OEMs awarded contracts for entire seats to asingle supplier, requiring them to establish dedicatedplants within a 30- to 60-minute delivery range andto produce seats just-in-time for a specific vehicle,delivered in exact match to the OEM’s assemblysequence. Seats were, thus, the first module (thoughthis name wasn’t used until later), carried out bysuppliers such as Johnson Controls Inc. (JCI), Lear,and Magna. This approach soon spread to Europe,Japan, and Korea, and by the mid-1990s, no OEMswere still making seats.

More modules defined and outsourced

By the early 1990s, OEMs extended this approach toother subassemblies, still oriented toward productionof large ‘chunks’ of physically proximate compo-nents, e.g., the instrument panel—that had previ-ously been produced in their assembly plants. Newmodules were defined, following the logic of com-bining bulky or heavy parts, e.g., the front-endmodule, which includes the front bumper and a frontcarrier section holding the radiator, cooling fan, andair conditioner condenser, as well as headlamps,airbag sensors, and associated wire harnesses. Notall such modules were transferred to suppliers; doorscontinued to be built in assembly plants due to thenecessary coordination with the paint process.

‘Modular factory’ as a new organizational concept

In the early 1990s, GM purchasing head IgnacioLopez de Arriortua declared his goal of building a‘modular factory’ to achieve assembly plant produc-tivity of less than 10 hours per vehicle, surpassingthe world benchmark. Lopez abruptly left GM forVolkswagen in 1993 to implement this strategicvision. VW’s first experiment was at a greenfieldbus/truck plant in Resende, Brazil. Resende’s‘modular consortium’ brought together key suppliersat the common site, under a single roof. VW retainedownership of all assets — land, buildings, machineryand equipment, inventories —so the outsourcing oftasks was not accompanied by an outsourcing ofassets (Sako, 2009). Supplier employees built eachmodule within the Resende plant and installed it onthe final assembly line. VW purchased the module atthe moment of installation; otherwise its role at thesite was system integrator, quality inspector, andadministrator of human resource policies for allemployees at the site (Lung et al., 1999). Subsequentnotable experiments in Brazil included a greenfieldGM assembly plant at Gravatai (named ‘BlueMacao,’ after birds that mate for life) and a green-field Ford plant at Camacari, in the Amazon; at both,suppliers were co-located and built large subassem-blies for in-sequence delivery. Brazil attracted theseinitiatives because of government subsidies forco-location, suppliers willing to build dedicatedfacilities tied to a single OEM, and flexible laborarrangements.

Global extensions: the ‘supplier park’ concept

The concept of ‘supplier park,’ like that of ‘modularconsortium,’ refers to the co-location of suppliers in



close proximity to an OEM assembly plant. Soonafter Resende opened in Brazil, supplier parks beganto spread throughout Europe. OEMs found supplierproximity an appealing way to implement lean pro-duction concepts of minimal inventory. Here OEMswere often retrofitting existing assembly plants tobuild new models and were able to attract govern-ment subsidies for consolidating automotive jobs at asupplier park. (The Mercedes-Benz assembly plantbuilt in Hambach, France, to manufacture the Smartmini car is a greenfield exception.) Suppliers ownedtheir factories and equipment and leased land fromthe OEM or from the government. A strategic movetoward modularity was another stated rationale forthese initiatives (Sako, 2003).

Deverticalization intensifies

By the late 1990s, based on their experiences withoutsourcing of large subassemblies, automotiveOEMs increasingly used the idea of modularity as arationale to move away from vertical integration forcomponents. This was not a new trend—the U.S.industry reached its peak of vertical integration inthe mid-1950s and has declined ever since, and theJapanese industry was never highly vertically inte-grated. But deverticalization accelerated during thisperiod based on arguments that it was necessary tosupport ‘modularity.’ Ideas of ‘core competence’were gaining influence in OEM strategy officesaround the world. From this perspective, an OEMshould not try to maintain excellence in making allcomponents when it could rely instead on the spe-cialized competence of ‘best in class’ suppliers. Thisfacilitated adoption of the production definition of‘module.’

Outsourcing of design

Once this idea took hold for production, it was ashort step to thinking about outsourcing designresponsibilities. Japanese automakers alreadyworked closely with their Japanese suppliers ondesign tasks, although not in explicit pursuit ofmodularity-as-property. U.S. and European OEMsnow also sought the allocation of design tasks tosuppliers, under the frame of ‘module design,’ to taptheir specialized knowledge. Suppliers welcomedthese overtures (and sometimes initiated them),seeing modules as a means to take on higher value-added activities (Whitford and Zirpoli, 2012).

Thus, increasingly, the auto industry interestin modules went beyond the idea of outsourcing

production to suppliers to more ambitious notions ofa new division of labor and new methods of coordi-nation. Executives and engineers were stimulatedby the powerful example of the computer and elec-tronics industries, where modular design providedpowerful combined benefits of cost reduction andspeedy, independent innovation (Sturgeon, 2002;Berger, 2006). Automotive OEMs envisioned aworld in which they could turn to suppliers for inno-vative design advances in modules while reducingtheir own in-house product development staff; wherecoordination complexity was reduced by the stan-dardization of module specifications and interfaces;where responsibilities for capital investment, qualityverification, logistics, and product liability could beshifted to suppliers; and where they could reducetheir heavy load of assets and become more agile,nimble companies with improved ROAs (Sako andWarburton, 2002).

These ambitions first had to face the reality thatautomotive modules are defined within a closed (i.e.,proprietary to a given firm and its suppliers) ratherthan an open (i.e., industry-level standards) architec-ture (Takeishi and Fujimoto, 2001; Garud andKumaraswamy,1995). Open modular architecturemakes it possible for any capable supplier to builda given module, while closed modular productarchitecture needs to be supported either by a fullyvertically integrated firm or by ‘black box’ subcon-tracting, where the customer provides proprietaryspecifications to a chosen supplier.

Creation of megasuppliers

As OEMs sought to outsource more modules tosuppliers, moving beyond seats, they found fewsuppliers capable of taking on the full set of designand production responsibilities. Industry analystsand investors predicted a necessary consolidation tocreate suppliers capable of handling this incipientdemand. The pace of deverticalization increaseddramatically at U.S. OEMs when first GM and thenFord spun off their captive parts divisions, creatingDelphi in 1998 and Visteon in 1999, respectively, asthe first ‘megasuppliers.’ More megasuppliers weresoon formed from horizontal merger and acquisi-tions of existing automotive suppliers to competefor module contracts; these firms were expected totake over critical design and engineering tasks, tohandle more complex manufacturing and logisticstasks, and to assume a larger role in the manage-ment of second- and third-tier suppliers. European

16 J. P. MacDuffie


megasuppliers included Faurecia, Valeo, andSommer Allibert in France and Bosch and Siemensin Germany; North American megasuppliersincluded Johnson Controls, Lear, and Magna.

Analysts greeted these developments with furtherpredictions of computer industry-like dynamics. ABain and Company report (Donovan, 1999: 6) wroteabout ‘the dawn of the megasupplier,’ predicting thatthey would quickly ‘design vehicle systems that canbe standardized within and across OEMs—in otherwords, used in multiple models of an OEM andeventually by multiple OEMs.’ Fine (1999: 62) pre-dicted that Chrysler, in taking the lead in outsourcingdesign work to megasuppliers, would be:

‘. . . the Compaq of the auto industry. Just asCompaq helped to drive the entire computerindustry to a horizontal/modular structure,Chrysler’s strategy allows suppliers—even Ford’sand GM’s suppliers—to strengthen their capabilityto develop whole automotive subsystems, therebypushing the entire structure of the industry fromvertical toward horizontal.’

Regional variation in modularity emphasis

Beyond Brazil and Europe, the emphasis on modu-larity in other regions varied widely. GM had plansto bring the ‘Blue Macao’ concept from Brazil to theU.S., under the code name ‘Project Yellowstone.’The United Auto Workers opposed the plan as a ployto increase outsourcing to nonunion suppliers, soGM backed away from Project Yellowstone per se,but nevertheless continued to outsource aggressively.For example, it moved responsibility for full interi-ors to Magna and the GM engineers supporting inte-riors became Magna employees. Ford, building onits experience in Europe, was the first to introduce asupplier park in the U.S. during the retrofitting of itsChicago assembly plant. Chrysler, historically lessvertically integrated than GM and Ford, was the firstof the Big Three to outsource both design and pro-duction to suppliers. Overall in the U.S., the pursuitof modular production provided the rationale foraccelerating outsourcing already underway at thecomponent level. Ford’s modularity initiative iscovered later.

In Japan, Toyota and Honda showed little interestin modules, believing they were already gaining thebenefits of design and production collaboration withtheir suppliers without any shift in product architec-ture, and that the pursuit of modules could lead toquality problems. These companies did, however,

create more subassembly lines within their assemblyplants, following a modularization process ofseparating once-bundled activities (Takeishi andFujimoto, 2001). Nissan, during its 1999 to 2001‘revival plan,’ restructured its relationship withkeiretsu suppliers substantially, selling equity stakesand redefining contract terms. One newly indepen-dent supplier, Calsonic Kansei, started doing moduleassembly inside Nissan assembly plants, a version ofthe modular consortium.

In Korea, dramatic restructuring in the late 1990sbrought Hyundai’s acquisition of Kia; their verti-cally integrated suppliers came together as HyundaiMobis, which was then spun off. As Hyundai set outto improve its brand image and quality reputation, itsmanufacturing strategy revolved around an extensiveuse of modules produced by Mobis; this case will becovered in more detail later.

Analysis

The historical account (summarized in Table 1)shows that the concept of ‘modularity’ was not ini-tially evoked when the movement of productiontasks off the main assembly line began. The initialframe was closer to a ‘focused factory’ idea(Skinner, 1974) in which specialized lines would bematched with specialized products. The frame ofoutsourcing was next, as OEMs sought to lower theirlabor costs and, at this point, the idiosyncraticproduction-based definition of an automotive‘module’ became well established. This definitionthen affected subsequent attempts to pursue designgains that became the heart of the strategic rationalefor ‘modularity’ at both OEMs and megasuppliers.Knowing that modules were already defined and inproduction before the frame of ‘modularity’ sweptthrough the auto industry is essential to interpretingthe subsequent dynamics of modularity initiatives.

Also crucial is a grounded understanding ofexactly how a ‘module’ is defined differently in theauto industry than in the information technologyworld (Sako, 2003). As noted earlier, automotiveusage defines a ‘module’ as a chunk of physicallyproximate components that are subassembled inde-pendently from the rest of the vehicle, tested forfunctionality, and installed in a single step in finalassembly. Taken strictly on these production-basedterms, modularity did increase in the automotiveindustry in the 1990s.

However, these automotive modules clearlyviolate the Ulrich definition. More than one function



Table 1. Emergent production-based definition of automotive modules

Event/activity Timeframe Goals/advantages Consequences for moduledefinition

Separate production(moving work frommain assembly line tosubassembly lines insame plant), stillvertically integrated

Beginning in1970s

Subassembly tasks physicallyeasier to do, ergonomicallybetter, logistics more efficient;quality testing done beforeinstallation; labor flexibility.Frees up space, reducescomplexity on main assemblyline.

Logic of which subassembliesbecome ‘modules’ becomesclear. What is moved tosubassembly line is ‘chunk’ oflarge, heavy, bulky, physicallyproximate components, forwhich production advantages ofseparation are greatest. Firstdone for seats.

Separate production,outsourced toproximate suppliers

Begins in late1980s

Take advantage of lower (oftennonunion) supplier labor;supplier has delivery and qualityresponsibility.

‘Chunks’ begin to be called‘modules.’ Emphasis on supplierbeing able to produce anddeliver in time for in-sequenceproduction. Module boundary/interface designed for one-stepinstallation at assembly plant.

Experiments with‘modular factory’

Early tomid-1990s

On-site suppliers coordinateclosely with OEM and eachother; take on moreresponsibility for quality,logistics.

