Temporary De-Embedding Buyer-Supplier Relationships: A Complexity Perspective

Research on buyer-supplier relationships has debated the advantages and disadvantages of embedded relationships. We join this debate by developing theory on the performance implications of relaxing embedded buyer-supplier relationships for a limited period of time — a previously neglected phenomenon we refer to as temporary de-embedding. To capture this phenomenon’s dynamic and complex nature, we use a combined-method approach. First, we conducted a longitudinal case study of the relationship between Nissan and a strategic first-tier supplier. This case study suggests that temporary de-embedding reinvigorates search and leads to higher performance for both the buyer and supplier. Second, we built a computational simulation model using the search perspective from complexity theory to complement the theory grounded in our case study. Our simulations confirm the case findings while shedding additional light on how frequency, duration, and intensity of de-embedding affect supply chain performance.


Introduction
Supply chain management is inherently complex and dynamic (Nair et al., 2009) because decisions made by one member of the supply chain affect subsequent decisions of other actors. In this dynamic and complex setting, the existence of an "optimized" masterplan proves elusive.
Instead, supply chain members engage in co-evolutionary search to advance and to innovate (Chandrasekaran et al., 2015, Giannoccaro, 2011, Kim et al., 2015, Levinthal, 1997, Sting and Loch, 2016, much like BMW's ongoing "Industrie 4.0" initiative to digitalize manufacturing processes and technologies. As part of this initiative, BMW has scanned its entire Rolls Royce plant in Goodwood, UK within a two-millimeter tolerance. Supply chain gains from such firmlevel improvements, however, critically depend on suppliers' compatible interfaces. BMW COO Zispe reflects: "How will our suppliers connect with these emerging systems?" (Mayer and Klein, 2015). This quote illustrates how pivotal buyer-supplier relationships are for supply chain innovation and performance (Chen and Paulraj, 2004, Choi and Kim, 2008, Kim et al., 2015, Terpend et al., 2008. Surprisingly however, little is known about how buyer-supplier relationships affect co-evolutionary search processes, let alone how these processes subsequently drive supply chain performance-the main motivation for this study. Prior research has shown that closely embedded buyer-supplier relationships foster joint problem-solving activities and information exchange (Dyer and Singh, 1998, Dyer and Chu, 2000, Gulati and Sytch, 2007, Uzzi, 1996, which, in turn, boost buyer performance (e.g., Cachon and Lariviere, 2005, Choi and Kim, 2008, Chopra and Meindl, 2007, Kim et al., 2015. Yet, embedded relationships can also trigger complacency, limit access to non-redundant information, and lead to poor performance for both the buyer and supplier Zsidisin, 2006, Villena et al., 2011). Consequently, prior research has often advised a constant, moderate degree of Electronic copy available at: https://ssrn.com/abstract=3058630 fact that firms can strategically relax, and later re-instate, embedded ties-a phenomenon we refer to as temporary de-embedding. Developing a theory for temporary de-embedding is our goal.
To this end, we employ a combined-method research design-a prerequisite for a broader understanding of complex supply chain management phenomena (Boyer and Swink, 2008). First, we collected longitudinal and dyadic data at Nissan and a strategic first-tier supplier over a 12-year timespan. We examine how the relationship's embeddedness evolved over time and how that evolution interacted with search and performance. We observe that when the relationship was over-embedded, both firms were limited to only incremental improvement initiatives that proved insufficient in breaking the relationship's deadlock that was freezing each firm's innovative potential. De-embedding, in contrast, shifted priority to intra-firm goals over inter-firm goals, reinvigorating both Nissan's and the supplier's independent search initiatives. Thus, de-embedding helped the supply chain to escape its sticking point. Notably, five years after the abrupt commencement of its de-embedding effort, Nissan opted to re-embed supply chain ties in order to ensure compatibility among independent search initiatives.
Secondly, to generalize and augment the theory emerging from the case study, we devised a computational model to simulate de-embedding in supply chains. The model uses the search notion of complexity theory where a supply chain gradually explores a rugged performance landscape in search of improvement and innovation. Generalizing the case findings, we show that temporarily de-embedding supply chain relationships enhances performance by promoting broader search for improvements in complex environments. Nevertheless, de-embedding can be a doubleedged sword: an intense cut reinvigorates search, but too frequent or prolonged cutting of ties leads to incompatible outcomes that can hamper supply chain performance.
We offer several contributions to research on buyer-supplier relationships. First, we investigate the phenomenon of temporary de-embedding and develop theory on how it affects supply chain performance. This is important since prevailing research advises balanced degrees of embeddedness (e.g., Uzzi, 1996, Villena et al., 2011 while largely ignoring the question of how an ideal balance can be achieved. We address this issue by proposing a dynamic balance. More specifically, we argue that the level of embeddedness can be altered dynamically over time to reinvigorate supply chain innovation and performance. Second, we integrate supply chain embeddedness with complexity theory and its fundamental notion of search. The search perspective offers new theoretical insights on the outcomes of buyer-supplier relationships that go beyond the current explanations based on transaction cost economics (e.g., Williamson, 1985), the relational view (e.g., Dyer and Singh, 1998), and social network theory (e.g., Gulati and Sytch, 2007, Uzzi, 1996. Third, we propose an agency view of buyersupplier relationships. This view qualifies extant approaches that consider these relationships as either relatively stable (e.g., Villena et al., 2011 or as following a presaged course (e.g., Jap andAnderson, 2007, Vanpoucke et al., 2014). Instead, we explicitly recognize the agency of firms in deliberately and strategically tuning their relationships. Supply chain researchers can benefit from this fresh agency viewpoint because it sheds light on the pivotal role of endogenous changes in buyer-supplier relationships.
Supply chains exemplify such complexity because numerous decisions interact among supply Electronic copy available at: https://ssrn.com/abstract=3058630 chain members, and organizations cannot fully oversee or understand, let alone globally optimize, all decisions simultaneously (Levinthal and Warglien, 1999). Instead, organizations dynamically explore the landscape of possible actions in an evolutionary, path-dependent search process. While search processes have been discussed in complex manufacturing and high-tech settings (e.g., Chandrasekaran et al., 2015, Sting andLoch, 2016), to the best of our knowledge this lens has not been applied to supply chain embeddedness. Given the dynamic, co-evolutionary, and complex nature of buyer-supplier relationships, this omission is surprising since the strength of such relationships is likely to affect the way supply chain members search jointly (Kim et al., 2015).
Buyer-supplier collaboration and knowledge exchange were shown to be key in managing complexity (Kim et al., 2015). Such relationships can vary from strongly to weakly embedded.
Embeddedness strength is shaped by structural, relational, and cognitive factors. Structural factors include joint projects, operational assistance, and cross shareholdings (Clark and Fujimoto, 1991).
We investigate the phenomenon of temporary de-embedding of buyer-supplier relationships defined as relaxing embeddedness for a limited period of time. Temporary deembedding thus implies the dissolution and subsequent resumption of embedded ties. This concept does not denote the state of a relationship but, rather, a process of deliberately altering the relationship. In this way, temporary de-embedding advances prior research by offering a dynamic and malleable viewpoint.
To delineate the nomological network of embeddedness and temporary de-embedding, we reviewed related concepts such as reciprocal interdependence (Thompson, 1967), near decomposability (Simon, 1962), weak ties (Granovetter, 1973), and loose coupling (Orton and Electronic copy available at: https://ssrn.com/abstract=3058630 Weick, 1990). Reciprocal interdependence refers to one party's decisions influencing another party's decisions; near decomposability entails the grouping of decisions that are strongly interdependent. These concepts thus tally strengths of interdependence in a system, whereas embeddedness describes relationship strength. As such, embeddedness is composed of structural (e.g., concentration of business and equity stakes), relational (e.g., frequency of interaction), and cognitive (e.g., shared norms) dimensions. Loose vs. tight coupling refers solely to the degree of operational integration, hereby capturing only the structural dimension of embeddedness (for a recent discussion, see Kim et al., 2015). Therefore, embeddedness uniquely offers sufficient conceptual breadth. What is more, embeddedness, coupling, and tie strength all denote merely the state of a buyer-supplier relationship, while de-embedding refers to changing it. In the following sections, we develop theory on temporary de-embedding.

