Performing Modularity: Competing Rules,Performative Struggles and the Effect of Organizational Theories on the Organization

The performative effects of theories

There is an increasing awareness of the links between organizational theories and their “performative” effects. In recent years, performativity-inspired approaches drawing on practice theory (Schatzki, Knorr Cetina, & von Savigny, 2001; Suchman, 1987), science and technology studies (Callon, 1998; Latour, 1987; MacKenzie, 2006; Pickering, 1995) and contemporary philosophy (Barad, 2007; Butler, 1993; Mol, 2005) have been gaining increasing recognition in mainstream organizational and management theory (Beunza & Stark, 2004; Cabantous & Gond, 2011; Ferraro, Pfeffer, & Sutton, 2005).

For instance, authors have described theories as constitutively shaping the design and performance of management practices and institutional arrangements, the social norms that regulate organizational and individual behavior, and the language, categories and assumptions that actors use to interpret the world around them (Ferraro et al., 2005). Further contributions have argued for the need to study organizations as sets of relations recurrently performed and reproduced over time and space (Orlikowski & Scott, 2008). According to these authors, organizational boundaries should never be conceptualized as solid or static but as constantly emergent and enacted in recurrent activities (Feldman & Orlikowski, 2011). A similar observation applies to organizational practices or routines: the routines’ ostensive—or abstract—pattern can never be taken for granted but needs to be “brought to life” every time through situated performance and enactment (Pentland & Feldman, 2008).

In this paper we focus on a recent perspective on performativity which has emerged out of studies of economic and financial markets as this, we believe, can provide a particularly useful template for conceptualizing the influence of organizational theories. Authors in this stream of work have suggested that theories and models are not simple descriptions of a setting but powerful “engines” which can profoundly transform the contexts they describe (MacKenzie, 2006). The key tenets of neo-classical economic theory, for example, do not simply describe the economy but are actually being performed within markets (Callon, 1998). Such is the strength of this performation that markets come to resemble the theory. Similarly, the mathematical models that underpin financial derivatives (theory) do not simply rest outside the market (reality) but are a constituent part of it (MacKenzie, 2006, MacKenzie & Millo, 2003). This, however, does not imply that theories are deterministically applied (MacKenzie, 2009). Rather, they are enacted through iterative cycles of “framing”, where theories begin to shape reality, “overflowing”, whereby differences emerge between reality and the theory, and “reframing”, whereby there is again convergence between theory and reality (Callon, 1998). These contributions thus provide the basis for a study of the emergent effect of theories over markets or organizations.

This dynamic approach, we suggest, is especially useful in highlighting how the influence of theories is never complete, and may vary significantly in degrees and levels. MacKenzie (2006), for example, has distinguished between theories that—when they are applied—have no or little observable effect on a setting (“generic” performativity), those that “make a difference” in some way (“effective performativity”), those that bring about the “state of affairs” for which they are a good “empirical description” (“Barnesian performativity”), and, finally, those theories that change economic processes so they conform less well to their depiction by theories (“counter-performativity”). Through foregrounding the dynamic adaptation between theory and reality, performativity can help us account for the nuanced and dynamic effects of modularity, and wider organizational theories, on the organization.

Performativity: Key notions

In order to understand how performativity can help us capture the co-production of modularity and the modular organization, we first need to look closer at its language and features. To this purpose we introduce four of its key constructs, starting from the notion of theory. Theories are understood broadly as including “formulas, methods, tools and instruments as well as verbal formulations” (MacKenzie, Muniesa, & Siu, 2007, p. 15). In the context of this paper, for example, the dominant theory relates to the organization’s imperative to follow a “strategy of modularity [which] entails both belief (‘this is the best course open to us’) and commitment (‘we are going to follow this course’) to modularity” (C. Baldwin, personal communication, September 2010). A theory is considered “successful” (Callon, 2007), or performative, when it brings about the vision of the world contained within.

In order to become performative, a theory or statement has to create a sociomaterial assemblage, or agencement (Callon, 1998), our second key notion. This must contain the theory and all the elements required to realize it (Callon, 2007), including a heterogeneous collection of material features (including technology, texts, and bodies) that “participate in its design, elaboration, experimentation, change, maintenance, extension and operation” (Callon & Caliskan, 2010, p. 23). Thus for a theory to be performative, it has to involve a heterogeneous configuration of actors and materials that supports its assumptions.

A third key idea, one which is as yet underdeveloped in the literature, is that the effect of a theory can only be decided in relation to other theories. Multiple theories typically coexist at any one time within any organization. Some of these may be complementary, whilst others may compete and clash substantially or irremediably (Callon, 2007). In order to theorize the influence of a theory in its context and over time, we thus need to take a further step and capture how it may operate “at large”, that is, in conjunction with other organizational theories and their rules. So far, for example, little is known about “how alternative [theories] compete within organizations … through ‘performative struggles’” (Cabantous & Gond, 2011, p. 582) and “influence behaviours and outcomes” (Cabantous, Gond, & Millo, 2014). In the case of modularity, this includes theorizing the process by which the modular organization emerges over time and the (organizational and contextual) sociomaterial dynamics that underpin its inner workings.

The point about sociomateriality takes us to a final key tenet. In a context where multiple statements struggle to establish themselves, one at the expense of the other, the potential of a theory to gather strength and prevail is closely related to its ability to involve not only actors but also artifacts and material devices (i.e., texts, documents, technologies). For example, by delegating their objectives and views as “scripts” (Akrich, 1992) to stable, durable and visible artifacts such as software, flowcharts or computer systems, actors might, at least in principle, be able to reinforce their own views and objectives over those of other agencies (i.e., organizational communities, functions or teams) that are not so well equipped (Callon & Muniesa, 2005). Performativity therefore provides opportunities to theorize the sociomaterial mechanisms that underpin the co-evolution of organizational theories and the organization.