‘Module’ is now central conceptfor this production experiment.More production-definedmodules established; e.g., frontend, instrument panel, andchassis, as well as seats.

Diffusion of supplierparks

Mid- to late1990s

OEMs improve logisticalefficiency. Supplier proximitysupports in-sequence assembly.OEM keeps integrator role.

Production-based division of laborbecomes well established.Suppliers develop mastery inproduction of modules, begin toseek design role.

Deverticalization byOEMs intensifies

Late 1990s OEMs seek more agility, ability toaccess new suppliers, increasein competitiveness of spun-offparts divisions.

Spin-off of Delphi (from GM);Visteon (Ford); Mobis(Hyundai) changes relationshipwith parent OEMs; spin-offsseek more independence, moreautonomy, bigger role indesigning modules.

Creation ofmegasuppliers

Late 1990s toearly 2000s

Pursuit of scale; amassingcapabilities to support OEMglobalization, take on highervalue-added production anddesign tasks vis-à-vis modules.

Megasuppliers created aroundstrategy of building modules,based on production definitionthat brings together componentsfrom many functionalsubsystems. Rationale is tohorizontally combine theexpertise needed to build anddesign those modules. Designgains difficult to achieve,knowledge doesn’t encompassall functions included in amodule.

18 J. P. MacDuffie


is mapped to the chunk; Ulrich would take this‘many-to-one’ characteristic as evidence of integral-ity. Furthermore, automotive modules, like theoverall product, have a closed rather than open archi-tecture; designs are OEM specific, with no standard-ization of modules at the industry level. Indeed, thereis no standardization of modules even at the firmlevel. Module boundaries (including which com-ponents are included) typically differ across dif-ferent models in an OEM’s product line (Sako andWarburton, 2002).

The organizational architecture associated withautomotive modules was also optimized for produc-tion. Production tasks were first disaggregated butkept within firm boundaries, then moved outside thefirm to take advantage of lower wages and to draw on(and/or develop) the specialized production expertiseof suppliers. Organizing for supplier proximity, eitherin a ‘modular consortium’ or a ‘supplier park,’ waspursued to reduce inventories and improve coordina-tion efficiency. None of these organizational arrange-ments furthered design improvements in modules.

Indeed, the prioritization of production goalscreated barriers to making designs more modular.Consider the instrument panel module (seeFigure 1). Functional interdependence is high, giventhe tight intermingling of components that supportsteering, climate control, entertainment, driver infor-mation, and safety. But interdependence is also highacross the module boundary, since most of these

functions require components elsewhere in thevehicle—e.g., in the body, electrical system, trans-mission, cooling system—to be operable. A lot ofdesign coordination is required to optimize its fit to aspecific vehicle’s space constraints, driving charac-teristics, and market positioning.

Furthermore, emergent phenomena such as NVH(noise, vibration, and harshness), either within oracross modules, can be modeled but not fully antici-pated or assessed before the final vehicle is com-plete. In the absence of standardization across firms,or even models, suppliers can’t gain advantages ofdesign specialization and scale that would obtainfrom a fully specified, industry standardized module.

As OEMs defined more modules and sought tooutsource their production, their need for morespecialized supplier expertise grew. For example,megasupplier Sommer Allibert was created in amerger intended to facilitate one-stop access to spe-cialized expertise in electronics and plastics for out-sourced production of instrument panels. Since itsexpertise did not encompass all the functions con-tained in an instrument panel, Sommer Allibert wasnot immediately well positioned to achieve designgains, although it had the potential to amass knowl-edge over time about interdependencies within andacross the instrument panel boundary through modu-larization processes.

Even with this increase in outsourcing tomegasuppliers (the extent of which varied by

Figure 1. Instrument panel (cockpit) as ‘module’



module, see Fixson, Ro, and Liker, 2005), OEMsretained a central role in design engineering, productdevelopment coordination, and system integration.Indeed, the difficulties megasuppliers had in pursu-ing modular designs caused them to engage in inten-sive coordination with OEMs rather than innovatingautonomously as expected under the ‘modularity’frame. Modularization-as-process within megasup-pliers and in coordination with OEMs did lead(in some, but not all, cases) to slow, incrementalincreases in design modularity. However, even thesemodest changes were the ex post outcome ofintegrative organizational processes, rather than anex ante condition that facilitated reduced coordina-tion, consistent with Cabigiosu and Camuffo’s(2011) findings in the air conditioning industry.

Overall, the barriers to achieving design modular-ity, among other factors, prevented the anticipatedmigration of value from OEMs to suppliers and thepredicted shift from vertical to horizontal structure.The computer industry analogy simply did not hold.

Thus, it is no surprise that the ambitious strategicgoals of automotive executives for moving towardgreater modularity-as-property could only be par-tially met. Attainable were the goals of movingasset-intensive activities to suppliers; tapping sup-plier production expertise; reducing manufacturingcosts; and reducing complexity on the main assem-bly line. Other design goals could not be readilyattained. Coordination costs during design and pro-curement were often higher, not lower; standardiza-tion of modules at an industry level was notsupported by any OEM; and innovation was stillconstrained by the tight design interdependenceamong suppliers and with the OEM.

The ‘modularity’ frame did not help executivesunderstand this misalignment; rather it raised expec-tations that gains seen in industries with moremodular product architectures, e.g., computers,could be attained for the auto industry, too. Theresulting frustration was a major contributor to whathappened with firm-specific modularity initiatives.We turn now to examine two such initiatives indetail.

Modularity at Ford Motor Company:production definitions that constraindesign ambitions

This case study explores the modularity initiative atFord Motor Company from three perspectives: (1)the corporate-level executives, managers, and staff

who mobilized the initiative from the start; (2) asupplier’s product development team working on amodule design to propose to Ford; and (3) a Fordchief engineer and his team of product developmentengineers who were making decisions about whetheror not to incorporate modules in their current vehicleproject. We started data collection with corporate-level interviews in 1999; we continued with supplierinterviews in 2000 and Ford product developmentteam interviews in 2001; and we returned tocorporate-level respondents, also in 2001, for anupdate.

Antecedents and initial steps

Ford was one of the first OEMs to implementsupplier parks extensively, particularly in Europe.Accordingly, Ford had experience with theproduction-dominated module definitions and out-sourced organizational arrangements with suppliersthat manufactured seat, instrument panel, and front-end modules in nearby plants.

Suppliers sought a broader design role at a module(rather than component) level and Ford was not orga-nized to respond to this request, prompting themodularity initiative. An engineer told us:

‘We have integrated suppliers very well into the com-ponent approach to product development. It’s hard totell who’s a Ford person and who’s a supplier personon the platform teams. But we don’t know how toevaluate a module proposal because we’re organizedaround components. It’s difficult to figure out if sup-pliers have all the capabilities we think they need formodules.’

To increase its ability to evaluate module propos-als, Ford established a corporate-level modularitytask force with cochairs from product developmentand manufacturing plus representatives from corpo-rate strategy and purchasing. After internal study andexternal benchmarking, the task force identified 19modules making up the entire vehicle, usingproduction-based definitions consistent with indus-try usage at the time; according to task forcedocuments, a module was defined as ‘assembly of asignificant number of components designed andoptimized as a subassembly before installation onthe vehicle on the main assembly line’ (Ford MotorCompany, 1999). Also stated in these documentswere goals for the shift to modules: reducing thenumber of parts; defining interfaces with adjoiningcomponents; setting common ‘hard points’ so a

20 J. P. MacDuffie


module could be positioned correctly for automatedsections of the production line; testing functionalquality before final assembly; making installationeasier; and improving logistics. The 19 modules(called Class 1) were later subdivided into submod-ules (Class 2 and Class 3), totaling 44 in number.Having submodules, it was believed, would reducethe barrier to adoption of modules by product devel-opment teams within Ford; vehicle projects couldease into modularity by adopting some Class 2 andClass 3 submodules and then move up to Class 1 fullmodules in the next product cycle.

Despite the production-based definition of moduleboundaries, the ambitions of the modularity initia-tive were clearly to leverage the benefits of designmodularity. The CEO’s aim was to restructure thedivision of labor between the OEM and the supplybase. According to an executive on the task force:

‘[The CEO] sees modularity as a way to share capitalrisk and assets with the supply base, so we don’t haveto be so big. Plus, modules should get us to the nextlevel of design innovation. We want to design andbuild a vehicle quicker, better, cheaper, closer towhat customers want. If we do modules right, it willchange how we think about building the vehicle. Itwill put the expertise in the supply base, where itshould be.’

Ford executives had seen these design-derivedbenefits enacted in the computer industry andwanted to reach beyond the already underway out-sourcing of production for the small number of well-established modules. By defining the entire vehiclein terms of 19 modules, the task force presentedFord’s organization with multiple challenges. First,Ford had not been granting design responsibilities tosuppliers on a modular level of aggregation; supplierinvolvement with product development had beentaking place within Ford’s product developmentteams, often with co-location and at a componentlevel. Second, many of the 19 modules were stillmanufactured entirely within Ford, so now theissue arose of whether suppliers were capable ofmanufacturing them efficiently.

Ford proponents of modularity cited both techno-logical and organizational advantages of givingdesign responsibility to suppliers. Some focused onthe prospects for more innovation, plus cost savings,from outsourcing design to suppliers. According toone engineer, ‘We should simply give suppliers atarget price and say ‘you figure out how to make it.‘

If we keep doing all the engineering, we don’t saveanything.’ Other managers emphasized accountabil-ity and the reduced coordination requirements fromassigning an entire module, e.g., seats, to a singlesupplier, as opposed to the past practice of dealingwith many suppliers providing their respective partsto Ford’s seat manufacturing line at the assemblyplant.

Yet many Ford engineers saw risks to productperformance and brand identity in outsourcing toomuch to suppliers. One engineer said, ‘We’ve got tobe clear on what makes a Ford a Ford. We can’t giveaway control of those things.’ Another engineeragreed: ‘Most modules affect look and feel.’

At the same time, there was recognition thatFord’s capability for module design was limitedand needed development. A product developmentmanager said, ‘Our first intention has been to designin-house a generic module design and then give it tocapable suppliers. But we haven’t been so goodabout designing for modularity. Our engineers havea component mentality.’

Ford’s managers were aware that the shift tomodules would put many new demands on suppliers,beyond the issue of their design capabilities. Accord-ing to one manager, ‘We need to certify that a sup-plier is module ready. They will need lots of projectmanagement skills. We expect them to manage Tier2 suppliers. We will want them to test modules thesame way we would.’

For those focused on outsourcing the manufactur-ing of more modules to suppliers, a different set ofconcerns—about cost—arose. As one product engi-neer explained, ‘It may not work to take our designand build it somewhere else. There could be a verygood rationale for keeping it inside.’According to oneengineer, ‘We’re not really optimizing the design of amodule—often we’re just seeing a subassembly thatwe used to build now being pushed out to a supplier.The freight costs can be very high unless the supplierlocates nearby.’ Many felt that design gains, notmanufacturing savings, were the key to modularitybeing a success: ‘Most modules we’ve checked outare cost plus; it costs the supplier more to make thanfor us to do it. We need more design improvementsfrom suppliers to justify doing modules.’

From these differing viewpoints, either Ford or itssuppliers could be the preferred location for eitherdesign or manufacturing, with the net benefit depen-dent on how successfully each mastered its assignedtasks. While each interview subject expressed his/her own preference for task allocation (e.g., design



by Ford, design by supplier), it seemed likely thatchoices, at least at this initial stage, would be con-tingent, varying by which newly defined module wasunder consideration.

Yet the corporate imperative behind the modularityinitiative was more absolutist. According to oneexecutive, ‘I don’t see anything that we’re not willingto outsource. We should focus on system engineeringand common standards.’ This message encouragedthe ambitions of Ford’s suppliers, who were eager totake on full responsibility for module design.