Case Selection and Method
We conducted a longitudinal, inductive case study of the relationship between Nissan and one of its strategic first-tier suppliers (referred to as "the supplier") during the period 1999-2012. 2 Our unit of analysis is the relationship between Nissan and the supplier. Our approach allowed us to trace different periods of embeddedness in real time while assessing the improvement and performance implications for both the buyer and the supplier.

Case selection
We identified Nissan as a critical setting (Barratt et al., 2011) because Nissan deliberately and publicly adjusted the level of embeddedness with its suppliers during the period of our study.
This adjustment targeted its entire supply base: Nissan purposefully discontinued the "Japanese way" of managing supplier relationships-a practice that has played a prominent role in the literature on close coordination, and that had received admiration since the 1980s when Japanese carmakers overtook their American counterparts.
We selected one supplier across different embedding periods in order to gain in-depth understanding of one temporary de-embedding process. Furthermore, we selected the supplier ourselves rather than asking Nissan, in order to avoid social desirability bias. Because we assured anonymity to the supplier, we cannot disclose the component delivered by the supplier. Our sampling criterion was that both companies had to view their relationship as strategic. First, from Nissan's perspective, the supplied component's strategic relevance is reflected by its "just-in-time" and "in-sequence" delivery from co-located supplier factories directly to Nissan's production line.
Also, the business volume between Nissan and the supplier has been substantial throughout our research period. During the period preceding the NRP, the supplier accounted for approximately one third of Nissan's domestic demand for this particular component. Moreover, there is a direct interface between the end user of the car and the component, which has an important impact on driving experience. Second, from the supplier's perspective, Nissan was its most important customer. During the 1990s Nissan accounted for one third of the supplier's production volume.
To encourage supplier interviewees to talk freely, we did not mention their company's name during our interviews with Nissan, choosing instead to query Nissan interviewees about all three main suppliers of the component in question. The length and in-depth nature of our interviews enabled us to gather sufficient information from Nissan about our focal supplier.