Research context

To capture the influence of modularity theory over the modular organization we draw from an in-depth, longitudinal study of a pioneering electronics manufacturing firm—an early producer of workstations, storage and high-end computer servers—whose history has been interwoven with the history of modularity in important ways. As part of our investigation, we were able to establish the presence of close historical links between this company and modularity theory from the early years when this company was a fast-rising high-technology start up. Sources report how the pioneering methods used by this organization to manage its design and production activities had generated an early interest in management scholars towards understanding the laws that underpinned modularity and its principles, therefore directly inspiring the ground-breaking Harvard research on modularity (Russell, 2012). From the beginning, in the early 1980s, the company pursued modularity as an efficiency-seeking strategy aimed at streamlining and standardizing products and operations. The firm’s progressive vision of modularity, which supported at the same time superior performance and cost efficiencies, by mixing and matching standardized third-party components, allowed it to “solv[e] a critical paradox of computing” (Baldwin & Clark, 2000). This “open systems strategy” (company archives) in turn gave the company outstanding advantage over those competitors that were following more conventional “closed” trajectories (West, 2003).

A relentless faith in modularity distinguished the company from its competitors from the outset. Everything in their system was modular, “from the lowest level circuits within microprocessors to the highest level system software and application programmes. Its products could be connected easily to one another and other systems, like so many digital tinker toys” (company press archive). The company’s flamboyant CEO was himself a strong advocate of modularity. In his public speeches he often drew from the Lego metaphor, the shorthand for modularity, by comparing their products to “a car built out of Lego blocks that [will] all fit together very nicely, and if you don’t like [our component], take that Lego block out and put [a competitor’s module] in it” (company press archive ). From the early 1980s, therefore, and throughout the company’s phenomenal hike up the electronics industry ranks for the rest of the 1980s and the 1990s, the company’s belief in the power of modularity appeared unshakable.

The emphasis on modularity remained strong even when, during the mid-1990s, the company switched from building only workstations—aimed at the expert engineering user who could independently resolve any reliability issues that might have arisen—to high-end enterprise servers, aimed at the less competent corporate end user. The acquisition of a new server product line, purchased through a multi-million dollar deal that had made the history of computing, had in fact highlighted potential tensions between modularity and the further rationale of “reliability”. Reliability appeared to gain relevance in a server context, where computers are used to run mission-critical activities in organizations such as hospitals, banks and internet service providers, which are unable to tolerate any significant server downtime. Reliability thus threatened to challenge modularity’s so far uncontested dominance in the company strategy, vision and value set.

Even the acquisition of the new server product line (including the very same server which is the focus of this study), and the new reliability ethos supported by the various management and computer experts that had come across with it, however, did not dampen the company’s enthusiasm towards modularity. Modularity made mixing and matching components appear a straightforward activity which was unlikely to bear any repercussions on the performance of other components or on the system as a whole (Fleming & Sorenson, 2001). Modularity and reliability were thus not considered incompatible as, in a modular product, any errors and defects detected during testing could be, at least in theory, “swapped out” of the system simply by replacing any failed standardized modular components with new ones (Baldwin & Clark, 2000). According to our practitioners this principle applied to workstations as much as servers, the latter simply requiring more extensive testing compared to the former, in order to detect and swap any potentially defective components.

Modularity and reliability thus appeared (initially at least) to happily coexist side by side. The company’s CEO illustrated the supposedly harmonious relationship between reliability and modularity in computers in a speech when he stated: “Imagine if I had an airplane hangar at home, I designed my own plane from different parts, and I told you to fly it. It wouldn’t be safe, yet that’s what we’re doing” (press archives). Encouraged by this high-level rhetoric which acknowledged the importance of reliability but at the same time relegated it to the background, the company basked in the certainty that modularity would continue to reign unchallenged. And reign it did—at least for another few years.

Between the end of the 1990s and the beginning of the new century the electronics sector bubble finally burst. We were already working on site when the company began to realize that profit margins were significantly shrinking. The effect of the crisis was to create a strong drive towards further efficiency/cost-cutting, which was to be achieved through modularity and further standardization. The CEO, at this point, predicted “a ‘massive standardization’ in the way technology is deployed and developed” (press archive). This move was supported by an even greater emphasis on modularity, not only at the product but also at the organizational level, which provided greater flexibility to switch suppliers to acquire the latest innovations and best practice. This also allowed increasingly delegated production and testing activities for (both non-core and core but non-critical) components to external manufacturers thus relinquishing responsibility for component-level faults and avoiding the costs of repair. It is therefore during this period that the company’s belief in the power of modularity began to take hold even further. Modularity, and the associated quest for superior efficiencies, was going to lead them out of the crisis.

In the spectrum spanning from integrality to modularity, they therefore set out to place themselves at the far modularity end by adopting a totally “disintegrated” organizational structure (something which, however, they were never fully able to realize, as we discuss later). They intended to achieve this by becoming the “perfectly modular” organization. This involved focusing on one or a few core modules and activities and coordinating the rest through arm’s length transactions (Campagnolo & Camuffo, 2009). The principle behind this involved buying rather than making: they would no longer need to be manufacturers but would become pure integrators (Hobday et al., 2005) focusing their capabilities on assembling computers out of third party components. It was precisely at this time that the “rework and repair” process for one of their server system’s core components, the computer board that is the focus of this study, was outsourced to its external manufacturer.

The drive towards standardization/efficiency thus seemed inexorable. Further down at the organizational level, this involved an increased push towards “standardised work procedures (across departments and business units) and documented steps, the presence of electronic databases, and procedures and systems for transferring knowledge across projects and business units” (Campagnolo & Camuffo, 2009, p. 276). As part of this, the company started looking for additional efficiency-bearing strategies which would allow production costs to be reduced while further streamlining their processes. The “zero inventory” philosophy, for example, which includes the Autoswap system that we discuss later, involved introducing new IT-based lean tools that would allow them to substantially reduce inventory overheads. It was not yet evident at this stage that the increasing emphasis on efficiency, supported by product and organizational modularity and complementary inventory rules, could conflict with reliability with important consequences for the organization.

In the sections that follow we analyze the rise and fall of modularity within and beyond the boundaries of this outstanding organization. In so doing, we focus on the performation of modularity and complementary efficiency-bearing rules and their confrontation with the competing rules underpinning the philosophy of reliability. Through an in-depth case study focusing on the changes that occurred in context and over time to the board rework and repair process we theorize how and how far modularity was able to prevail over reliability and produce a modular organization.