Visteon develops a new modular design

Our interviews in 2000 focused on Visteon (thefirst tier supplier that spun-off from Ford) and,as a specific example, its modular design for asuper-integrated instrument panel, known as SI-IP.Visteon’s strategy was to emphasize its capability tobe the system integrator. As the chief engineer forSI-IP told us, ‘We always did the full integration forFord in the past [when vertically integrated], so wefully have that capability. Most electrical suppliersdon’t have that breadth.’ Another SI-IP engineerexpressed his ambition to ‘break down barriersbetween the modular design approach in electronicsand the more integrated design approach in autos.’Visteon invested heavily in advanced technology forthe SI-IP, filing 100 patents and 500 disclosures.

The traditional IP is built around a beam of steelthat bolts across the back of the engine compartment.Plastic moldings, which provide the IP’s shape andsurface texture and define cavities for componentsand storage (i.e., glove box), are then attached.Various components are then assembled into thestructure, including instrument cluster (speedometer,gauges), HVAC controls (heating, ventilation, airconditioning), audio system, and at times, steeringcolumn and brake pedal. A traditional instrumentpanel is a tangled mass of clunky metal boxes weigh-ing 80 to 100 pounds. The bulk and weight wereconsequences of component-based design. Eachcomponent had its own protective housing, heat sink,and microprocessor. With many separate compo-nents, assembling the IP involved many electricalconnections, each a potential source of quality defectsand ergonomic problems for assembly workers.

Visteon’s SI-IP design aimed to move away fromthis component-based approach. Using a singlecasing, with structural metal embedded in the plasticmolding, reduced size and weight substantially. Inte-grating the housing for a fan motor into this casing

allowed for a smaller and lighter motor. Electricalfunctions were consolidated onto a small number ofintegrated circuit boards. The module was designedwith a hinge along the back edge, allowing the tophalf of the IP to open up to facilitate replacement ofboards and installation of software. Based on con-sumer electronics designs, engineers chose a singlemotherboard designed for longtime functionality andput electrical functions that might be upgraded morefrequently on separate boards. In place of wire har-nesses with bulky connectors, SI designers used a flatwire technology that allowed highly flexible posi-tioning of components and eliminated heat sinks.

The gains claimed by Visteon for the SI-IP designwere impressive—a 44 percent reduction in weight,a 30 percent reduction in the number of components,a 20 percent reduction in cost, a 10 percent reductionin NVH (noise, vibration, harshness), and a 30percent reduction in quality defects, plus much lowercubic feet requirements for the overall physicalpackage. At the same time, Visteon engineers wereaware of technical challenges facing SI-IP, given anautomobile’s demanding operating requirements:10+ year durability, high reliability in extreme tem-peratures, holding up under constant vibration, andresistance to oxidation, corrosion, and water. Butthey were optimistic that their design would stand upto Ford’s rigorous testing.

However, they worried about organizationalarrangements with Ford. First, Ford was losing thepotential cost benefits of having Visteon do themodule design by insisting on having its own engi-neers monitor—through ‘shadow engineering’—Visteon’s work, as the SI-IP chief engineerexplained:

‘OEM engineers are monitoring engineers at suppli-ers who are fully capable of doing the job. Each oneis afraid of having something bad happen on hiswatch . . . So the cost benefits of modules aren’tbeing realized.’

Second, they saw how Ford’s purchasingorganization— in structure, routines, and incentivesvis-à-vis risk tolerance—was not well suited toevaluating a module design proposal from a supplier:

‘When they have to evaluate something like this,their organization begins shaking. Purchasing wantsus to break everything down to the subcomponentlevel so they can call in their technical experts tocheck out the costs. Then either they can’t quiteassess the technical characteristics of the integrated

22 J. P. MacDuffie


design, which makes them nervous, or they see thatit’s great, which is threatening . . . Ultimately, even ifthey like the design and improvements we offer, theydon’t want to take the risk of the whole package.’

Ford’s reaction to Visteon’s modular design

In 2001, we spoke with the development team for aproduct being redesigned for the 2005 model yearabout how they thought about utilizing Ford’sdefined modules. The Ford project team had evalu-ated the super-integrated instrument panel design wehad studied at Visteon. As the engineer involved inthat evaluation explained:

‘The SI-IP from Visteon was a great concept. Wewould have gotten a huge glove box— it would havesold the car. I was willing to take a $5 to $10 penaltyon it. But it didn’t work. The ribbon wiring didn’tmeet our temperature or vibration requirements. TheIC cards installed on their edge also had a big vibra-tion problem—laptop computers don’t use them forthat reason. The idea of opening the instrument panelto put in new cards for programming or new function-ality was great, but with the windshield there, howcould you do it? And even though SI-IP passed theinitial tests, it failed all the extreme condition tests.’

This engineer acknowledged that the SI-IP, as a newmodular design, faced an uphill battle amid the tighttime lines, quality tests, and cost targets of a typicalproduct development project:

‘Even if SI-IP had passed all the tests, you wouldworry a lot. You’d throw double engineers on it,really harass the suppliers to make sure everything isworking. It would cost a lot, you might get a pat onthe back, or you might not.’

The chief engineer described other examples ofwhy his team judged various modules to be unwork-able. With the ‘door inner’ module, Ford’s purchas-ing assessment concluded that there was no availablesupplier with all the capabilities needed to designand build the entire module. Furthermore, its heavyweight would mean having a supplier build a partialmodule with remaining manufacturing steps done onFord’s assembly line. The team judged the cost ofthis mixed approach to be too high.

The project team also considered using a front-end module (FEM). This would reduce weight by 20pounds, and quality would improve from installingthe FEM straight into the open end. But the currentassembly plant layout didn’t allow for installing theFEM in one step. This time, the project manager

decided, it would be better to build the front endfrom components, pending the plant’s retrofit.

While component-oriented purchasing routinesand factors driving up manufacturing cost weremajor barriers, knowledge-related issues were agreater challenge, according to one engineer:

‘We know how to design a component (and how totell a supplier to design a component), but I’m notsure we have the systems engineering capabilities todesign a module. When you’re designing a whole setof components brought together in a module, youneed to understand every interface deeply.’

All in all, the chief engineer described a strikingreversal in their thinking during the project:

‘Our goal was to use all 19 modules . . . But in theend, after many false starts, we didn’t use a singlemodule.’

His ultimate conclusion was that:

‘There’s no point in modularity for its own sake. It’slike Don Quixote pursuing the impossible dream. Ican see why suppliers want to do it. It brings morevalue-added into their plants. But despite my bestintentions, I simply couldn’t justify doing it.’

Analysis

The Ford modularity initiative demonstrates howdifferently groups within an organization canframe ‘modularity’ and the consequences for bothmodularity-as-property and modularization-as-process. Tensions rose due to conflicts among theproduction-oriented definition adopted by the modu-larity task force, the design-oriented ambitions ofsenior executives, and the skepticism of engineersabout whether suppliers had the capabilities to meetthe new strategic vision.

As noted in the earlier historical account, theemergent production-based industry-wide definitionof module arose before the term ‘modularity’ hadbecome common currency, at Ford or at any auto-maker. Ford’s pursuit of modularity-as-property washeavily influenced by path dependence in howmodules came to be defined at the industry level aschunks of physically proximate components. Theexternal pressure from Ford’s suppliers, seeking theopportunity to bid on a full module rather than com-ponents, also affected the task force’s definitionalchoices and subsequent activities.



The action by the modularity task force to definethe entire vehicle in terms of 19 Class 1 modules ledalmost inevitably to a full set of production-baseddefinitions, given that certain modules were alreadywell established on that basis throughout the indus-try. These definitions also conveyed an idea of fixityabout module boundaries that was perhaps inappro-priately rigid at the early stage of a modularizationprocess. This confirms that the task force saw its taskas achieving modularity-as-property and gave littleexplicit attention to modularization-as-process; thisthen corresponded with production-oriented moduledefinitions that paid little attention to the potentialfor design gains.

Yet despite this orientation, ambitious goalsrelated to module design were being set from thestart. Senior executives were particularly likely toapply the analogy of the computer industry in pur-suing a strategy of modular design innovations gen-erated by specialized suppliers. The CEO’s strategicintent for modularity was to change the boundary ofthe firm, allocating more tasks, more investmentrequirements, and more risk to suppliers—essentially the ‘modularity’ frame as enacted in thecomputer industry and projected onto the auto indus-try. Particularly appealing was the idea of being ableto set a target price and then turn everything over toa capable and accountable supplier; this set up anexpectation that suppliers would work autonomouslyon modules, achieving lower costs and faster, betterinnovation.

Ford managers and corporate staff tended toaccept this framing as legitimate and warranted, cog-nitively adjusting to how this new strategy differedfrom the past. For example, whereas the earlier cog-nitive frame of ‘core competence’ had originallymeant retaining control of such key components asengines, the overlay of the ‘modularity’ frame nolonger required such choices. Ford was already out-sourcing engines and, in the words of one manager,‘I don’t see anything we’re not willing to outsource,’provided that Ford focused on ‘system engineeringand common standards.’

In contrast, Ford’s engineers tended to be morefocused on the pros and cons of achievingmodularity-as-property. The perceived advantagesincluded assigning full responsibility for cost,quality, and delivery to a single supplier, thus reduc-ing coordination requirements, simplifying procure-ment, and improving accountability. Their list ofdisadvantages was longer: difficulty in assessingsupplier module proposals due to knowledge at Ford

being organized around components; higher costsbecause of lesser supplier efficiency; loss of branddifferentiation because suppliers would have lesscommitment or ability to generate the Ford vehicle‘look and feel;’ loss of quality and reliability becauseof lesser project management and testing capabilitiesat suppliers; and redundancy in engineering effortbecause of Ford’s need to monitor suppliers. Notethat most of this list revolves around skepticism ofsupplier capabilities.

Ultimately, neither senior executives and manag-ers nor engineers explicitly addressed the centraldilemma—the not-very-modular-for-design natureof the defined modules, i.e., large interdependenciesexisting across module boundaries due to functionalsystems that spanned modules—that made decisionsabout whether design should stay at Ford or beoutsourced so difficult. Their debates, rather thanreflecting disagreements about the wisdom of out-sourcing (all respondents viewed an increase in out-sourcing as necessary and mostly advantageous),were instead addressing fundamental questionsabout how to set the appropriate knowledge and taskboundaries between OEM and supplier, given mul-tifaceted cost, quality, and product performancerequirements. Yet discussion of these issues was dis-torted by lack of recognition of the consequencesof the idiosyncratic path to a production-definedmodule.

When we interviewed Visteon’s engineers as theyworked on an innovative module design, essentiallymoving to adopt the new role proffered them by thestrategic vision of Ford’s CEO, we found they com-bined the managerial and engineering frames foundat Ford. Like Ford’s executives, they saw a poten-tially new division of labor and welcomed the oppor-tunity to be more autonomous, take more designleadership, and achieve higher margins. Yet theyshared many of the worries of Ford’s engineers,albeit with an understandable difference in perspec-tive; they felt Ford didn’t trust their capabilitiesenough, was too inclined to monitor them, andcouldn’t break out of its component mentality. Alsounderstandable is their belief that they could handlemodule design on their own, without much coordi-nation with Ford. Only by making this convictiona reality would they be able to escape from thesubsidiary and dependent position that automotivesuppliers were accustomed to occupying in theautomotive design process.

Ford’s pursuit of modularity-as-property signifi-cantly affected how both Ford and Visteon

24 J. P. MacDuffie


approached modularization-as-process. Ford definedmodule boundaries on its own and then engaged inheated internal debate about whether design respon-sibility for modules should or should not be trans-ferred to suppliers. This decision was oftendescribed in either/or, zero-sum terms, i.e., eitherFord would transfer responsibility to a supplier orwould keep the module in-house. Ultimately, Ford’smanagers favored supplier autonomy, while Ford’sengineers (and purchasing agents) feared it. Thisdebate also reflected the under-the-surface (andperhaps not entirely conscious) divergence inopinion about whether senior management’sframing of ‘modularity’ using the analogy of thecomputer industry was, in fact, applicable to the autoindustry.