Data collection
Our data consist of official company reports of both Nissan and the supplier; patent data; as well as coverage in the media and business press. To better understand de-embedding at the Electronic copy available at: https://ssrn.com/abstract=3058630 systemic level, we gathered such data also from other Nissan suppliers. We collected these data continuously from the announcement of the NRP in 1999 up to 2012. In addition, we conducted face-to-face interviews over a 12-year period. At both Nissan and at the supplier, we interviewed key employees central to the relationship-several times when needed. With the exception of Renault-Nissan's CEO Carlos Ghosn, all interviewees spoke on the condition of strict anonymity. Table 1 provides a summary of our interviews. These interviews were conducted by the second author who is fluent in Japanese. The interviews with CEO Carlos Ghosn and one non-Japanese Nissan employee were conducted in English in Japan. Two senior executives and three senior purchasing managers were interviewed in English in France. All other interviews were conducted in Japan in Japanese. We designed the interview protocol in English and translated it into Japanese. A native Japanese speaker translated the protocol back into English to fix minor issues. The duration of the semi-structured interviews varied between 30 minutes and five hours, leaving ample room for new topics to arise.
Most interviews were taped, and all taped interviews were transcribed verbatim the same day. When interviewees declined to be taped, detailed interview notes were taken. Japanese interviews were translated into English by the interviewing author with a Japanese native speaker subsequently double-checking the translations of the main episodes of the recordings. Any unclear content was resolved during subsequent meetings or through follow-up phone calls or emails.

Data analysis
Our interest in embeddedness took hold when learning through the media of Nissan's deembedding moves in 1999. As a first step, our study analyzed Nissan's public announcements as well as media coverage of the NRP. Recurring mention of "destruction of the keiretsu" and references to Ghosn as the "keiretsu killer" pointed us to extant research on the keiretsu approach to organizing the supply chain (e.g., Dyer, 1996, Nishiguchi, 1994, Sako and Helper, 1998. Based on these sources and our interest in supply chain relationship performance, we drafted our interview protocol of semi-structured questions (see Appendix A). Our aim was to capture all main dimensions of buyer-supplier relations while allowing enough room for elaboration as well as the introduction of new topics by our interviewees. Apart from data about the pre-NRP period, all information was collected in real time. We asked respondents about the period "before 1999" during all interview rounds. We compared the responses given in each period about the pre-1999 period with each other and found no noticeable differences, implying that retrospective error is not a concern with respect to our understanding of this period.
The initial analysis of our interview data revealed two major triggering events that "bracket" the different periods of our study (Langley, 1999). The first triggering event was the We proceeded with an in-depth analysis of each of these periods. Following Gioia et al. (2012), we first relied on "open coding" by adhering closely to informant terms. We grouped similar remarks together to come to our first-order codes. In the next step we axially coded these informant-centric first-order codes, resulting in our second-order codes that capture commonalities in our first-order codes. We proceeded to substantive coding of our data into aggregate dimensions, using concepts from the buyer-supplier relationships and search literature as a point of reference.

Case Findings
This section discusses the three embeddedness periods: (1) embedding, (2) de-embedding, and (3) re-embedding. We also detail the two triggering events that mark these periods, describing the relationship and the outcomes for both Nissan and the supplier.

Relationship
A well-known characteristic of the keiretsu approach is shareholding in suppliers; a system-wide practice also followed by Nissan (Cusumano, 1985). Our case is no exception: in the pre-NRP period Nissan held a 20 percent equity stake in the supplier. Figure 1 depicts Nissan's ownership in the supplier as well as in 20 other strategic first-tier keiretsu suppliers. We observe that Nissan's average equity ownership in any of its suppliers did not change drastically before the NRP. Another characteristic of the keiretsu system is the co-location of supplier factories with buyer factories. Our case also exemplifies this practice and the first plant and headquarters of the supplier were built in the direct vicinity of a Nissan factory. As the relationship evolved, Nissan assisted the supplier in opening co-located production facilities at all of Nissan's six domestic mass production plants.
Also, Nissan dedicated senior purchasing managers to serving ties with each of its affiliated suppliers. When nearing retirement, these purchasing managers moved from Nissan to comfortable positions at these suppliers-a practice known in Japan as tenseki. This practice helped Nissan since the senior purchasers would switch to the supplier's payroll, while the supplier benefitted from the close ties with Nissan of these ex-Nissan envoys. Having assigned purchasing managers to a supplier firm rather than to a component group often incentivized Nissan purchasers to champion "their supplier". Suppliers, in turn, invested heavily in ensuring that "their buyer" remained closely tied to them. By meeting for after-hours drinks, dinner, or overnight trips (including trips abroad), and through personal gifts from the supplier, work duties and social lives became deeply entangled. Shaded area indicates the standard deviation of ownership. Dotted line represents equity in our case supplier. Data were retrieved from Nissan and suppliers official company documents.
We also observed that Nissan and the supplier referred to the supplier as Nissan-kei (Nissan group). 3 In addition, the keiretsu practice of referring to Nissan as oyagaisha (parent firm) or oyabun (parent part) and to the supplier as kogaisha (child firm) or kobun (child part) was used.
Amplifying the family analogy, Nissan's wholly owned supplier was dubbed Nissan's first wife, while the supplier called itself Nissan's second wife (using English terms). The usage of familial terms signaled a far-reaching commitment where each party internalized a prescribed relationshipserving role to enact. To illustrate, our interviewees reflected on the keiretsu period with comments such as "this is the way it always has been," or "we simply belong to Nissan." 3 Kei is used to signal group membership. It is the same Japanese character kei as the first syllable in keiretsu, and means "linked." Looked at from the supplier perspective, this affiliation is exclusive; suppliers are not affiliated with more than one OEM. It is based on the history of the supplier and generally does not change over time, even when the customer base of the supplier changes.