Data collection

Exploring the contrasting forces that drive towards—or away from—the modularization of product and organization requires the close observation of organizational processes to identify changes occurring to product, technology and organizational interfaces. To this purpose we draw inspiration from the notion of an “artifact’s biography” (Appadurai, 1986; Pollock & Williams, 2009), which consists in following the key stages of a technology’s life-cycle and the practices and routines that surround it, as they develop over time and across contexts. Drawing on the biographical approach, we mapped the changes occurring to the computer board rework and repair routine over time, as rework was outsourced and later on re-insourced. To this purpose, we followed the rework routine at different stages to see how this changed as a consequence of the rework being conducted in-house or at the supplier’s premises. This entailed “following the board around” as it crossed the boundaries between supplier and integrator in a typical repair cycle. The biographical approach provided us with insights into modularity and its innermost workings, including how it might affect the emergence of organizational boundaries.

Data analysis

The data analysis and coding was undertaken jointly by both authors. The first author, acting as the “insider”, was able to draw upon the substantial knowledge and expertise acquired through her extended engagement with the research site to formulate possible explanations. The second author, acting as the “outsider”, played as the devil’s advocate (Rerup & Feldman, 2011) in challenging those emerging explanations. The second author also contributed towards analyzing the data by proposing explanations that were either accepted or refuted by the first author, based on her intimate knowledge of the field (Rerup & Feldman, 2011).

Alongside this dialectical role play we proceeded to analyze the data consistently with an inductive, “grounded theorising” approach (Glaser & Strauss, 1967). This involved proceeding in an iterative manner, scanning the extensive field data for recurrent concepts and themes, consolidating these into emergent categories (Nag, Corley, & Gioia, 2007), and subsequently evolving categories through constant comparison with related notions in the extant theory. This process led to developing new concepts, themes and dimensions, which captured how modularity was performed over time and within the context of the rework process.

Through applying the techniques of content analysis (Krippendorff, 2004) and constant comparison (Glaser & Strauss, 1967) we were able to identify a number of emerging first-order codes (Van Maanen, 1979). Initially, notions surfacing from the observation of the rework process and interview transcripts revolved around concepts such as “swapping”, “modular products”, “interchangeable components”, “a flexible organization”, “mix and match”, “open interfaces”, “delegating responsibility” and “the organization as a Lego system”. At a later stage, alternative notions began to take hold. These revolved around concepts such as “fixing”, “fault-finding”, “a reliable product”, “an error-free organization”, “debugging”, “errors”, “quality as a top priority”, “we must never fail”, “owning the problem”, and “failure-is-not-an-option”. Further axial comparison (Strauss & Corbin, 1990) between and across the codes revealed that these converged around two main patterns, which, using the informants’ labels (Van Maanen, 1979) we coded as “swapping” and “fixing”.

The juxtaposition of these emergent categories with those in the relevant literature produced a crucial discovery: each set of concepts (“swapping” or “fixing”) was largely consistent with the rules and assumptions typical of a specific organizational theory, modularity or reliability, as it may be played out in an actual organizational context. Put in other words, our sets of first-order concepts pointed to the fact that the principles of modularity and reliability theory were distinctly at work within the rework process, as highlighted by the fact that their rules and their effects were being constantly invoked during the course of everyday actions and discourses. This creative—but firmly grounded—leap (Suddaby, 2006) allowed us to “lift” our first-order codes to a more abstract level and led to coding (modularity and reliability) rules and errors as second-order themes. In addition, field quotes also showed how rules were supported by being embodied, embedded and enacted (M. Feldman, personal communication, August 2012) in and through heterogeneous ensembles of artifacts and communities which appeared to exert a key influence over the rework process and its outcomes. This led us to code artifacts and communities as further second-order codes.

Constant immersion in the data (Langley, 1999) combined with comparison against the extant literature allowed us to take our analysis to a further level of abstraction. As the evidence clearly indicated, modularity and reliability rules were not simply nominally present, but were being enacted and were shaping the rework process. This, in turn, suggested that modularity and reliability theory and their assumptions were having an effect on the process. The second creative leap in our analysis thus involved the realization that capturing the patterns emerging from the field required a framework able to explain the effect of organizational theories over organizational processes. Our search for a “theory of theories” or “meta-theory” led us to engage with performativity theory’s emergent constructs. This insight led us to characterize the effect of the rules and errors over the rework process in terms of “framing” and “overflowing” (Callon, 1998) and the influence of theories on the rework process as “performing” (our overarching dimension). This process led to developing a new framework which theorizes how modularity was performed over time within the context of rework.

Finally, further examination of interview and observation data confirmed that the balance of power between the two theories was unstable. This was highlighted by the fact that terms pointing to “fixing” or “swapping” were alternatively gaining and losing emphasis over time. Following these key insights, we were able to identify three temporal stages which captured the co-performation of modularity and reliability over time and within the board repair cycle: Stage 1: swapping prevailed almost unquestioned whilst fixing remained in the background; Stage 2: swapping and fixing were opposing each other in every aspect and at every level of the rework cycle; and Stage 3: swapping prevailed once again but this time was modified.

While drawing on a single case, our findings afford further generalizable insights about the complex relationship between modularity theory and the modular organization, and, more in general, organizational theories and the organization. Powerful generalizations (Feldman & Orlikowski, 2011; Siggelkow, 2007) can in fact result from combining strong inductive theorizing with a practice perspective (Schatzki et al., 2001) based on performativity theory (Callon, 1998; MacKenzie, 2006), which involves embracing the complexity and ambiguity of organizational life and emphasizing its multiplicity and emergence. A deeply grounded inductive approach drawing on in depth and longitudinal case study evidence can therefore provide fertile terrain for theory development.

The biography of a computer board

To capture the influence of modularity and its rules we draw from the detailed study of the computer board “rework and repair” cycle. A computer printed circuit board is a flat insulating sheet to which integrated circuits and other components are attached and is a critical component of the overall computer server system (Soanes & Stevenson, 2005). Servers are complex, customized, high-performance, -quality and -reliability machines that provide essential services across a network of computers. For example they can be used to host databases, files and web email. They are individually configured to contain several printed circuit boards, based on their performance and customer specification. Our server can contain up to 36 boards, which can perform together or as 36 individual computers.