Visteon’s engineers also approached the modular-ization process as an autonomous activity, believingthey had system integration capabilities superior toother suppliers. Indeed, they found ample opportu-nity, in their SI-IP design, to achieve reductions inweight and number of components, and they confi-dently predicted reduced costs, fewer defects, andgreater flexibility via software upgrading. Theychafed under what they saw as an OEM’s outmodedinsistence on control (manifested in demandsand oversight from Ford), which they felt con-strained them from increasing Visteon’s share ofvalue-added.

Explicit in this complaint was the idea that Fordshould be matching the new technical architecturewith a new organizational architecture. Unacknowl-edged in this anticipation of ‘mirroring’ was thatthis new way of organizing would require moreintegration of expertise, more interaction, and moreattention to interdependencies—both intra- andinterorganizationally—than in the long-establishedcomponent approach to design. This created tensiongiven expectations (shared by Ford and its suppliers)about the benefits of lower coordination, reducedcost, increased speed, and more innovative featuresthat would arise from granting capable suppliersautonomous control over module design.

When the chief engineer we interviewedexplained his reasons for not accepting any of themodule proposals submitted for his current project,his list for most modules was dominated byproduction-related concerns, e.g., additional auto-mation needed for one-step installation of the doorinner; insufficient space for installing the front-endmodule at the assembly plant. However, his objec-tions to the Visteon instrument panel proposal were

based on Ford’s assessment that the SI-IP hadserious design failings—e.g., not meeting tempera-ture or vibration requirements; not functioning reli-ably under extreme conditions; not feasible to openthe instrument panel to install new IC cards given thewindshield angle.

Recall that Visteon engineers criticized any Fordorganizational routines that interfered with theirability to move fully into module design. Yet the factthat the Ford chief engineer found so many problemswith Visteon’s instrument panel prototype hardlysuggests that the design would have been improvedby more supplier autonomy. Earlier and more fre-quent interaction between Ford’s engineers andVisteon engineers, understood not as excessivemonitoring but as a modularization process toincrease mutual learning about interdependenciesand interfaces, might have resulted in faster andbetter resolution of problems before the SI-IP designwas completed.

Those Ford organizational routines also held backthe Ford chief engineer. He worried about choosinga module that would inflate the costs of his currentproject, even if it paved the way for company-wideimplementation in the future. At the same time, hefelt putting lots of extra Ford engineers on the task ofassessing a supplier’s design proposal was the onlyway to mitigate his personal risk; since this wouldadd cost, a critical metric, he was even less likely tochoose a module.

So while the modularization process hypotheti-cally advances steadily toward a better understand-ing of interdependencies and, hence, more beneficialchoices of module boundaries and interfaces, theFord case reveals how that process can be blocked,interrupted, or misdirected. When expectations ofcost reductions and lessened coordination weren’timmediately met, the powerful imperatives of com-pleting a vehicle project on time and at budget(which drove the chief engineer’s incentives) over-rode corporate-level goals of increased adoption ofthe task force’s defined modules.

It seems likely that the overlap of task and knowl-edge boundaries (however much complexity thisadded), and even Ford’s ‘shadow engineering’(despite the risk of redundancy), were necessary toachieve design gains with production-definedmodules, certainly during a transition of movingfrom components to modules as the focal level ofanalysis. Neither party seemed to recognize this,despite (or perhaps because of) Visteon so recentlybeing a vertically integrated division within Ford.



Coda: reconsidering modularity strategy

In 2001, we returned to ask members of the originalmodularity task force about Ford’s modular strategy.A senior manufacturing executive emphasized howmuch time and cost and capability building would berequired to make modularity work and acknowl-edged that Ford had not understood this in imple-menting the strategy:

‘Doing this right requires a high level of skills onboth sides [OEM and supplier] and we’re justcoming to grips with that. We at Ford need to bebetter at the systems engineering than we are.Neither we nor the suppliers really understand howthe electronics in an instrument panel module need tointeract with the electrical system in the rest of thevehicle. Plus the suppliers need to understand awhole lot more about the customer, the warrantysystem, our dealers, etc. Right now, the suppliershave less engineering talent than we do, so we’rereluctant to pass the design to them. It’s easier to gowith components and take the “pass the screwdriver”approach . . . We built a great strategy and we’vebeen abysmal about implementing it. It’s still theright strategy and it still presents us with huge oppor-tunities. But there are lots of skeptics in the companynow because of how we’ve gone about this.’

The skeptics prevailed when Ford disbanded itsmodularity task force in 2001 and reduced its effortsto persuade chief engineers to adopt the definedmodules and submodules in their vehicle projects.Modules did not disappear by any means; front-endmodules and traditional IPs, built up as modules atsupplier plants, are visible to this day as one-stepinstallations in a number of Ford assembly plants.But the more ambitious goals for design gains frommodularity-as-property went by the wayside.1

To the end, the framing of modularity by Fordexecutives and corporate-level managers and staff

remained the same. Their emergent recognition wasthat Ford’s capabilities for system engineering ofmodules were not all that strong, that supplier capa-bilities for doing module-level design innovationwere also underdeveloped, and that both partiesneeded to do much more work to understand all ofthe interfaces. Thus, Ford’s framing perceivedincomplete achievement of modularity-as-propertyas a consequence of inadequate capabilities, ratherthan seeing modularization-as-process as a steeptechnical learning curve, given high interdependen-cies across production-defined module boundaries,requiring a patient, long-term commitment to seeingwhat design innovations might emerge over time.This latter view is better represented in the next case.

Reorganization for modules at HyundaiMotor Company

Hyundai Motor Company is arguably the automakermost heavily engaged in using modules to managecomplexity, improve quality, and reduce costs. Thiscontrasts with OEMs like Ford that have backedaway from modularity initiatives (even as modular-ization processes arguably continue there and at allautomakers). Hyundai’s approach to modularitydepends heavily on a unique relationship with itssole-source module supplier, the now-separatemegasupplier known as Mobis that was once a ver-tically integrated division within the huge Hyundaichaebol (conglomerate). Together, Hyundai andMobis evolved to an approach that stays fundamen-tally oriented toward production advantages frommodularity while also making slow but steadyprogress toward design gains—accomplished withinhighly integrated organizational arrangements thatsupport an ongoing modularization process.

History of Mobis: its relationship withHyundai Motor

Established in 1977 to make shipping containers,Hyundai Precision became a contract manufacturerfor Hyundai Motor in the 1990s. The financial crisisin Asia in 1997 forced Korean automakers Daewooand Kia into bankruptcy. Hyundai acquired Kia, andHyundai Motor was created as a freestandingcompany in 2000, separate from the rest of theHyundai chaebol. Meanwhile, Hyundai Precisionwas restructured to include just automotive partsbusinesses and was renamed ‘Mobis’ in 2000 afteracquiring parts operations from Hyundai and Kia.

1 The controversy of Firestone/Bridgestone tires and rolloveraccidents involving the Ford Explorer SUV was unfolding in2000 and 2001, with the first federal government alert in May2000, a Firestone recall in August 2000, and a further Fordrecall in the spring of 2001 after a severing of the 100+ yearrelationship between the two firms. As a result, CEO JacNasser, who had been the lead champion of the modularityinitiative, stepped down from his position in October 2001.Investigators did identify a distinctive interaction betweenvehicle design and tire design as the most likely cause for thehigh number of Explorer rollover accidents, bringing home theunusually high level of design interdependencies affecting auto-mobiles. This event, combined with Nasser’s departure,undoubtedly affected Ford’s inclination to continue its high-level pursuit of modularity-as-property for the entire vehicle.

26 J. P. MacDuffie


Moving toward modularity was an important strate-gic rationale for the restructuring. The initial productlineup consisted of three core automotive modules—chassis, instrument panel, and front end. Moduledefinitions fit the industry’s production-based norm.Hyundai Motor was initially the only customer.

Spinning off Mobis was in keeping with the trendof deverticalization in the auto industry worldwide.But from the point of view of structure and gover-nance, Mobis has certain unusual features. WhenMobis was created, the founder and chairman ofHyundai, Chung Mong-koo, insisted that theHyundai/Kia aftermarket parts business be includedso that its high margins could subsidize the newmodule business; as a result, Mobis’ R&D expendi-tures are higher than virtually all other Korean sup-pliers. Mobis is also the official holding company ofHyundai Motor, as well as its largest shareholder;plus the CEO and other senior executives at HyundaiMotor worked previously at Mobis. As a result, thefinancial and managerial relationship betweenMobis and Hyundai Motor is much closer than thecounterpart examples of Delphi (spun off from GM)and Visteon (spun off from Ford) in the U.S.

Hyundai saw many production advantages inhaving Mobis as a separate company. While HyundaiMotor’s plants are entirely unionized, Mobis is non-union and often relies on outside agencies to providetemporary or contract employees, allowing substan-tial wage differentials between Hyundai Motor andMobis plants in Korea and elsewhere. Mobis plantsare deliberately located close enough to Hyundai andKia assembly plants to provide frequent sequenceddelivery of modules and keep inventories low, whilealso being far enough away to draw on a differentlabor pool and reduce wage comparisons.

During Hyundai/Kia’s aggressive expansion inthe past decade, with new assembly plants in India,China (2), Slovakia, the U.S. (2), Turkey, and Russia,Mobis built new plants nearby, sometimes followingthe not-too-close siting logic and sometimes inclosely proximate supplier parks. Hyundai modeledits new assembly plants in China and India on itsnewest Korean plant at Asan Bay, and the supportingMobis plants followed a similar replication strategy.Furthermore, Mobis manufactures its modulesalmost entirely for Hyundai (the exception is a con-tract with Chrysler) and, thus, is guaranteed consid-erable business; it is heavily dependent on Hyundai’ssales, but also does not face competitors.

Supporting Hyundai’s rapid global expansion tobecome the fourth-largest automaker, Mobis became

the tenth-largest global automotive supplier, withsales that grew from $1.7 billion in 1999 to $14.4billion in 2010. Mobis’ profitability exceeds that ofHyundai Motor; its profits increased 50 percent year-on-year from 2009 to 2010, reaching $1.6 billion.The great increase in the number of Hyundaivehicles on the road provides high demand for after-market parts; as I’ve mentioned, this is a high marginbusiness that subsidizes R&D investments at Mobis.

We visited the Mobis plant, built in 2003 andlocated 12 kilometers from Hyundai’s Asan Bayassembly plant, in the fall of 2005. All chassis,cockpit, and front-end modules for Asan Bay aremade in this plant, at unit volumes of 300,000 peryear. The modern facility is nonunion and staffedwith contract workers from five different agencies,with staggered contract lengths so that turnover orcontract renewals affect only a subset of the work-force at any one time. Wages are about two-thirds ofthe level paid at Hyundai’s unionized assemblyplant.

We also visited the Mobis plant in Montgomery,Alabama in 2006. Two different instrument panelmodules as well as chassis modules were being builtthere for the Sonata sedan and Santa Fe SUV, with150 minutes of lead time before sequenced deliveryto Hyundai’s assembly plant, 12 miles away. Openedin 2002, volume had reached 500,000 modules peryear, with assembly lines and quality proceduresvirtually identical to the Mobis plant near Asan Bay.Wages for the 850 employees at the nonunion plantstarted at $10.50 per hour and increased to $15 perhour over two years, compared with wages at theHyundai plant that started at $15.60 per hour andincreased to $23.60 per hour. Both Mobis plantswere clean, well lighted, and spacious—in contrastwith other Korean supplier plants we saw in bothcountries.

Design and R&D collaboration

In its early history, Mobis was simply assemblingcomponents in the same manner as a Hyundaiassembly plant would have done. But it has movedsteadily toward more integrated module designs,with particular attention to reducing the weight andnumber of parts as well as better integrating themultiple functions embedded in each module.