Improvement and Innovation Initiatives
Nissan's focus was improving manufacturing excellence as a goal for its supply chain relationships in themselves, expecting suppliers to improve exclusively through joint, incremental process advancements (kaizen). In case of the focal supplier, it would first be awarded a specific contract, and subsequently be expected to achieve cost reduction targets-all within the dominant design and in concert with Nissan. Informants at both sides of the relationship reported joint improvement initiatives, noting that such improvement initiatives delivered incremental outcomes (i.e., adapting to smooth out existing problems in the production system) rather than radical or proactive ones (i.e., generating ideas that would advance the supplier's technological capabilities).
Cross-organizational improvement teams staffed by senior individuals from both partners focused on retaining harmony (wa) rather than on generating radical ideas. Thus, joint efforts were geared towards preserving the dominant design at a loss of momentum as reflected by the supply chain products' reputation of "smelling of old men"-an image often cited by our interviewees both at the supplier and at Nissan.
Our interviewees at the supplier stressed that during the embedded period, they were obliged to obey their "parent" and follow its course. This, in turn, relieved the supplier of making strategic decisions such as which technologies or markets to pursue. A manager at the supplier diagnosed the technological standstill in the following way: With shigarami [strong bonds] in keiretsu, it is really hard to create new technology, and this is needed to win in the new market. Shigarami inhibits progress.
At the systemic level, internal supplier championing at Nissan frequently led to the rejection of new suppliers and technologies by Nissan's purchasing department (Sugiyama, 2002).
Beyond the focal relationship, this "lock-in" was evidenced by Nissan's refusal during its downturn to change its supply base in terms of the number of suppliers. This was the case even when production at its seven domestic factories dropped to about 50 percent of capacity during the 1990s. Illustrating this point, a senior executive at Nissan described the snares of the gridlock into which both Nissan and its supplier were trapped: The pre-99 situation was unsustainable in its coziness and inefficiency because of a lack of tension-it was far too comfortable. There was no tension to promote performance improvement.
In a situation where relationships will be too long-term, there are barriers to entry for some of the new suppliers or more agile suppliers. Can you imagine in this situation […] the sourcing of components in low-cost countries, or leading-cost countries? It would not have happened because there is no drive-no impetus to change.
One startling drawback of this combined technological and relational lock-in was Nissan forgoing the technological opportunity of hybrid electric engines. Main rival Toyota had launched the hybrid electric Prius in Japan in 1997, but Nissan did not have the required capabilities inhouse, nor in its keiretsu, to enter this market. 4 Ultimately, Nissan's embedded supplier relationships coincided with poor efficiency and slowing growth. For example, Ghosn said during our 2001 interview that a benchmark study of component prices revealed that Nissan had been paying 20-25% more for its parts than its alliance partner Renault. These observations became the basis for the subsequent radical reform.

Nissan Revival Plan: De-embedding
Nissan's belief that keiretsu-style relations had become a liability led it to craft the NRP.
At its announcement, Ghosn stated one maxim: "No sacred cows, no taboos, no constraints." Nissan's expectations regarding the supplier had changed dramatically, as summarized by a senior Nissan purchasing manager: The ability to deliver high quality products at a low price now really has become the responsibility of suppliers. Nissan is not willing to just pour money into the suppliers anymore … QCD [Quality Cost Delivery] is the basis of the sourcing decision. The trend is very much discrete: "Do you want this? You are in straightforward competition-best price, you get it; otherwise, you won't."