There are several reasons why a board might fail: it can contain fabrication defects, it can become damaged during transportation, or it can be accidentally broken through manual or robot handling in the factory. Whatever the source of failure, this has to be identified and corrected. Repair must be followed by thorough tests to verify that the fault has been fixed. Undetected board-level faults are in fact likely to generate system failures that can occur at any stages of the product life-cycle.

While being a crucial aspect of high-end server production, the fault-finding and bug-fixing process appears at first sight to be rather simple and repetitive. The reality, however, could not be more different. The repair process is situated at the front line of the emergence and persistence of the modular organization, reflecting some of the key tensions that contribute to make an organization truly, or at least partially, modular. This is because it is precisely in this kind of context that organizational processes and boundaries are typically re-negotiated. In the following sections, we set out to follow the computer board around the rework and repair process, as it unfolds across the boundaries between systems integrator and component supplier.

Stage 1: Modularity “rules”

In the early days following the acquisition of the server division from an outstanding competitor organization whose fate was declining, board rework and repair activities were conducted in-house, in a dedicated portion of the shop floor known as the “Failure Analysis” (FA) area. Here, specialist tools and training were made available to technicians to ensure that repairs were conducted in a focused and timely manner. This all changed following the onset of the downturn in the electronics industry. Fully embracing a modular production strategy led our firm to focus on systems design and integration, while mostly delegating the production of specialist product components (including computer boards) to their external manufacturers. This provided the organization with the increased flexibility to switch suppliers, as well as increasing or decreasing production following the fluctuations in demand that are typical of this sector (Sanchez & Mahoney, 1996; Sturgeon, 2002), especially at a time of crisis.

There was another consequence to the modularization of production and production-related knowledge, one that is mostly unacknowledged in the extant literature which locates product testing activities entirely within the boundaries of the systems integrator organization (Sturgeon, 2002). Following the outsourcing of board production, key problem-solving routines such as fault-finding and failure recovery, which had been until then conducted in-house, had become distributed across the integrator/supplier boundary. This generated some important implications which we identify by examining the computer board’s rework and repair cycle at the time of fieldwork (see Figure 1).

OPEN IN VIEWER

Figure 1.

Computer Board Rework and Repair Process.

A computer board is produced at the supplier’s facility where it undergoes basic testing. It is then shipped to the systems integrator where it is put through further advanced testing trials. When a fault is found and a board needs to be repaired (or reworked, in technical jargon) it is sent back to the supplier which then issues a new board and sends it to the integrator in exchange for the defective one. The supplier repairs the defective board and sends it out again to the integrator as a new board, as further requests for materials come through. Once it arrives at the integrator, the repaired board is once again scanned and put through the testing process. This process, as observed at the time of fieldwork, appeared to be in all respects very efficient and straightforward. This was until the day in which a rather low-key and seemingly harmless email was circulated around the organization.

Stage 2: The investigation: Reliability gains strength

The investigation analyzed the rework process in the greatest detail and produced a report summarizing the outcomes. Findings confirmed that, within the rework cycle, there were instances where the integrity of information and data flows broke down. This carried significant implications for problem-solving and learning processes, as it created opportunities for faults within the boards to escape detection. In the following section we consider in detail the main issues at play within the rework cycle as exposed by the investigation.

Outcome of the investigation: Vicious cycle of failures

The investigation conducted on the rework and repair cycle thus produced an unexpected outcome: defective boards continued circulating without getting fixed (see Figure 1b).

Once a (still defective) board arrived at the integrator re-labelled as a new board, shop floor practitioners were unable to recall that the board had previously failed. Because the new identity provided a clean slate, the board was not put through Failure Analysis (where repaired boards should have undergone system-level testing for verification), but was instead put once again through production as a new board. Here, the still defective board would almost certainly produce a new fail. Because the board remained the component most likely to have caused the failure, this was once again chosen as a candidate for replacement and sent back to the supplier for repair. There it was not repaired, as the supplier did not have system-testing capabilities. System-level faults did not show up in the supplier’s tests and consequently did not get fixed. The board was re-serialized and returned unrepaired to the integrator only to fail again.

These dynamics resulted in a vicious failure cycle of undetected faults and potentially repeated failures: test techs at the integrator did not know what they were swapping, the supplier did not know what they were fixing, the integrator did not know whether the board that arrived from the supplier was new, or, if it was a repaired board, whether it had been actually fixed. A defective loop was generated, whereby boards kept coming back and forth between the integrator and component supplier but remained unrepaired. In the previous context where debugging was performed in-house, this problem had been prevented by an easy and quick route from the production line to the nearby FA area where the board would get repaired and re-tested (see Figure 1a). The investigation concluded that an element of unpredictability and a source of error had been created as a consequence of bug-fixing having become decoupled across the integrator/supplier boundary.

OPEN IN VIEWER

Figure 1a.

In-house Rework.

OPEN IN VIEWER

Figure 1b.

Rework Outsourced.

Stage 3: Modularity persists but modified

As highlighted by the investigation, modularity theory, and its key rules and assumptions, were being performed fiercely in almost every aspect of the rework cycle where modularity rules had been able to prevail practically unchallenged. The errors brought up by the investigation, however, were about to produce some dramatic changes. The evidence of potential reliability and quality losses was promptly enrolled by those supporting reliability, whose side began to gain traction. From a reliability viewpoint, it was becoming increasingly clear that modularity and complementary rules had made the organization so efficient that it had become too lean to learn, both from its own and the supplier’s errors.

Following the investigation, a new network supporting the reliability argument thus began to emerge. New statistics began to surface that challenged practitioners to reflect on the consequences of undetected failures in production and in the field. Failure information that had once been ignored began to be foregrounded. People in corridors debated the pros and cons of letting errors proliferate uncorrected. Meetings were called that provided opportunities to discuss the possible implications of bringing rework back in-house. At the same time, arguments were being advanced from the modularity side of the debate warning that re-internalizing rework would imply incurring the costs and responsibility of repairing boards which they themselves had not broken, costs which would have normally been incurred by the supplier. These drawbacks, however, appeared less significant compared to the loss of reputation brought about by errors of the “fixing” kind which could potentially result in substantial outages in the field.