To understand how Hyundai/Kia and Mobis thinkabout achieving design gains from production-defined modules, I did interviews at the R&D facilityof Mobis in Korea in 2008, 30 kilometers from Seoul



and not far from Hyundai’s R&D center. A chassisengineer described how Hyundai benefits fromMobis taking over the manufacturing of modules andthen explained why it is a challenge for Mobis tobenefit from module production to the same extent:

‘Chassis modularity is not as beneficial for Mobis.We need to develop our own components. Otherwise,if we get components from suppliers, those costs arevery clear to the OEM and we don’t control them. Wecan only manage labor cost in that situation. So weare trying to develop our own electronics to controlall functions in the module—suspension, ABS,power steering—to get the benefit of software-basedintegration and receive higher margins.’

From its founding, Mobis has emphasized build-ing capabilities to achieve greater design modularity.Initially, Mobis entered into technology allianceswith Bosch, Textron, and Siemens Automotive. Overtime, Mobis has steadily increased hiring into itsR&D function, growing from 150 engineers in 1999to 700 in 2007. Its stated R&D goals are to progressrapidly from modules done mainly for assembly andlogistics to functionally integrated modules charac-terized by reduced weight and number of parts,increased convenience of assembly, efficient inven-tory management, and cost savings.

Mobis also aims to be a technology innovator inits modules. This is facilitated by the close relation-ship between Mobis R&D and Hyundai R&D. As aMobis manager explained:

‘We have a lot of useful overlap in knowledge. Thirtyto 40 percent of our engineers come from Hyundai,with lots of experience in car design. This makescommunication with the OEM design engineerseasy. The remaining 60 to 70 percent of our engi-neers are from component manufacturers. This helpsus understand both sides—vehicle and component.Hyundai’s system engineering capability is good, butthey don’t know the details of components. Mobisengineers can provide this knowledge linking.’

A project on electronic controls illustrates howclosely Mobis and Hyundai work together:

‘For unified electronic controls in the chassismodule, we have a software team and a hardwareteam working together on this, located in the sameoffice, talking all the time. Our software engineersare from electronics companies like Samsung andLG and we are using their techniques for software.We need to get lots of information from Hyundai and

Kia—body data, engine characteristics. This is veryeasy for us with Hyundai because we are family—this is very difficult with other OEMs.

Collaboration is also necessary for problems likeNVH (noise, vibration, harshness):

‘We can’t address NVH issues within chassis alone:it is tied to many other aspects of product design.When we have NVH issues, Hyundai and Mobisengineers meet frequently to resolve them.’

Thus, the design trend is greater levels of interde-pendence within these modules over time (as thenumber of parts is reduced, each part carries more,not fewer, functions) with continued interdepen-dence across module boundaries for product perfor-mance criteria like NVH that are emergent andsystemic. The paradox is that the performance ofthese ‘modules’ increases as they become moreinternally integral in terms of product architectureand as increased learning about cross-module inter-dependencies leads to an ever-more-integrated orga-nizational architecture.

External customers

Although Mobis’ sales are dominated by Hyundai, itseeks to diversify its customer base. Its most notablesuccess to date was winning a 2006 contract toprovide the front-end module to the Jeep Wranglerproduced at Chrysler’s Toledo, Ohio, plant; Mobisbuilt a new plant nearby to support sequenced deliv-ery. The relationship with Chrysler expanded in2010 when Mobis landed a $2 billion contract tosupply front-end and rear cradle modules for theJeep Grand Cherokee; it reopened the closed Detroitplant of a U.S. supplier for the needed proximity. Sofar, Chrysler is its only non-Hyundai group cus-tomer. Developing new customers has proven diffi-cult, according to a Mobis senior executive:

‘The module customer and the module supplier needa long-term business relationship, like Hyundai andMobis—that way, we know we can invest. With anoutside OEM, it is very difficult to know if we caninvest. When we are asked to bid, we need to thinkabout how long a relationship may be possible. Wehave a 10-year relationship with Chrysler in theirsupplier park. If Chrysler wants a change after that,they have to buy the plant back from us.’

In this executive’s view, there is another necessarycondition:

28 J. P. MacDuffie


‘Mobis wants to expand its business to other OEMs,but first Mobis needs to have control over componentsuppliers and over module design— now both areowned by Hyundai. Without this control, building ageneral business isn’t so easy.’

This executive explained that Mobis decided tobring development of some technologies inside inorder to do more integration, e.g., proprietary workon air bags, brake systems, steering devices, head-lamps, and unified electronic control software.Another motivation is that ‘developing advancedtechnical capabilities is very important for Korea.’He said they would not keep these technologies ver-tically integrated indefinitely, mindful of the advan-tages of working with external suppliers to keep pacewith technological change. Thus, while the tightlyintegrated relationship with Hyundai will be subjectto disintegration pressures, it is likely to remainstable for some time.

Analysis

The framing of the modularity initiative at Hyundaiand Mobis was multifaceted. Adopting modules atHyundai was initially given a production rationale.At the same time, pursuing design improvementswithin modules was presented as a primary source ofMobis’ future competitive advantage. By includinglong-term design innovation in the cognitive framewhile also aggressively pursuing short-term produc-tion gains, Hyundai and Mobis managers justifieda high investment in R&D and in integratedorganizational arrangements that would movemodularization-as-process forward.

Mobis was initially created on the productionrationale that Hyundai, in adopting modules, wouldbenefit from the organizational separation of manu-facturing activities aligned with the module bound-ary. The three modules that Mobis builds all followthe industry’s production-oriented definition.Hyundai, thus, accepted an already-established levelof modularity-as-property and, through Mobis,pursued the manufacturing advantages this couldprovide. (These are exactly the same as those iden-tified in the historical account, i.e., there is noevidence that Mobis found new manufacturingadvantages to exploit.)

As its module business has grown, Mobis increas-ingly emphasizes the ‘design innovation’ framing ofits modularity strategy. Indeed, Mobis presents itsJust-in-Sequence modularity system as a ‘third revo-lution’ for the automotive industry, following Henry

Ford’s moving conveyor line and Toyota’s Just-in-Time system—both production process innovations.The commitment to this strategy is cast in a longtime horizon, well beyond other OEMs. Theseprocess innovations are framed almost entirely asarising from the relationship between Mobis andHyundai, which was established in the change oforganizational architecture that kicked off the modu-larity initiative.

From its inception, the relationship betweenHyundai and Mobis has stayed closely integrated.The equity cross-holdings and governance structure,the frequent interaction of Mobis and Hyundai engi-neers, and the sharing of design tasks between Mobisand Hyundai R&D make this a quasi-vertically inte-grated relationship, marked by tight interpersonaland organizational ties across firm boundaries. Farfrom a reduced amount of coordination betweenOEM and supplier, collaboration between Hyundaiand Mobis has intensified over time. Module designsremain closed, proprietary, and customized bymodel; Mobis’ limited foray into working withanother customer (Chrysler) has not shifted modulespecifications toward industry standardization.

What Mobis and Hyundai are doing together isnot revolutionary; both production gains and designgains are logical outcomes of incremental improve-ment processes. What is most striking in their col-laboration is the patient pursuit of module-leveldesign innovation within the confines of production-defined module boundaries under conditions wherestandardization prospects are low, facilitated bytightly integrated organizational arrangements thatfoster learning-oriented modularization processes.

DISCUSSION

A concept such as modularity that can be simulta-neously presented as source of the increased paceand extent of innovation, the surge of outsourcing/offshoring, and the feared loss of capabilities andadvantage by incumbent firms and developed nationsis powerful indeed. It is this very power that hasmade modularity into a Rorschach test; observersmay discern the shape of whatever they are mostinclined to see amid its murky outlines. Paying atten-tion to the antecedents of decisions to pursuemodularity-as-property, to what is learned frommodularization processes, and how ‘modularity’-in-frame influences the interaction between propertyand process is a priority in order to progress beyond



this confusion and indeterminacy (Kotabe, Parente,and Murray, 2007). Here I compare the two automo-tive case studies and draw out the implications formodularity research and for global architecture ofthe multinational firm.

Comparing Ford and Hyundaimodularity initiatives

I summarize my analysis of the Ford and Hyundaimodularity initiatives in Tables 2 and 3, identifyingantecedents, key aspects, and consequences withrespect to modularity-as-property, modularization-as-process, and ‘modularity’-as-frame. I comparethe two cases in Table 4, pointing out common fea-tures as well as differences.

As shown in Table 4, the two modularity initia-tives arise at roughly the same time (1999 to 2000)and both start from the production-based moduledefinitions already prevalent in the industry. Both setstrategic goals that link modularity to outsourcing,and both begin outsourcing the production of largesubassemblies to suppliers before they have fullycrystallized their plans to pursue design modularity.Both have shifted their cognitive frame from com-ponents to modules and anticipate that performanceadvantages on multiple dimensions will result fromthis shift. Both recognize that product architecturechanges need to be linked to changes in inter- andintraorganizational arrangements. Finally, bothcome to understand the need for OEM and suppliersto have a high level of knowledge of all components

Table 2. Ford case

Modularity-as-property Modularization-as-process ‘Modularity’-as-frame

Antecedents Production-defined modules,following industry usage;reinforced by supplier parkexperiences.

Suppliers ask to bid on designat module, rather thancomponent, level.

Outsourcing manufacturing tosuppliers to take advantageof lower labor costs, improvequality, reduce Ford’sassembly plant complexity.

Key aspectsof initiative

Define entire vehicle as 19modules, each one asubassembly of manyphysically proximatecomponents fulfillingdifferent vehicle functions.Move quickly to pursuedesign innovations withinmodules. Implement moremodules in each new vehicleprogram.

Debate within Ford aboutwhether design andmanufacturing knowledgeand tasks should stay insideor be outsourced tosuppliers.Executives/managers wantbig supplier role, engineersare skeptical. Suppliers seekautonomous design role,resist Ford’s ‘shadowengineering.’

Analogy drawn to computerindustry. Relying on supplierexpertise to increase designinnovation. Sharing capitalrisk, reducing coordinationwith suppliers. Achievingcost savings, qualityimprovements and more andfaster innovation. ‘Putexpertise in the supplier basewhere it should be.’

Consequences Modules as defined can beoptimized for production(costs sometimes higher), butdesign gains difficult toachieve due to highinterdependencies acrossmodules. Few new modulesadopted due to cost andfunctional performanceconcerns, so entiremodularity initiative isabandoned.

Supplier proposals for modulesdifficult for Ford purchasingto evaluate. Supplierspropose design innovations,but don’t understand allcomponents and interfaceswell enough. Limitedinteraction between Ford andsuppliers, beyondmonitoring, due toexpectations of supplierautonomy. Few supplierproposals adopted.

Viewed as a valid strategy, withpoor implementation due toskill and knowledgeshortfalls. For Ford, needingbetter systemsengineering/integrationcapabilities. For suppliers,needing higher level ofengineering talent, moresystems knowledge of allinterfaces with allcomponents; lackingexpertise in understandingcustomers, warranty andrepair.

30 J. P. MacDuffie


and interfaces related to modules—although thisawareness comes slowly to Ford (only at the end ofits modularity initiative), while it emerges earlier atHyundai and Mobis.

The company cases differ primarily in thecognitive ‘modularity’-as-frame that spurred andinformed each initiative, affecting module definition,strategic goals, the lessons drawn, and the extent ofmirroring of product and organizational architecture.Each ‘modularity’ frame sets a particular directionfor the initiative and a particular pace and sense ofthe necessary time frame for achieving strategicgoals.

Ford drew on a computer industry analogy toframe ‘modularity’ as a way to change the division oflabor between OEMs and suppliers and, hence,industry structure. These ambitions of Ford’s CEOand senior managers, building on production expe-riences already underway with a few modules atsupplier parks, drove an immediate partitioning ofthe entire vehicle architecture into modules. Underguidance of Ford’s cross-functional modularity task

force, newly identified modules were givenproduction-based definitions, following the patternset by existing modules. Ford’s senior managers andstrategy staff expected that suppliers could immedi-ately begin to work autonomously on moduledesigns on the basis of these definitions, thus movingquickly toward the modular organizational arrange-ments that would mirror the product architecturechoices. Chief engineers, it was also expected,would adopt more and more modules with each newvehicle development project. That these expectationsof achieving modularity-as-property were not ful-filled both confirmed the skepticism of Ford’s engi-neers and reflected their close monitoring of supplierefforts and unwillingness to adopt supplier moduleproposals.