Relationship
Nissan divested shares in all but four of 1394 affiliated firms (see Figure 1). During our interview with Ghosn in 2001, he reflected on shareholding in keiretsu times the following way: Before, the relationships were confused by shareholdings. You felt obliged to buy from a company because you owned part of this company.
Nissan further announced that the total number of component suppliers would be reduced to 600, those suppliers were obliged to reduce the price of their parts delivered to Nissan by 20% within three years. Our focal supplier was even asked for a 26% price reduction of its components.
Also at the dyadic level radical changes were executed. Pre-NRP practices of Nissan engineers aiding kaizen process improvement activities at the supplier's facilities were abolished. Likewise, Nissan no longer trained supplier personnel at its own facilities. Furthermore, some supplier facilities near Nissan factories were shut down by the supplier.
After-hours socializing among Nissan purchasers and supplier representatives was banned by Nissan, which severely affected personal relationships. Mid-year gifts (o-chūgen) traditionally exchanged in Japan with key business contacts were shunned by Nissan, poignantly symbolizing a rejection of the "Japanese way." Such radical changes earned Ghosn the nickname "keiretsu killer" (Nezu, 2000). Ghosn made clear that preferential treatment of suppliers was now deemed a bad practice. According to Ghosn: Sourcing decisions in all categories will be strictly based on credible performance commitment. This clearly means that sourcing from our affiliates will be no exception to this guideline.
In addition, Nissan's purchasing department was completely overhauled. The previous structure with purchasers dedicated to one specific supplier was replaced with purchasers in charge of component groups. As a result of this system-wide change, the supplier lost its champion at Nissan and eventually had to negotiate with a French purchaser in Paris who had broader knowledge of all offers in the field. Because socialization between purchasers and supplier personnel was strictly forbidden during this period, in one stroke, the supplier lost its access to information regarding developments at Nissan.
Our interviewees at the supplier described the relationship with Nissan as having changed from a parent-child relationship to "having become adult" (otona ni natta). This change was also reflected in the business language used. During our interviews, we noted that interviewees at the supplier referred to the pre-NRP period as Nissan "offering" or "giving" them business; post-NRP, they talk about "winning" business.

Improvement and Innovation Initiatives
In its efforts to build a globally competitive supply base, Nissan committed itself to evaluating any supplier, incumbent and new, according to exactly the same standards. The supplier responded immediately after the NRP had taken effect by channeling its improvement activities into its own "Survival Plan." This plan was a strategic roadmap to "design products that cost less." First, the plan outlined the supplier's increased contribution to product development and design. Incorporating these value-added activities was paired with building in-house capabilities in computer aided design (CAD). A manager at the supplier explained that before the NRP, they relied on their second-tier suppliers for design. After the NRP, they increased their investments in R&D and prototyping.

Re-embedding
A few years into the NRP, Nissan's more flexible and efficient approach to supplier selection began to face the strain of weakened coordination. One striking example was the Japanwide steel shortage in November 2004. Nissan's domestic competitors all received their ordered quantities and could continue production as planned (Ibison and Fifield, 2004). Nissan, however, had cut the number of steel suppliers from five to two during the NRP, and had drastically decreased the volume purchased from its previous keiretsu steel supplier to below 10 percent (Itoh, 2001 Consequently, Nissan became increasingly concerned about reduced supplier responsiveness.
According to a senior Nissan purchasing official: In the past … we could ask suppliers without a commercial negotiation "Can you make sure that you work Saturday morning this week because we just had a breakdown?" or "We want to bring some visitors around-sorry you got only half-an-hour's notice-they'll be arriving at 4 o'clock." But we probably have seen some deterioration in responsiveness to change. Now, it's a bit more formal: "Who is paying for Saturday morning, please, before we commit?" "Is this visitor really important?" Or: "We are too busy, can we put it off to next week?" Or: "Can't you tell him to go see a different supplier?"

Relationship
These developments triggered the second change in the relationship: a re-appreciation of embedded ties with suppliers announced by Ghosn at a meeting with suppliers on November 18 th , 2004. According to Ghosn: "from now on, it is necessary for Nissan to strengthen cooperative ties [with suppliers]" (Nikkei Industry Journal, 13 January 2005). Ghosn also publicly stated a reappreciation of the keiretsu approach: "Not everything about the keiretsu was wrong. It simply did not function properly at Nissan in the past. With Nissan's subsidiaries, the keiretsu system was too cozy" (Mikawa and Okudaira, 2005).
Subsequently, a number of initiatives were aimed at strengthening the relationship with suppliers. First, through the Alliance Supplier Improvement Program (ASIP), Nissan began to offer specialized, on-site production process support to its suppliers. Subsequently During this time, the supplier also opened engineering facilities in France close to the technical center of Nissan's partner Renault. In Japan, the supplier started to attend design and engineering meetings organized by Nissan at Nissan's technical center. Amongst our supplier interviewees, the belief that Nissan had regained technological capabilities that were worthwhile learning from started to take root again. Moreover, they believed Nissan's initiatives to regain a close relationship were not only sincere, but would also directly benefit the supplier because Nissan had successfully fended off its crisis and seemed to be on a solid trajectory of recovery, particularly abroad. One supplier interviewee explained their interest in renewed closeness to Nissan in the following way: Before, [Nissan] purchasing people were very lazy. With the arrival of Carlos Ghosn they all changed their way of working dramatically. And that was a very good thing for the company I think, because Nissan is very profitable today.
Some of our interviewees at the supplier however were skeptical about Nissan's attempts at closer coordination. They viewed these initiatives as a Trojan horse that would give Nissan access to their facilities and information systems, which they believed would be used to enforce further cost reductions. According to one of our interviewees at the supplier: Nissan wants to have transparency to the supplier, but not the other way. They want to have transparency to find an opportunity for reducing costs. That's the motivation for asking us to be transparent.
As exemplified by the two quotes above, supplier employees' opinions on their firm's reembedding with Nissan were mixed. Concluding we see that despite the reservations, the supplier took the opportunity to expand internationally alongside Nissan, and engaged more frequently and closely with Nissan.