An instance of reliability loss was highlighted during the course of a heated debate around an extended server outage incident which had recently occurred at a Japanese customer running mission-critical manufacturing operations. The outage, which had caused extreme disruption to the customer’s time-sensitive shop floor operations, had clearly exposed the risks of server quality and reliability loss. Server failure, in this case, had not only implied unacceptable consequences for the customer itself, which on the basis of that single episode was threatening to rescind the contract, but also for our organization which, following the incident, was facing a significant loss of reputation. During the course of this debate the reliability and modularity coalitions came to confront one another. The reliability side used the example of the outage to support the argument that errors of the kind highlighted by the Japanese incident could not be tolerated. The modularity side, instead, tried to dismiss the incident as minor and the customer as unreasonably demanding. The threats and dangers highlighted by the outage episode, however, were very real and therefore not so easy to downplay. This realization caused some practitioners to change sides during the debate, moving across from the modularity to the reliability-supporting coalition and thus fomenting the growing momentum in favor of reliability.

The Japanese customer incident, and the related discussions highlighting instances of reliability loss in the server domain, therefore produced the effect of strengthening the outcome of the investigation, thus providing further ammunition for the proponents of reliability who were able to canvass increasing support for their arguments. Once appropriated by a network that was rapidly growing in power and recognition, the evidence of actual and potential server failures uncovered by the investigation and related incidents such as the Japanese outage, could therefore no longer be easily dismissed and called for a solution to be reached concerning the extent of integration in the area of rework. The solution that was reached took the form of a compromise among different views and errors and resulted in re-internalizing a limited portion of the rework, pertaining to some of the most common failure causes for that particular board type (see Figure 1a). This solution, which reflected a partial victory for reliability, was supported despite implications for cost (the need to take responsibility for boards damaged on the premises), skills (the requirement to retrain technicians to undertake rework), technical implications (the need to re-acquire tools necessary to perform repair in-house), and despite the fact that this component was classified as critical, but ultimately non-core.

Yet, the solution did not imply that they were ready to simply abandon modularity. While reliability posed a strong challenge, modularity rules had been successfully entangled with practically every aspect of the rework cycle. They were embedded and enacted in the design of the product and the interfaces, in the software tools and methodologies used to manage inventories, in key aspects of the relationship between integrator and supplier, in the distribution of repair competences and skills, in the philosophy that avoided trust and relinquished control, and finally in both formal and informal shop-floor procedures. All these aspects added to the strength of modularity rules and meant that they were hard to overturn. For example, artifacts like the Oracle database or Autoswap, which was the material embodiment of the “swapping” imperative, were difficult to work around. Even those practitioners who had hinted at possible limitations in performing the ideal modular organization, including the Engineer who circulated the unintentionally incendiary email that set off the investigation, interpreted this just as a temporary setback in the path towards full modularity.

The decision to bring some of the rework back in-house, therefore, must be seen in the light of the predominance of modularity and its rules. The compromise solution did not in fact invalidate the theory. Rather, this temporary and partial remedy allowed modularity to continue to be performed (embodied, embedded and enacted) in almost every aspect of the rework process, although in a slightly modified, less perfectly decoupled form. Despite these problems, the practitioners’ belief in the power of modularity remained very much alive.

Modularity framing the rework process (and ensuing errors)

As illustrated by the company history, modularity had been reigning almost unchallenged within the rework process and the wider organization. The performation of modularity depended upon the realization of a specific set of rules, or “conditions of felicity” (Austin, 1962; Callon, 2007) which, for modularity to have an effect, had to frame the rework process. Modularity rules included product design rules, which are “principles that define how an artifact works, what it does and how it is manufactured” (Brusoni & Prencipe, 2006, p. 179), as well as the related organizational principles and properties that are required for the rules to be performed. More specifically, modularity rules included: the modularization of product and organizational architectures, the codification and standardization of product and organizational interfaces, maintaining an arm’s length relationship with suppliers, and performing a clear division of knowledge, problem-solving and productive labor, responsibility and accountability across the supplier/producer boundary (see Table 1).

OPEN IN VIEWER

Table 1.

Modularity and Reliability Rules (Framing) and their Errors (Overflowing).