Hyundai, in contrast, initially emphasized the pro-duction advantages of separating module manufac-turing made possible by the acquisition of Kia andthe creation of Mobis, focusing on three modulesalready well established in the industry and movingquickly to high-scale implementation at a time of

Table 3. Hyundai (Mobis) Case

Modularity-as-property Modularization-as-process ‘Modularity’-as-frame

Antecedents Hyundai reorganizes afterAsian financial crisis,decides to adopt industrystandard modules, seekingproduction advantages.

Mobis is spun off as separateparts subsidiary, givenmodule responsibility.

Organizational separation ofMobis from Hyundai Motorwill provide cost savings(lower labor costs), moreflexibility, more speed insupporting Hyundai’s rapidglobal growth. (Hyundaiperspective)

Key aspectsof initiative

Focus on three modulescommon in industry (chassis,cockpit, front end) usingindustry’s production- baseddefinitions, pursue long-termdesign gains within module.

Close governance ties, closeproximity, frequentinteraction of Mobis andHyundai engineers allowsprogress on designintegration.

Doing only production tasksfor Hyundai brings limitedbenefits to Mobis. Forcompetitive advantage,Mobis needs to innovate forbetter design integrationwithin module. (Mobisperspective)

Consequences Rapid growth of Hyundai andMobis, plus high marginsfrom Mobis’ aftermarketparts business support highinvestment in R&D.Production gains achieved,slow but steady progress ondesign gains.

Recognition that a high level ofinformation exchange andinteraction between Hyundaiand Mobis is needed fordesign gains in modules;cautious addition of newcustomers.

Mobis needs to gain morecontrol over module designand over (lower-tier)component suppliers to makefurther design progress;should do more proprietarytechnology development.Long time frame needed toachieve this. (Mobisperspective)



Table 4. Ford and Hyundai (Mobis) cases compared

Common Ford Hyundai (Mobis)

‘Modularity’-as-frame

Both see significance inshifting frame fromcomponent level tomodule level in termsof achieving multipleperformanceadvantages.

Applies computer industryanalogy: modules developedby autonomous suppliers asa way to reduce coordinationcosts, speed innovation,share risk. Focus onachieving modularity-as-property.

Applies historical auto industryanalogy: modules as next bigproduction process innovation,after moving assembly line (Ford)and JIT (Toyota). Closeorganizational relationship ofHyundai and Mobis seen as sourceof lower costs, more designinnovation. Emphasis onmodularization-as-process.

Moduledefinition

Both start withproduction-basedmodule definitionscommon in autoindustry, i.e., physicalchunks of proximatecomponents that mapto different vehiclefunctions.

Defines entire vehicle as set of19 modules. Some are wellestablished at industry level(though not standardized)and outsourced, others arenewly conceptualized asmodules and still verticallyintegrated.

Keeps primary focus on threewell-established modules, e.g.,chassis, instrument panel, frontend, that will be produced inseparate Mobis plants. Establishedin Hyundai’s production systemduring period of rapidglobalization.

Strategicgoals

Both link modularityinitiatives to theseparating ofproduction and theoutsourcing of largesubassemblies and topursuit of designimprovements.

Move quickly beyondoutsourced production toimplement 19-modulearchitecture. Invite suppliersto submit module designproposals. Urge chiefengineers to put modulesinto new vehicle projects.Prioritize reliance onsupplier expertise, supplierautonomy.

Pursue production gains first for threecore modules. Over time, realizedesign gains through high R&Dinvestments and close coordination,frequent interaction between Mobisand Hyundai engineers. Maintainexclusive relationship, withHyundai as only (primary)customer. Decision to develop mostnew technology within Hyundaiand Mobis.

Lessonsdrawn

Both come to understandthe need for bothOEM and supplier tohave high level ofknowledge of allcomponents and allinterfaces.

Barriers to modularity seen asevidence of capabilityshortfalls at both Ford andits suppliers, rather than lackof emphasis on mutuallearning throughmodularization processes.

Gains from modularity seen whereintegrative organizational processesbetween Hyundai and Mobis andinternal technology developmentbring design improvements.Expecting difficulties in developingmodules with otherOEMs/suppliers.

Mirroring ofproduct andorganizationalarchitecture?

Both recognize thatproduct architecturechanges need to belinked to changes in(intra- and inter-)organizationalarrangements.

No. Expects productmodularity (definitions thatidentify thin crossing places)to allow organizationalmodularity (low coordinationwith suppliers), but theseexpectations are frustrated.Initiative is abandoned.

Yes. Expects product modularity(with production-based definitions)will allow production-basedorganizational separation, but thatdesign gains will require integratedOEM-supplier relationship givenfunctional interdependence acrossmodules. Making gradual progresstoward design gains focused onmore integration within modules.

32 J. P. MacDuffie


global expansion. ‘Modularity’ was framed as anindustry innovation in production process; Mobisasserted that it would be the third ‘revolution’ ofautomotive manufacturing, joining lofty predeces-sors (Henry Ford’s moving assembly line and TaiichiOno’s Just-in-Time production system).

In the Ford case, all parties involved focused onachieving modularity-as-property, including execu-tives who sought more agility, better ROA results,and more (and faster) innovation from outsourcingmodule design and production; product developmentand purchasing managers who anticipated greatreduction in coordination costs and scale economiesfrom industry standardization of modules; and sup-pliers who hoped to move into a new role as autono-mous providers of innovative design proposals. Eventhe engineers and manufacturing managers whowere most skeptical about supplier capabilitiesdidn’t so much question the vision of mirroredmodularity in both product and organizationalarrangements as they cautioned going slowerand advocated closer monitoring. None of theparticipants emphasized the potential gains from amodularization process of understanding interdepen-dencies across components and module boundaries.Even when the initiative was shut down, the collec-tive attribution was to inadequate capabilities forreaching modularity-as-property, rather than to aprocess failing.

Meanwhile, for Hyundai, the primary framing forthe initiative was modularization-as-process, arisingdirectly from its origin in a major change in organi-zational architecture, i.e., the creation of Mobis inthe turbulent period after the 1997 Asian currencycrisis. Mobis, as it spun off from Hyundai Motor,was very central in the OEM’s organizational struc-ture and in its sense of key competencies. BothHyundai and Mobis executives saw the value ofquasi-integrated organizational arrangements fromthe beginning. But it was Mobis managers and engi-neers who realized over time that solely manufactur-ing modules offered them little long-term benefit.Achieving design improvements to boost the Mobisshare of value-added would require deepening theknowledge-intensive exchanges between Hyundaiand Mobis engineers. Accordingly, interactionsamong Hyundai and Mobis R&D engineers, i.e.,those with systems integration expertise and thosewith component expertise, respectively, intensifiedafter Mobis was established as a separate organiza-tion, the reverse of what is generally forecast whenorganizational architecture becomes more modular.

The ‘modularity’-as-frame identified in thesecases—each cognitively influencing attention, infor-mation processing, interpretation of events, andlegitimacy of contested perspectives—powerfullyaffected the direction, focus, and pace of each initia-tive, as shown in Figure 2. Ford’s ‘modularity’frame, based on a computer industry analogy andfocused on speedily achieving modularity-as-property, created barriers to learning that impededmodularization processes. Hyundai’s framing of‘modularity’ as an automotive process innovationrequiring integrative organizational arrangementskept Hyundai and Mobis engineers engaged inmodularization-as-process and alert to long-termopportunities for design improvement. While therelationship between modularity-as-property andmodularization-as-process is still bidirectionaloverall, the predominant frame applied in each casegave more weight to one of the two directions in thedynamics of each initiative.

While Ford’s initiative sought mirrored modular-ity in both product and organization and failed toachieve either, Hyundai and Mobis had a differentexperience. The integrative expertise developingbetween Hyundai and Mobis was a good match toinherent technical characteristics of the modulesbeing manufactured because these physical chunkscontained bits and pieces of many different func-tional systems whose other components were distrib-uted around the vehicle. The more a Mobis module‘knew’ about other modules and components andoverall requirements for the vehicle, the better itwould work. Mobis engineers needed that breadthof knowledge, which they got from sitting withHyundai engineers; in turn, the Hyundai engineerslearned much more about component technologies.

It is in this regard that the Hyundai case, which atfirst appearance looks like a classic misalignmentstory of integrated organizational arrangementslinked to modular product architecture, can be

Figure 2. Modularity as property, process, and frame



reinterpreted as supporting the mirroring hypothesis.Modules defined for production weren’t actually allthat modular from a design perspective; in fact, theirblend of integral and modular characteristics couldeven be said to tilt toward the integral end of thespectrum. But as such, the quasi-vertically integratedorganizational relationship of Hyundai and Mobiswas well-matched to the product architecture. Notethat the direction of influence in this case was froma change in organizational architecture—a conse-quence of Hyundai’s restructuring—that createdfavorable conditions for Mobis to explore changesin product architecture, shifting over time from aproduction focus to a design focus, all within theconfines of existing production-based moduledefinitions.

Implications for modularity research

Much research approaches modularity as somethingthat inherently resides in a product or organization,i.e., as a stable property that, once identified, can beapplied as a label. I argue here that this viewpoint isnaively deterministic and incomplete. Strategies fortechnological development don’t simply arise fromtechnical attributes of the product architecture orfrom how development processes are carried out, butrather are shaped, often dramatically, by the cogni-tive frames that firms bring to bear. The Ford case, inparticular, shows how much the analogy of the com-puter industry shaped the ‘modularity’ frame toanticipate a particular path to achieving ‘modularity-as-property.’

While it is always challenging to assess the extentof modularity-as-property for a complex, multitech-nology product, the automotive case supports Fix-son’s (2005) argument that function-to-componentmapping should be evaluated separately from inter-face coupling. IMVP researchers, when developingthese case studies, applied Ulrich’s (1995) ideal-typical definition to what the automotive OEMs andsuppliers were calling ‘modules’ and concluded thatthey didn’t deserve the name since there was noone-to-one mapping of function to component. Here,in the historical account, I present a revised perspec-tive that highlights automaker motivations for pursu-ing outsourcing, which constituted separation alonga production-defined boundary. The initial focus onoutsourcing was not related to modular productionper se, but it did influence the production-baseddefinition of modules; emphasis on moduledesign emerged only after this definition was well

established. Fixson’s (2005) approach helps makethe sequence of events in the auto industry’s emer-gent focus on modules more comprehensible.

With respect to the extensive literature on the‘mirroring hypothesis,’ this account offers qualifiedsupport, but also a challenge. The automotive caseis somewhat confusing, including to industry par-ticipants, because what they call a ‘module’ is dif-ficult to separate from a design perspective, i.e.,relatively integral, due to functional interdependen-cies across its boundary, while being easy to sepa-rate from a production perspective, i.e., relativelymodular, due to the minimal difficulty of achievingone-step installation at the OEM’s assembly plant.To call something a ‘module’ that is actually mostlyintegral would appear to set up a classic misalign-ment of product and organizational architecture,with the organization responding to the misleadinglabel and choosing modular organizational arrange-ments, i.e., low coordination, that doesn’t fitwhat the integral product architecture needs to beeffective.

However, as I have argued, once differences in theintended purposes of modularity are taken intoaccount, it is possible to see a mirrored match hererather than a misalignment. Automotive ‘modules’are easy to separate for production and, organiza-tionally, they are so separated; meanwhile, thesesame ‘modules’ are difficult to separate for design,so organizationally (e.g., at Hyundai and Mobis)integrated arrangements are established.