Improvement and Innovation Initiatives
In addition to Nissan's operational assistance to suppliers abroad and in Japan, this period saw a more active exchange of information with its suppliers. With regards to our focal dyad, Nissan invited supplier employees to its technical center for bilateral discussions on specific products alongside general training and brainstorming sessions. At this time, our interviewees at the supplier were no longer outright dismissive of Nissan's offers to dispatch engineers to their sites.
Importantly, the supplier's focus shifted to the needs of the end customer-a change made clear in the supplier's official "corporate philosophy" as featured in its brochures and at its website.
Consequently, Nissan and the supplier geared their improvement and innovation initiatives toward a common goal, customer satisfaction, they did not share during the NRP period.
It should be noted, however, that during this time Nissan did not relieve its cost pressure on the supplier. Final decisions on supplier selection were made, as in the de-embedding period, based on tough cost performance and price criteria. That being said, the RNPO panel membership embodied a medium to longer-term oriented collaborative ethos rather than the per-contract, transactional approach of the previous period, to reflect the emerging belief that joint improvement initiatives could benefit the relationship's performance.

Synopsis: The three periods
The embedding period was marked by ailing performance for both Nissan and the supplier.
During that time, improvement as well as innovation initiatives by both partners had stalled. Nissan narrowly focused (and enforced its supplier's focus) on manufacturing excellence to the detriment of innovation initiatives, resulting in poor cost performance and uncompetitive product design.
The supplier was not able to effectively forge substantial changes since it did not contribute any new technologies. More radical innovations, deemed disturbing to the relationship's status quo, were shunned. Thus, neither partner strove to liberate the relationship's deadlocked situation and generate more significant innovations. With the relationship having settled in a mutually acceptable-yet globally inferior-configuration, collaborative initiatives were no longer sufficient to advance the relationship's performance, resulting in a worrisome decline.
We find that when embeddedness was temporarily relaxed by the NRP, both partners were spurred to independently search for improvement and innovation. Both succeeded in improving individual results, with Nissan enlisting a broader set of suppliers that offered new technologies.
The supplier, in turn, was able to focus on advancing its own R&D capabilities, to streamline its processes free of Nissan's stifling protection, and to seek new customers. This independent search performed by both partners unleashed creativity and helped the relationship dissolve its deadlock.
As Nissan's post-NRP performance improvements started to plateau, it sought to re-embed its suppliers. Nissan could no longer reap the benefits of coordinated actions and was seeking to resume mutually aligned innovation initiatives. According to a senior Nissan purchasing official: We've bottomed out now in terms of the competitive, destructible, short-term approach. … If Nissan thinks it can survive, thrive, and prosper at the expense of suppliers, then clearly that's not the case.
Considering all three periods together, the temporary relaxation reinvigorated supply chain performance because it sparked improvement and innovation activity.
Clearly, Nissan did not revert to the identical pre-NRP constellation during re-embedding, which is not surprising given technological change, demand shifts, turnover of key personnel etc.
that any firm would most likely face during a 12-year period. Nonetheless, the keiretsu past was deliberately invoked in Nissan's efforts to regain closer coordination with suppliers. According to one of our Nissan interviewees: [Nissan is] starting to think about going back to some of the pre-1999 long-term-ism rather than what was subsequently. ... [There is] a realization now that maybe we have exhausted the other approach. And as these things tend to be quite cyclical we start talking more about trust, more about partnership, more about long-term relationships. … It won't look like what we knew from pre '99, but … there is a realization now that we've caught up from a benchmark position and we now need to decide how we can move forward and how we can be far more constructive.
Patent data, despite its flaws (Cohen, 2010), may serve as a proxy to depict the observed changed pattern of improvement and innovation initiatives. Figure   The data were retrieved from Nissan's Japanese financial statements and Thompson One database.
Based on this synopsis of the three periods, we formulate the following proposition: Proposition 1: A temporary relaxation of buyer-supplier embeddedness can aid overall supply chain performance via reinvigorated improvement and innovation initiatives.

A Computational Simulation Model of De-Embedding
Our model is grounded in the case and parsimoniously represents its core elements: (i) a supply chain's search on a complex technological environment by using canonical performance landscapes, (ii) the supply chain is constituted by a relationship between a buyer firm and a supplier firm in either an embedded or a de-embedded state, and; (iii) temporary de-embedding is a dynamic process that changes an embedded supply chain relationship into a de-embedded and then back to a re-embedded one. In the following we describe how these three core elements are operationalized in our model.

Supply chain search on a performance landscape
We build a variant of Kauffman's (1993) NK model to represent a supply chain's search for performance in a complex technological environment. The NK model is a widely used method to model complex organizational problem solving (e.g., Chandrasekaran et al., 2015, Giannoccaro, 2011, Giannoccaro and Nair, 2016, Siggelkow and Rivkin, 2005. In our application of this model, N represents the number of decisions made in the supply chain. For instance, one typical decision could consider the material flows at the first-tier supplier to be organized following a push or pull logic. In another decision, the buyer undertakes quality controls. Following the standard NK model, each of these decisions di is assumed to be binary, that  (d , ,d , ,d ).
Supply chain performance is a function of both parties' decisions, formally P(dS, dB), and will be specified next. In each period t, the firms carry out search sequentially. Each search randomly chooses one decision di, evaluates if changing specification (from 0 to 1, or vice versa) improves performance, and updates the configuration in the case of gains. Next, the other firm carries out its search, this cycle being repeated during T periods where T represents the time-to-market pressure of the search.