	“Fixing” (performing reliability) vs. “Swapping” (performing modularity)
Reserializing/loss of identity	Rule	Swapping. No need for coordination. Modular systems “hide information” behind standardized interfaces so that the component’s internal characteristics can be overlooked (Baldwin & Clark, 2000; Langlois, 2002; Parnas, Clements, & Weiss, 1985; Sanchez & Mahoney, 1996; Schilling, 2000; Sturgeon, 2002).
	Error	Fixing. Loss of memory, failure history, loss of traceability, no learning (Fixing => emphasis on “continuous learning”) (Weick, Sutcliffe, & Obstfeld, 1999).
Zero Inventory/Autoswap	Rule	Swapping. Full component interchangeability, “mix and match”. Clear standardization and decoupling of components within a system means that these can be swapped and substituted straightforwardly (Sanchez & Mahoney, 1996). No slack, “zero inventory” (Womack & Jones, 1996).
	Error	Fixing. No redundancy, combined with swapping means no tracking and tracing (Fixing => acknowledging the role of slack, redundancy) (Weick et al., 1999).
Problem ownership/ responsibility	Rule	Swapping. No need to take ownership of/responsibility for failure “shifting the problem”.
	Error	Fixing. Losing track of failures, shifting the problem (Fixing => “owning the problem”, emphasis on “heed”, “attention” and “care”) (Roberts, 1990; Weick & Roberts, 1993).
Sharp div. of lab/ asymmetric views/ uneven access to tools	Rule	Swapping. Sharp division of productive labor (Sanchez & Mahoney, 1996). Outsourcing allows systems integrator to relinquish ownership and responsibility for subcontracted components. Separation of component from system testing
	Error	Fixing. Unable to fault find (Fixing => coordination through “shared ideology” or “mindsets”, “collective mindfulness”) (Vaughan, 1996; Weick et al., 1999).
Arm’s length relationship	Rule	Swapping. Market-like, arm’s length relationship with suppliers (Langlois, 2002).
	Error	Fixing. Not being able to know if the fault has been fixed (Fixing => emphasis on coordination, collaboration, trust, “collegiality”) (LaPorte & Consolini, 1991; Rousseau, Sitkin, Burt, & Camerer, 1998), preoccupation with failure (Weick & Sutcliffe, 2001).
Judgment & problem-solving	Rule	Swapping. Elimination of uncertainty, ambiguity, judgment in problem-solving. A modular organization is “one in which tasks and responsibilities are distributed across separate groups/units, which do not engage in ongoing, real-time communication and joint problem solving” (C. Baldwin, personal communication, September 2010). In fully modular systems, interactions among components and overall system behavior entirely predictable, process of assembling components into a functioning system reduced to simple Lego-like activity.
	Error	Fixing. Fragmented problem-solving (Fixing => joint problem-solving by teams of experts, reluctance to simplify interpretations, Weick & Sutcliffe, 2001).
Formal & informal rules, procedures	Rule	Swapping. Clear and sharp division of knowledge and capabilities across the integrator/supplier boundaries (Sanchez & Mahoney, 1996; Ulrich & Eppinger, 1995). Loosely-coupled modular system (Orton & Weick, 1990) no need to fault-find (exploitation). Formal and informal procedures support automatic switching and shifting responsibility to suppliers.
	Error	Fixing. Unable to know for certain the cause of failure (Fixing => “zero tolerance” on errors, Failure is Not An Option) (exploration).
Systemic interactions	Rule	Fixing. Need for understanding systemic interdependencies, need for systems integration (Brusoni et al., 2001; Hobday et al., 2005). Rapid, uneven innovation (Nelson & Winter, 1982) destabilizes server system.
	Error	Swapping. Unable to unproblematically switch boards in a system (Swapping => no need for system testing (Sanchez & Mahoney, 1996).

A complementary set of efficiency-bearing rules contributed towards supporting modularity, including the principles of “Lean” or “Zero Inventory”, and “Autoswap”, regulating the automatic exchange of boards with the supplier. Throughout most of the company history up until the beginning of the observation period (first stage), this complex set of rules appeared to be firmly in place and was effectively framing the process. Over time, however, another competing theory, reliability, had begun to gain momentum.

In the second stage, modularity began to clash with reliability in the rework process. The reliability focus, a consequence of the acquisition of the high-end server product line, had been coexisting for a while alongside modularity and its complementary efficiency-bearing strategies. Following the economic crisis, however, the increasing pressure towards modularity and leanness at both product and organizational levels had led to the outsourcing of rework. Frictions had begun to emerge between modularity and reliability, culminating with an email being circulated, which exposed important potential instances of reliability loss. Reliability rules required that faults should be detected and traced, errors remedied and failures prevented; boards and all components tracked throughout their life-cycle and system-level diagnostics performed at the supplier’s premises; knowledge and problem-solving integrated and responsibility and accountability shared with the supplier. A set of quality principles complemented the reliability statement, setting the parameters within which the boards needed to operate to guarantee acceptable failure rates (see Table 1).

Beyond rules, our data shows the emergence of another category of elements: “errors” (see Figure 2). Similar to Callon’s “overflowing” (1998), errors followed from the competition between (temporarily) conflicting theories, modularity and reliability, as a consequence of their joint attempts to frame the rework process. Thus, in the first stage, those actors supporting reliability had drawn on a range of failure statistics and charts to expose how performing modularity was generating errors of the “fixing” kind (boards circulating and not getting fixed). In the second stage, modularity supporters had instead highlighted the possibility that performing reliability might produce “swapping” errors (the integrator would pay for fixing boards that it had not broken). Errors were therefore enrolled in the confrontations between modularity and reliability and helped shape the discussion.

Practitioners soon found themselves negotiating between different theories (modularity or reliability) and their (actual or potential) errors. For example, fixing errors were promptly enrolled as evidence of the drawbacks of modularity by actors supporting reliability and were used to strengthen their claims against the prevailing theory. Negotiations led to a compromise in the third stage, where modularity prevailed but was modified. Reliability, nevertheless, had at least partially succeeded in highlighting and problematizing the errors created by modularity, to the extent that practitioners decided to take action. The decision to change the organizational boundaries to re-insource aspects of rework represented a partial victory for reliability (partial because the fix ultimately meant that modularity continued to dominate in all other aspects of the rework process).

Performative struggles and materiality

An important related point is that modularity and reliability rules (and their errors) were able to frame and reframe the rework process by involving a heterogeneous set of material elements (Callon & Caliskan, 2010). Modularity rules, for example, which supported “swapping” (switching and returning a faulty board to the supplier), were embedded in different aspects of the rework process, including tools and technologies (such as the Autoswap system and the Oracle database, which together turned the swapping of faulty modules into the default option); were enacted through the participants physical bodies and ostensive understandings of the rework process and its purpose (i.e., “a swapping routine”); and they were performed through formal procedures and informal heuristics (as in the methodology used by shop floor technicians for diagnosing failed boards). Reliability rules, instead, which supported “fixing” (switching and repairing a failed board in-house), despite being embodied in some of the actors’ abstract ostensive understandings of the process (i.e., a “fixing routine”) did not succeed in enrolling the ERP cycle counting module and RFID technology which would have supported counting and the automatic tracking and tracing of boards across the supplier/integrator boundary and along the rework cycle. The extent to which practitioners were able to mobilize social and material features in support of one theory over another helps explain how a theory might begin to dominate in an organization and the consequences for its practices and boundaries.