But I do not take this as complete backing forthe mirroring hypothesis. There is little support inthis industry case for a priori expectations that ashift toward modular (integral) product architecturewill lead inevitably to modular (integral) organiza-tional architecture. Indeed, as noted, this contextseems highly likely to display the phenomenon thatgreater modularity, particularly of design, willemerge only as the ex post consequence of inte-grated organizational arrangements. Static cross-sectional assessment of product and organizationalarchitecture in search of matched mirroring cangenerate misleading interpretations, while patientlongitudinal tracking of the movements of productand organizational architecture (and how they relateto each other) can provide a better source ofinsight.

With respect to the boundaries of the firm,Baldwin’s work (2008) on ‘where do transactionscome from’ highlights the contingent reasons whytasks involving dense and complex transfers of

34 J. P. MacDuffie


information should be located in ‘transaction-freezones’ where transaction costs don’t overwhelm thecoordinative capacity of the system. Our mentalimage of a ‘transaction-free zone’ is often a firm(Kogut and Zander, 1996) where culture, sharedidentity, and management practices facilitating moti-vation, skill acquisition, and adaptability can boostcapabilities for effective coordination. But here, andin other research (see MacDuffie and Helper, 2007),I find it is not firms that are the most reliable‘transaction-free’ zones but rather that quasi-integrated organizational arrangements prove to bethe most effective means of knowledge integrationfor design improvement. Future research couldpotentially identify the conditions under which afirm really is the best ‘transaction-free zone’ versuswhen quasi-integrated relationships with suppliersprovide that opportunity.

This research speaks to Campagnolo andCamuffo’s identification (2010) of opportunities toadvance the modularity literature: (1) it explicitlyincorporates a life cycle perspective on the evolu-tion of module definitions and the interrelationshipof production and design; (2) it pays close attentionto the OEM’s system integrator role as a barrier tomodularization processes and a source of stabilityunderlying persistent integrality; (3) it attends toperformance issues, acknowledging that not alltechnically feasible modular designs will beadopted if their performance lags more integraldesigns; (4) it questions the unidirectional causalityof the ‘product designs organization’ assumptionand, hence challenges technologically deterministviews of the consequences of modularity; and (5)similarly, it shows that product and organizationalarchitecture are not necessarily directly correlated,but may have different trajectories for their inter-relationship, due to different starting points andemergent misalignments.

Implications for global architecture of themultinational firm

This research can also inform how global firms thinkabout the strategic uses of modularity. When out-sourcing production and/or design activities inpursuit of the benefits of modularity-as-property,such firms may make the mistake of believing thereis a technologically determined path leading towardinevitable outcomes, i.e., that this action will reduceproduction costs and speed innovation as global sup-pliers apply their specialized expertise to rapid

improvements in the components going into keymodules, allowing the firm to focus just on thosetasks (e.g., R&D; high level product architecture)that are core to its competitive advantage. Instead, amultinational firm should approach both globaliza-tion and modularization as quasi-experimental learn-ing processes that can provide ongoing data on howwell capabilities are being extended into new coun-tries, how successfully technical tasks are beingtransferred to other firms, and what knowledge itneeds to maintain to oversee these externalizingactivities.

When global firms view their efforts to decom-pose complex problems and to understand interde-pendencies as ubiquitous and perpetual, they aremore likely to see how the learning processes ofmodularization can result in more integral, as well asmore modular, products and organizations. This isespecially true as new players, many from emergingmarkets, enter the auto industry value chain (Kuma-raswamy et al., 2012). The fact that quasi-integratedorganizational arrangements advance these learningprocesses when considerable interdependenciespersist across module boundaries suggests payingattention to proximity of design activities betweenOEMs and suppliers in global locational choices.Planning for thick communication across organiza-tional boundaries may be particularly importantwhen systemic characteristics important to productperformance are unpredictably emergent from cross-module interdependencies.

Generalizability and limitations

The research is primarily a single-industry casestudy, underpinned by implicit (and sometimesexplicit) comparisons to the computer industry.While we originally made that choice because ofhow often during our automotive data collection thelatter’s highly modular product and organizationalarchitecture was invoked, I do so here to highlight‘polar types’ and to establish a basis for predictingwhat will happen vis-à-vis modularity in otherindustries. Asking whether an industry is more likeautomobiles or more like computers is a good firststep, particularly since the latter has been overusedas a generalizable analogy.

Automobiles are an extreme example of a multi-technology complex product that is difficult to char-acterize under a single label. While some claim thatassessment of such product architecture is only reli-able if done at the component level of analysis, this



research highlights the value in examining the entireproduct and seeking to understand those systemiccharacteristics that emerge from component-levelinteractions. Thus, component-level and fullproduct-level investigations are complementary;both are essential.

This research looks closely at what the autoindustry chose to define as ‘modules,’ yet theprocess of modularization ubiquitously and perpetu-ally affects all components and subcomponents atall levels in a product hierarchy. Future researchcould study modularization processes at variouslevels, departing not from what industry partici-pants define as a ‘module;’ but rather examining anyoccasion on which managers and engineers makedecisions about function-to-component mappingand the potential for separability. For example, inthis same context, such an expansion of scope couldexamine functional systems that aren’t physicallyproximate but are still designed with clearly speci-fied interfaces to manage interdependencies withother functions, e.g., the safety system (Whitfordand Zirpoli, 2012).

Finally, this research demonstrates the importanceof paying heed to the multiple purposes to which‘modularity’-as-frame is applied, and how differentits influence can be if it directs attention primarily tomodularity-as-property versus modularization-as-process. As noted earlier, a primary focus on achiev-ing modularity-as-property can constrain attentionto the learning from modularization-as-process;accordingly, a primary focus on modularization-as-process can also reduce the salience of the goal ofreaching a high level of modularity-as-property,instead creating a trajectory that leads to greaterintegrality. Framing effects, thus, help explain theoften divergent findings in modularity research.

CONCLUSION

I examine three different but interrelated aspects ofmodularity to analyze product architecture initiativesin the global auto industry, in search of broaderstrategic lessons for global firms. Modularity is aproperty of the architecture of a complex system;modularization is a process that involves ongoinglearning about interdependencies among elements inthat system, i.e., where boundaries can be drawn andinterfaces specified to create ‘thin crossing places’between modules; and ‘modularity’ is a cognitiveframe that affects strategy by prompting a particular

dynamic—and directionality—in the interplaybetween property and process. I provide an industryhistory to document how the idiosyncratic definitionof automotive modules emerged, and I analyze twocompany cases, at Ford and Hyundai, of modularityinitiatives that both started from this definition butplayed out very differently.

The contributions of this research are fourfold.First, I demonstrate the crucial importance of exam-ining the antecedents of a module definition, whichconstitute the earliest stages of the intertwineddynamic between property and process. Differenttrajectories are triggered depending on how amodule is defined, due to path dependence at thefirm or industry level; the choices of an ‘architect’targeting a particular stage of the product life cycle;or negotiation among different interests. In the auto-motive case, the production-based definition of amodule arose from the production logic of separa-bility, but once established, it created difficulties foradvancing design modularity.

Second, I show how powerfully ‘modularity’ as acognitive frame can shape the interplay betweenproperty and process. Even when departing from thesame production-based definition of modules, theFord and Hyundai modularity initiatives displayeddifferent dynamics due to Ford’s emphasis onspeedy achievement of modularity-as-property andHyundai’s expectation that it would only be able togo beyond production advantages to achieving long-term design gains by emphasizing modularization-as-process through integrated organizationalarrangements with Mobis.

Third, I highlight factors that create barriers formodularity initiatives and can lead to persistent inte-grality, particularly from the reciprocal feedbackbetween product and organizational architecture andthe influence of potentially conflicting productionand design goals. Automobiles have a stable, long-standing dominant design and many interfaces arewell understood, but certain performance criteriacannot be predicted in advance due to complex inter-actions among functional systems and the consumerand regulatory requirements the product mustsatisfy. Understanding of these interdependenciesemerges only as iterative cycles of design decisionsoccur. Furthermore, transaction costs may favorkeeping a production or design activity inside anOEM, while emergent technical issues may create arequirement for OEM system integration knowledgebeyond a supplier’s capabilities. Thus, even wherecomponent interfaces can be well specified, systemic

36 J. P. MacDuffie


interactions and associated cost and capability con-siderations may render difficult the transfer of pro-duction and (especially) design responsibilities to asupplier. Overall, these cases provide evidence thatthe process of achieving modular design in automo-biles is likely to be slow and only partially attainable,at least with the current dominant design.

Fourth, I provide a fresh look at the mirroringhypothesis—which predicts a convergence over timein product and organizational architecture—basedon distinctive features of this context. The casestudies demonstrate that what progress has beenmade in moving toward modular design in the autoindustry has arisen within tightly coupled, quasi-integrative organizational relationships, showingthat industry structures and organizational architec-tures do not necessarily map to, or evolve toward,technological architectures. Indeed, in the Hyundaicase, the organizational design choice to establishquasi-integrative relationships between OEM andsupplier came well before the modest movementtoward greater design modularity that those arrange-ments made possible. ‘Organization designsproduct’ is as much an explanation for this particularcase study as ‘product designs organization.’

Most significantly, I hope to drive home twolessons for modularity researchers: (1) calling some-thing ‘modular’ imposes a powerful cognitive framethat is just as likely to obfuscate as to enlighten(particularly when not informed by context-grounded research); and (2) ‘modularity’ framesdrawn from one context are an unreliable guide towhat will happen in a different context. Even if acomponent or subsystem is called a ‘module’ tofocus attention on its potential separability andmanaged via ‘modular’ organizational arrange-ments, i.e., outsourced, careful scrutiny of interde-pendencies across the boundaries of that ‘module,’ toother components or subsystems, may well justify aclassification closer to the ‘integral’ end of the‘modular-to-integral’ dimension of product architec-ture. Witnessing separability and outsourcing andinferring the many attributes and meanings associ-ated with ideal-typical ‘modularity,’ e.g., as in thecomputer industry, carries high risk of creating ananalytic blind spot.

Managerial implications

Managers need to be aware of how modules aredefined in their industry context, since these defini-tions influence the pursuit of modularity-as-property.

Module boundaries defined for production don’tnecessarily coincide with the boundaries most favor-able to design, and vice-versa; therefore, a moduledefinition that privileges one may create difficultiesfor achieving the other.

Furthermore, managers often make the mistake ofexpecting that, having separated technical tasks, theycan move quickly to a full separation of all relatedorganizational tasks, i.e., that once they have out-sourced work on a module to a supplier, they can relyon the supplier to proceed independently based onclearly stated module boundaries, interface specifi-cation, and functional requirements. This can beineffective both because OEMs often need to retainthe ability for system integration (Zirpoli andBecker, 2011) and because it is the dialogue abouttechnical tasks after separation that often providesthe insights prompting design innovations. A centralparadox, therefore, is that modularity-as-propertymay emerge only after a modularization process thatis interaction-intensive and quasi-integrated, theopposite of what we expect from a ‘modular’ orga-nizational arrangement that mirrors the productarchitecture.

Thus managers should be careful not to overem-phasize, in their framing of ‘modularity,’ the notionthat modularity-as-property is inherent in a productand just needs to be identified through moduleand interface definition. Instead, it is better tode-emphasize whether modularity-as-property is (oris not) achieved and to emphasize insteadmodularization-as-process for learning about inter-dependencies within and across the boundaries ofboth product and organizational architecture. Thatknowledge, when developed together with suppliers,can provide the basis for sensibly refining the inter-organizational division of labor, managing coordina-tion costs, maintaining a balance across productionand design goals, appropriately adjusting moduleboundaries, and improving module performance.

Finally, managers need a healthy respect for thepower of ‘modularity’ as a cognitive frame, given thewide-ranging associations and macroconsequencesit evokes. ‘Modularity’-as-frame can become a cog-nitive trap in raising expectations of autonomousinnovation, strategic agility, frictionless coordina-tion, and industry scale efficiencies, a risk that growsall the greater when based on an analogy from a verydifferent setting. To avoid this risk, managers shouldpay careful attention to context and contingencies,process and property, and learning how to learn fromthe barriers to modularity they face.