Embedded and De-embedded Relationships
Both our literature review and case observation characterize an embedded relationship as attending to the other party's objectives: each firm considers not only how its decisions will affect its own performance, but also how its acts will influence the results of other firm's decisions. For example, in the embedded period, Nissan and the supplier engaged in collaborative kaizen programs that were geared to offer joint gains to both firms.
In contrast, a de-embedded firm gauges solely how its choices will affect its own results.
Here, when the supplier and the buyer search the landscape, they will adhere to partial objective functions. As our focal relationship de-embedded, Nissan and the supplier strove for improving their own performance by nurturing new capabilities (e.g., modularizing platforms and prototyping, respectively); those initiatives did not aim at improving the supply chain performance between

The de-embedding process
De-embedding the supply chain refers to shifting a relationship from an embedded (default) state into a de-embedded state. When this shift happens, supply chain members no longer search for mutual benefits, but only for their own. For a given period t, the supply chain becomes deembedded with probability δ-referred to as inclination to de-embed. To broaden the perspective from our single case and to unearth the performance tradeoffs tied to varying inclinations to deembed, this strategic variable will be systematically varied such that [0,1]   . As a default in the baseline model, the de-embedded supply chain is re-embedded (i.e., resumes the embedded state) after one period. However, later we also investigate the duration of de-embedding.
Our notion of de-embedding as a tunable probability has three important advantages. First, it provides a parsimonious representation of relationship de-embedding free of designing specific policies that is more in line with ideal agent-based modeling. Second, this representation accords well with the search notion from complexity theory, which argues that organizational decisionmaking is not perfectly rational, transpiring via simple heuristics such as probabilistic decision rules (Cyert and March, 1963). Third, since our goal is to characterize supply chain performance effects from a conservative angle, we are chiefly interested in the theoretical advantages and disadvantages of de-embedding events that may occur stochastically. We thus sidestep having to tailor (perhaps improperly) an 'optimal' rule for de-embedding the supply chain. In the robustness section, we investigate a model extension that is built around the special case of de-embedding contingent on a standstill in performance.

Simulation Results
We implemented the model described above in Matlab and ran 10,000 replications of each experimental instance. When not specified otherwise, we parameterized the NK technology landscape with N=12 and K=6. While this parameterization clearly does not capture the magnitude and complexity of real-world supply chain decision-making, it allows us to represent sufficiently complex behavior that can still be managed computationally (cf. Chandrasekaran et al., 2015).
Allowing for a replication of our results (also through replicative coding in other programming languages), we included a pseudo-code of our model in the Appendix. All presented performance figures are statistically robust in that they exhibit sufficiently small standard errors (<0.001)-too small to be meaningfully displayed in the ensuing performance graphs.  Figure 4 shows supply chain performance (P, depicted on the y-axis) as a function of the (varied) inclination to de-embed the relationship (δ, depicted on the x-axis). The two curves characterize an environment of moderate (T=100) versus high (T=50) time-to-market pressure.