The battle between contrasting theories can therefore be understood as the confrontation between multiple competing ordering systems (the theories, their rules and the array of sociomaterial elements that support them). Sometimes the ordering systems underpinning a theory, corresponding to different “logics” (Greenwood, Raynard, Kodeih, Micelotta, & Lounsbury, 2011; Purdy & Gray, 2009), interpretations, or understandings of the process (Bechky, 2003; Carlile, 2004; Dougherty, 1992), plus the sociomaterial infrastructures that support them (or “logistics”) may overlap substantially (Berg & Timmermans, 2000). Yet at other times, as our case illustrates, they may be opposing each other to a greater extent. The outcome, in this case, was an “amalgam of logics” (Berg & Timmermans, 2000, p. 43) as exemplified by the compromise between a “swapping” and a “fixing” logic, and the related ostensive understandings of the rework routine, eventually reached in the rework area.

What did this compromise look like and how did it configure the rework process? While modularity, by embedding itself in most aspects of the rework process, had prevailed over reliability and retained its dominant position in the rework process, it was clear that the latter had managed to secure at least one crucial victory. The hidden “fixing errors” that had been highlighted by the performative struggle against reliability, embedded in the material form of failure stats and charts used as the basis for negotiations, had been acknowledged and addressed by bringing rework back in-house. This brought renewed (albeit temporary) stability to the process, as reflected in a renewed balance of power between the two competing theories. Modularity’s main rationale, “swapping”, was therefore able to continue to thrive, not despite, but because it could draw upon its errors (“fixing”).

Organization modifies the theory

Performative struggles therefore resulted in changes to the organization which became more closely coupled than modularity’s abstract ideal of a “perfect integrator”—which our practitioners were pursuing—would have allowed. The outcome of the competition between two theories in the context of the rework process, however, also implied that modularity itself was modified as a result of the performative struggle against reliability. This particular instance of performation of the theory had challenged—and at least marginally deflated—the company’s unwavering faith in the power of modularity. Or, perhaps more precisely, it had suggested that a more sophisticated temporary truce was required at this point in time, in this particular organization, between efficiency- and reliability-bearing strategies, two important (usually complementary, but in this particular case temporarily conflicting) sides of the lean production philosophy (Womack & Jones, 1996). As a consequence, modularity was therefore still being performed, albeit to a lesser extent. The performative struggles between modularity and reliability over time and in the context of this organization had eventually resulted in the modification of the theory itself.

Theories (and their errors) perform organizations

The first key contribution resulting from studying modularity from a performative perspective is a theoretical framework that captures the more complex, dynamic and nuanced effects of modularity rules on the organization. As our evidence shows, while modularity rules can indeed be fully prescriptive or merely descriptive, they are often performed by actors to different levels and with varying outcomes. We therefore add to those contributions in the modularity literature which have offered opposing arguments for modularity theory—i.e., that it either succeeds or fails to “design” (Brusoni & Prencipe, 2001) organizations—by showing how and how far the theory is enacted and shapes the organization, and how it may be shaped in return. In our case, the evidence shows that modularity was being performed successfully in most aspects of the rework process. We can therefore place the outcome of performing modularity, in the rework context and during the period of observation, at the “strong” end of the performativity range (D’Adderio, 2008; MacKenzie, 2006).

At the same time, however, our evidence also shows that the “translation” (Callon, 1986) of modularity rules into a modular organization was incomplete. In framing the rework process, modularity rules had created errors which were promptly enrolled by the competing paradigm, reliability, and used to modify both the theory and the organization. The re-internalization of a portion of rework, in response to fixing errors, had thus allowed modularity to prevail, albeit in a modified form, marginally closer to the integral than the fully decoupled end of the spectrum. “Errors” (or, in performativity terms, the overflowing) have been a neglected aspect of modularity studies, having been implicitly viewed as either the outcome of the theory not being fully or correctly implemented into technological and/or organizational interfaces (the mainstream approach); or as the result of the failure of the theory itself to account for organizational and technological complexity (the contingency approach). Through our performative lens, instead, we were able to theorize the emergence of errors not simply as a failure of technology/organization to reproduce the theory or a failure of the theory to reflect the complexity of technology/organization but as the continuously emergent outcome of performative struggles among competing theories in their context. The emphasis on context has also provided important insights into the organization’s ability (or lack thereof) to harness overflows and use them to reconfigure its boundaries and/or the theory.

In further theorizing the role of errors, our findings also add to performativity-inspired approaches in organizational theory (Cabantous & Gond, 2011), and the discussion of “overflows” in performativity theory (Callon, 1998). In the first case, our results add to the extant theory by showing precisely how overflows may (or may not) contribute to constituting and modifying organizations and their theories. In the second case, we shed new light over the possible sources of overflows. Whereas the extant theory (Callon, 1998) has focused on the emergence of overflows as resulting from “imperfections” within the frame, we demonstrate that overflows may be at least partially constituted outside the frame, from a theory’s interactions with other theories operating within and beyond the organizational context.

Performative struggles shape organizational boundaries

Our second core contribution therefore consists in theorizing the coexistence of modularity with competing theories, their performative struggles and their implications for the organization and the theory. Our in-depth, longitudinal study set in a contemporary high-technology organization exposed to multiple organizational and institutional pressures showed how modularity did not operate alone but instead coexisted together with other theories which were also attempting to configure the organization at the same time. As our evidence illustrates, within organizations and beyond, theories rarely exist in isolation. Rather, organizations are typically configured by multiple ordering systems, including a plurality of theories, each with its own rules and assumptions. Some theories may be mostly complementary (as, in our case, modularity and zero inventory principles, or reliability and quality), whereas others may be largely in conflict (as the efficiency-bearing modularity rules and reliability). To complicate the picture, alliances among theories may shift quite significantly both over time and in the context where they are enacted.

The presence of competing theories, and their influence over organizational practices and boundaries, has been largely neglected within studies of modularity. Building on the case of a diverse and distributed organization, and drawing on performativity’s emphasis on plurality and multiplicity, we were able to add to extant contributions by revealing how different sets of rules do not always merely substitute one another, but often become involved in dynamic confrontations. While at times confrontations can indeed result in the replacement of one rule, which becomes non-performative, for another, which becomes performative (as in Brusoni & Prencipe, 2006), at other times their outcomes tend to reflect a synthesis, involving mutual dynamic adaptation under conditions of uncertainty. Our results therefore add to the modularity debate by theorizing the dynamics underpinning performative struggles (Callon, 2007) and how they shape both theory and organization. In addition, as theories evolve through competition, their influence varies over time and across organizations, so that any outcomes can only ever be emergent. It is therefore the dynamic competitions among contrasting theories and their networks, which include but span beyond abstract rules (Sanchez & Mahoney, 1996) and knowledge (Brusoni et al., 2001), which ultimately modify theories and “design organizations”. While adding to modularity, this finding also responds to the recent call for studying how different “theories interact and influence organizational practices” and shape organizational behaviors and outcomes “through performativity struggles” (Cabantous & Gond, 2011, p. 582). While Callon (2007) has advocated for the need to re-imagine organizations and markets as sites for performative struggles, “this perspective has yet to be explored” (Cabantous et al., 2014). But what are the micro-level dynamics which define how theories compete?