ACKNOWLEDGEMENTS

I am grateful to my colleagues from the InternationalMotor Vehicle Program (IMVP) who worked withme on the Modularity and Outsourcing Project(Sebastian Fixson, Takahiro Fujimoto, Susan Helper,Fiona Murray, Frits Pil, Mari Sako, Akira Takeishi,and Max Warburton); to the managers and engineersat automotive OEMs and suppliers who graciouslygave their time and insight during interviews; and toIMVP for financial support. Thanks also to research-ers in the IMVP network whose independent workon modularity influenced my thinking: MarkusBecker, Arnaldo Camuffo, Charlie Fine, FrancoisFourcade, Christophe Midler, Josh Whitford, DanWhitney, and Francesco Zirpoli. During the lengthylife of this specific manuscript, I have been the ben-eficiary of helpful comments from (in addition to theaforementioned): Paul Adler, Edward Anderson,Lotte Bailyn, Carliss Baldwin, Robert Cole, MichaelCusumano, Annabelle Gawer, Rahul Kapoor, SteveKobrin, Glenn Mercer, Michael Piore, the specialissue editors, and three anonymous reviewers atGlobal Strategy Journal.

REFERENCES

Abernathy WJ, Clark KB, Kantrow AM. 1984. IndustrialRenaissance: Producing a Competitive Future forAmerica. Basic Books: New York.

Adner R, Levinthal D. 2001. Technology evolution anddemand heterogeneity: implications for product andprocess innovation. Management Science 47: 611–628.

Argyres N, Bigelow L. 2010. Innovation, modularity, andvertical deintegration: evidence from the early U.S. autoindustry. Organization Science 21(4): 842–853.

Baldwin CY. 2008 Where do transactions come from?Modularity, transactions, and the boundaries of firms.Industrial and Corporate Change 17(1): 155–195.

Baldwin CY, Clark KB. 2000. Design Rules: The Power ofModularity (Vol. 1). The MIT Press: Cambridge, MA.

Benford RD, Snow DA. 2000. Framing processes and socialmovements: an overview and assessment. Annual Reviewof Sociology 26: 611–639.

Berger S. 2006. How We Compete: What CompaniesAround the World are Doing to Make It in Today’s GlobalEconomy. Random House: New York.

Bingham C, Kahl S. The process of schema emergence:assimilation, deconstruction, unitization and the plura-lity of analogies. Academy of Management Journal.Forthcoming.

Brusoni S, Principe A, Pavitt K. 2001. Knowledge special-ization, organizational coupling, and the boundaries of

the firm. Administrative Science Quarterly 46(4): 597–621.

Cabigiosu A, Camuffo A. 2011. Beyond the ‘mirroring’hypothesis: product modularity and interorganizationalrelations in the air conditioning industry. OrganizationScience 23(3): 686–703.

Campagnolo D, Camuffo A. 2009. What really drivesthe adoption of modular organizational forms? An insti-tutional perspective from Italian industry-level data.Industry and Innovation 16(3): 291–341.

Campagnolo D, Camuffo A. 2010. The concept ofmodularity in management studies: a literature review.International Journal of Management Reviews 12(3):259–283.

Colfer L, Baldwin CY. 2010. The mirroring hypothesis:theory, evidence, and exceptions. Working paper, HarvardBusiness School, Boston, MA.

Donovan D. 1999. The Dawn of the Megasupplier: WinningSupplier Strategies in an Evolving Auto Industry. Bain &Company: Boston, MA.

Eisenhardt KM, Graebner ME. 2007. Theory building fromcases: opportunities and challenges. Academy of Manage-ment Journal 50(1): 25–32.

Ethiraj S, Levinthal DA. 2004. Modularity and innovationin complex systems. Management Science 50(2): 159–173.

Feldman MS, Orlikowski WJ. 2011. Theorizing practiceand practicing theory. Organization Science 22: 1240–1253.

Fine C. 1999. Clockspeed. Perseus Books: Reading, MA.Fine C, Whitney D. 1996. Is the make-buy process a core

competence? Working paper, Massachusetts Institute ofTechnology, Cambridge, MA.

Fixson SK. 2005. Product architecture assessment: a tool tolink product, process, and supply chain design decisions.Journal of Operations Management 23(3/4): 345–369.

Fixson SK. 2006. A roadmap for product architecturecosting. In Product Platform and Product Family Design:Methods and Applications, Simpson TW, Siddique Z,Jiao RJ (eds). Springer: New York; 305–333.

Fixson SK, Park JK. 2008. The power of integrality: link-ages between product architecture, innovation, and indus-try structure. Research Policy 37: 1296–1316.

Fixson SK, Ro Y, Liker JK. 2005. Modularization and out-sourcing: who drives whom? A study of generationalsequences in the U.S. automotive cockpit industry. Inter-national Journal of Automotive Technology and Manage-ment 5(2): 166–183.

Ford Motor Company. 1999. Modularity Task Force Report(internal document). Ford Motor Company: Dearborn,MI.

Fourcade F, Midler C. 2004. Modularization in the autoindustry: can manufacturer’s architectural strategies meetsupplier’s sustainable profit trajectories? InternationalJournal of Automotive Technology and Management 4(2/3): 240–260.

38 J. P. MacDuffie


Garud R, Kumaraswamy A. 1995. Technological and orga-nizational designs to achieve economies of substitution.Strategic Management Journal, Spring Special Issue 16:93–110.

Gavetti G, Levinthal DE. 2000. Looking forward andlooking backward: cognitive and experiential search.Administrative Science Quarterly 45: 113–137.

Herrigel G. 2004. Emerging strategies and forms of gover-nance in high-wage component manufacturing regions.Industry and Innovation 11(1/2): 45–79.

Hoetker G. 2006. Do modular products lead to modularorganizations? Strategic Management Journal 27(6):501–518.

Jacobides MG. 2005. Industry change through vertical dein-tegration: how and why markets appeared in mortgagebanking. Academy of Management Journal 48(3): 465–498.

Jacobides MG, Billinger S. 2006. Designing the boundariesof the firm: from ‘make, buy, or ally’ to the dynamicbenefits of vertical architecture. Organization Science 17:249–261.

Jacobides MG, Knudsen T, Augier M. 2006. Benefitingfrom innovation: value creation, value appropriation, andthe role of industry architectures. Research Policy 35(8):1200–1221.

Jacobides MG, MacDuffie JP, Tae CJ. 2012. When valuesticks around: why automobile OEMs still rule theirsector. Working paper, London Business School, London,U.K.

Jacobides MG, Winter SG. 2005. The co-evolution of capa-bilities and transaction costs: explaining the institutionalstructure of production. Strategic Management Journal26(5): 395–414.

Kaplan S. 2008. Framing contests: making strategyunder uncertainty. Organization Science 19(5): 729–752.

Klepper S. 1996. Entry, exit, growth, and innovation overthe product life cycle. American Economic Review 86(3):562–583.

Kogut B, Zander U. 1996. What firms do? Coordination,identity and learning. Organization Science 7(5): 502–518.

Kotabe M, Parente R, Murray JY. 2007. Antecedents andoutcomes of modular production in the Brazilian automo-bile industry: a grounded theory approach. Journal ofInternational Business Studies 38(1): 84–106.

Kumaraswamy A, Mudambi R, Saranga H, Tripathy A.2012. Catch-up strategies in the Indian auto componentsindustry: domestic firms’ responses to market liberaliza-tion. Journal of International Business Studies 43(4):368–395.

Langlois RN. 2002. Modularity in technology and organi-zation. Journal of Economic Behavior & Organization49(1): 19–37.

Langlois RN, Robertson D. 1992. Networks and innovationin a modular system: lessons from the microcomputer and

stereo component industries. Research Policy 21(4):297–313.

Lung Y, Salerno M, Zilbovicius M, Carneiro Dias AV. 1999.Flexibility through modularity: experimentations offractal production in Europe and Brazil. In Coping withVariety: Flexible Productive Systems for Product Varietyin the Auto Industry, Lung Y, Chanaron JJ, Fujimoto T,Raff D (eds). Ashgate Publishing: Aldershot, U.K.; 224–258.

MacDuffie JP, Helper S. 2007. Collaboration in supplychains: with and without trust. In The Firm as Collabo-rative Community, Heckscher C, Adler P (eds). OxfordUniversity Press: New York; 416–466.

Mudambi R. 2008. Location, control, and innovation inknowledge-intensive industries. Journal of EconomicGeography 8(5): 699–725.

Ocasio W. 2011. Attention to attention. OrganizationScience 22: 1286–1296.

Pil FK, Cohen S. 2006. The influence of modularity andcomplexity on innovation, imitation, and sustained com-petitive advantage. Academy of Management Review31(4): 995–1011.

Sako M. 2003. Modularity and outsourcing: the nature ofco-evolution of product architecture and organizationalarchitecture in the global automotive industry. In TheBusiness of Systems Integration, Principe A, Davies A,Hobday M (eds). Oxford University Press: Oxford, U.K.;229–253.

Sako M. 2009. Outsourcing of tasks and outsourcing ofassets: evidence from automotive supplier parks in Brazil.In Platforms, Markets, and Innovation, Gawer A (ed).Edward Elgar: London, U.K.; 261–289.

Sako M, Murray F. 1999. Modular strategies in cars andcomputers. Financial Times, 6 December: 4–7.

Sako M, Warburton M. 2002. Modularity and outsourcingproject: report of the European team. Working paper,International Motor Vehicle Program, MassachusettsInstitute of Technology, Cambridge, MA.

Sanchez R, Mahoney JT. 1996. Modularity, flexibility, andknowledge management in product and organizationdesign. Strategic Management Journal, Winter SpecialIssue 17: 63–76.

Schilling M. 2000. Toward a general modular systemstheory and its application to interfirm product modularity.Academy of Management Review 25(2): 312–334.

Siggelkow N. 2002. Evolution toward fit. AdministrativeScience Quarterly 47: 125–159.

Simon HA. 1981. The architecture of complexity. In TheSciences of the Artificial (2nd edn), Simon HA (ed). TheMIT Press: Cambridge, MA; 192–229.

Skinner W. 1974. The focused factory. Harvard BusinessReview 52(3): 113–121.

Staudenmayer N, Tripsas M, Tucci CL. 2005. Interfirmmodularity and its implications for product development.Journal of Product Innovation Management 22(4): 303–321.



Sturgeon T. 2002. Modular production networks: a newAmerican model of industrial organization. Industrialand Corporate Change 11(3): 451–496.

Takeishi A. 2001. Bridging inter- and intra-firm boundaries:management of supplier involvement in automobileproduct development. Strategic Management Journal22(5): 403–433.

Takeishi A, Fujimoto T. 2001. Modularization in the autoindustry: interlinked multiple hierarchies of product, pro-duction, and supplier systems. International Journal ofAutomotive Technology and Management 1(4): 379–396.

Tripsas M, Gavetti G. 2000. Capabilities, cognition, andinertia: evidence from digital imaging. Strategic Manage-ment Journal 21(10/11): 1147–1161.

Ulrich K. 1995. Product architecture in the manufacturingfirm. Research Policy 24(3): 419–440.

Utterback J. 1996. Mastering the Dynamics of Innovation.Harvard Business School Press: Boston, MA.

Weick K. 1969. The Social Psychology of Organizing.Addison-Wesley: Reading, MA.

Whitford J, Zirpoli F. 2012. Pragmatism, practice, and theboundaries of organization. Working paper, ColumbiaUniversity, New York

Yin RK. 1994. Case Study Research: Design and Methods(2nd edn). SAGE Publications: Thousand Oaks, CA.

Zirpoli F, Becker MC. 2011. The limits of design and engi-neering outsourcing: performance integration and theunfulfilled promises of modularity. R&D Management41(1): 21–43.

40 J. P. MacDuffie


modularity-as-property, …fixson and park, 2008). modularity-as-property and...

Documents