A related finding is that, to exert its influence, modularity had to build a scaffold, including an array of social and material features (similar to Callon’s agencement) which supported the realization of its rules in practice. Modularity and complementary efficiency rules, for example, were so thoroughly embedded in texts, technologies and artifacts that there were serious limitations on how far actors could operate “outside” the theory. A neglected aspect of current theorizing, capturing the sociomaterial aspect of rules performation, or “how [a] theory is translated into tools and material devices” (Cabantous & Gond, 2011, p. 583) thus further adds to current explanations for the success of modularity beyond abstract rules and knowledge. A performative framework therefore allows us to capture how the practices and boundaries of the modular organization are continuously challenged and constantly shift to reflect the particular array of (intra- and inter-organizational, social and material) relationships that produces them at any one time and over time. In so doing, our findings also add to performativity-inspired approaches to organizational theory (Cabantous & Gond, 2011) by capturing how various sociomaterial assemblages, supporting different knowledge, rules and assumptions and affording different practices or kinds of “calculation” (Callon & Muniesa, 2005), may confront one another in their effort to support competing theories, often influencing their organizational outcomes. Sociomateriality can therefore help us understand how a theory may be able to shape routines, practices and boundaries and contribute to “designing” an organization. Our evidence, in addition, also suggests that the organization itself might in return participate in “designing” the theory.

The mutual production of theory and organization

Our final key finding concerns the mutual shaping of theory and organization. Our case study has highlighted the subtle ways in which modularity rules might be able to (re)configure the organization, its structure, practices and boundaries, while also shedding light over how, in turn, the situated enactment of modularity in this particular organization ended up modifying the theory. In the first case, concerning the influence of theory over organization, we have seen how the company’s vision of modularity as an efficiency-seeking strategy aimed at standardizing and distributing products and organization and therefore supporting superior leanness and cost reduction, conflicted considerably with reliability, which instead demanded redundancy and the substantial overlap of production knowledge and activities.

In the second case, concerning the influence of the organization over theory, we have instead seen that, while modularity did manage to frame the rework process, at the same time it prompted the emergence of errors. These were readily enrolled by the competing rationale, reliability, which was able to win a marginal victory in the performative struggles between the two theories. This indicates that performing modularity had produced an extraordinarily lean and efficient organization, which was however brittle and fragile, and therefore prone to failure. The threat of failure, as highlighted and problematized by the supporters of the reliability argument, was considered to be such that it induced practitioners to revise their vision of modularity. As a consequence of the confrontation with reliability, modularity was modified. In capturing and theorizing the mutual shaping of theory and organization, this finding adds to both studies of modularity (Baldwin & Clark, 2000; Brusoni & Prencipe, 2006; Brusoni et al., 2001; Sanchez & Mahoney, 1996) and wider approaches concerning the performativity of (organizational) theories (Cabantous & Gond, 2011; Cabantous, Gond, & Johnson-Cramer, 2010).

Our findings beg a further and final question. Will the lessons from this outstanding organization, which had such a deep historical relationship with modularity, also succeed in modifying the wider theory, namely the scholarly debate surrounding modularity? While we are unable to predict any specific outcomes at this stage, we believe that it will depend upon the willingness of both the mainstream and the contingency sides of the modularity debate to engage with our results and novel approach, which at the same time both reconciles and adds to existing approaches. It would be interesting, for example, if this outstanding organization which has initially sparked the academic interest towards modularity, would end up once again teaching us scholars some important theoretical lessons about the mutual shaping of theories and organizations.

Pushing our insights even further, we can foresee promising grounds for future investigation that may emerge from our novel approach. We envisage a new stream of studies investigating how the struggles among different (contradictory and complementary) theories or strategies, and their associated rules, rationales and logics, may play out and transform different organizational contexts. This would add to extant approaches that have focused on how different rationales or logics play out within the same theory (as in Cabantous & Gond’s 2011 example of rational choice theory). Further advances that build on our framework might also include shedding new light over the emergence of macro theories and their organizational effects. Among the interesting features highlighted by our case evidence and related discussion, for example, is the hypothesis that the confrontation between modularity and reliability followed the unexpected emergence of tensions between two different (but usually complementary) sides of Lean Production theory, the “efficiency”, or “cost reduction” side and the “zero defects” side. Taking this insight further would prompt a timely investigation into the emergence of Lean Production as a meta-theory, incorporating both efficiency and reliability principles, albeit in different proportions. The question to arise then would be whether it was the imbalance created in our case by the overemphasis on the efficiency side of lean that brought two usually complementary features to clash irremediably. One possible answer, we posit, is that incorporating the errors and re-internalizing rework had the effect of temporarily restoring a truce between the two sides of this influential management theory, a truce which had been lost following the increasing emphasis on efficiency brought about by the electronics sector crisis.

While a comprehensive analysis of this topic lies beyond the scope of this paper, results indicate that our analytical framework based on performativity can contribute more generally towards advancing our understanding of fundamental aspects of contemporary organizing. In capturing the mutual constitution of multiple theories and organization, our approach begins to unravel the micro-level, sociomaterial mechanisms through which organizations are able to manage apparently unsolvable conflicts, contradictions and paradoxes (Adler et al., 1999; Gibson & Birkinshaw, 2004; Poole & Van de Ven, 1989) while dynamically adapting to change. Especially in tough and fast changing competitive environments, this crucial ability lies at the very basis of an organization’s potential to survive and prosper.