The performance implications of de-embedding for supply chain search
Starting from an ever-embedded supply chain (δ=0), we observe that a small increase in δ provides steep performance gains for the supply chain for both time-to-market-pressures. A maximum performance is reached at δ =20% for both time-to-market pressures. High time-to-market pressure, however, takes a toll on performance since less time is available for search. Post-peak performance drops with inclination to de-embed the supply chain. While the performance effects are more pronounced for moderate time-to-market pressure, both performance curves trace an inverse Ushaped pattern where a modest, positive inclination to de-embed appears most effective.
Why does a temporarily de-embedded supply chain search improve overall performance?
Search theory has shown that organizational adaptation on complex landscapes may eventually freeze (Rivkin and Siggelkow, 2003) as organizations reach a "sticking point". In the single organization setting, organizational search halts on a local hill where any small move would deteriorate performance. We also know that freezing on inferior configurations proves more likely and becomes especially damaging when organizations face more complex search environments on more rugged landscapes (Levinthal, 1997).
In our inter-organizational setting that embraces a dyadic relationship of two firms, search activities across the entire supply chain may likewise freeze, particularly when neither buyer nor supplier can identify specific actions that would advance the relationship en bloc. Such a scenario transpired with Nissan and its first-tier suppliers during the over-embeddedness period. Here, the supply chain relationship seemed to be stuck in deadlock as evidenced by declining innovation advances in Figure 2. At the same time, it is also possible that de-embedding the supply chain relationship reshapes how the supply chain deals with sticking points. To probe this further, we ran another experiment in which we varied the complexity of the supply chain's environment, exploring both lower complexity (K=2) and higher complexity (K=10). Figure 5 details results of the simulation experiment in which we varied the complexity of the supply chain under moderate time-to-market pressure (T=100). To compare performance across complexity levels, the y-axis plots a relative supply-chain performance gain or loss that a de-embedding supply chain (with varying δ) has above or below an ever-embedded supply chain (δ =0) facing identical degrees of complexity (K∊{2,6,10}). This relative perspective allows us to focus on the performance gains delivered by de-embedding normalized for each complexity level.
Notably, previous research has established that complexity also generally affects performance levels in NK fitness landscapes via (i) different likelihoods for local peaks and valleys, and (ii) the various expected performance heights and depths for such local positions (Levinthal, 1997).
Therefore, we isolate the effects of de-embedding from the other (known) effects of complexity. Two results from Figure 5 are particularly striking. First, the performance gain from deembedding is higher in environments of higher complexity. This can be seen where the K=10 performance curve dominates the K=6 performance curve sitting atop the K=2 performance curve.
Second, the most effective inclination to de-embed increases (shifting rightward in Figure 5) with complexity: For K=2, it is at about 6%; for K=6, gains peak near 20%; for K=10, benefits plateau at around 40%. Moreover, for lower levels of complexity, performance gains plunge more quickly with higher de-embedding. For K=2, increasing the de-embedding inclination beyond 50% would even reduce supply chain performance below that of an ever-embedded supply chain (as signified by performance losses with the high inclination to separate).
The first result highlights that de-embedding is more effective when complexity is high, that is, when it is more likely for the supply chain to find itself in deadlock on a local performance hill. In such a scene, de-embedding temporarily relaxes the strong ties between the relationship partners. Then, both the supplier and the buyer no longer pursue chain-wide performance, but they independently search for self-improvement. By doing so, they make decisions that may harm the supply chain as a whole in the short term while prodding the relationship to escape a local optimum to advance in the long term. Thus, de-embedding can broaden the entire supply-chain search.
Because the supply chain re-embeds after the de-embedding, the aligned supply chain partners will again make decisions to smooth out prior misaligned actions. However, the creative impulse obtained from de-embedding moves the supply chain forward from past local sticking points. This mechanisms of broadened search well befits the more radical improvement initiatives that Nissan and its suppliers undertook when their relation was de-embedded.
The second result highlights that while de-embedding may improve supply chain performance, it comes at a cost when used excessively. This cost results from the loss of alignment between the buyers' and suppliers' decisions. As they pursue independent improvements fitting their own objectives, incompatible paths may result. De-embedding costs are embodied in the waning (and even negative) post-peak trajectories in Figure 5. Here, too-frequent de-embedding spawns mismatches that hamper supply chain performance. Under low complexity, the advantage of broader search via de-embedding is limited, and the cost of de-embedding quickly counteracts any benefits.

Duration and intensity of de-embedding
So far, we have assumed that de-embedding periods last for one period (D=1) where the de-embedding features a complete cut of ties between the supplier and buyer. In order to shed light on the effects of both duration and intensity of de-embedding, we now relax these two assumptions.
First, we let de-embedding take a tunable duration of 1 D  periods, i.e., after the firms stay deembedded more than one period before they resume into an embedded relationship (i.e., reembedding). Second, within de-embedded periods, the supply chain members will not totally cut Here, the intensity of a de-embedding can be characterized by 1-J/N. When J=0, the intensity is 100% since there is no overlap in the objectives, whereas if J=N, intensity is 0% and the supply chain will stay embedded as the objectives of buyer and supplier match. Figure 6 portrays the results of this relaxation upon supply chain performance for T=100 and K=6 under tuned durationintensity combinations. De-embedding is most effective when implemented in the form of short and intense periods, as seen for the dominating curve with D=1 and 100% intensity. With a brief, intense de-embedding strategy, the supply chain strikes an effective trade-off: The supply chain enjoys creativity unleashed by intense periods of de-embedded search while not losing much time that is required for re-coordinating these initiatives through successive re-embedding efforts.
To summarize, our simulation reinforces the main finding of the case study; thereby it lends further validity to the simulation model, and generalizes our case insights by unearthing how temporary de-embedding affects search as a mechanism for supply chain performance (Proposition 1). Furthermore, the simulation-based results augment theory, giving rise to propositions on (i) how frequently de-embedding should occur to improve performance (Proposition 2), (ii) how technological complexity interacts with the performance gains of de-embedding (Proposition 3), and (iii) how long and intense the de-embedding periods should be (Proposition 4).

Proposition 2:
There is an inverted U-shape relationship between the inclination to de-embed and supply chain performance where a small inclination is most effective for supply chain performance.
Proposition 3: Temporary de-embedding is more beneficial to supply chain performance in more complex technology landscapes.
Proposition 4: Short and intense periods of de-embedding offer higher performance gains than prolonged or less intense periods of de-embedding.

Model Extensions and Robustness Analyses
In order to address specific "what-if" questions, and to broaden the perspective obtained from the case, we explored a number of model extensions.  Figure 7. First, de-embedding continues to create value when triggered contingent on standstill (Proposition 1); if implementable, such a practice would even outperform our base model. This is because the supply chain is de-embedded just when it can help reinvigorate search. Second, the characteristic performance curve of de-embedding remains (Proposition 2), yet its downside part appears less pronounced. This result is because deembedding takes effect only following a performance standstill and so cannot harm efforts to ensure compatibility after de-embedding as much.
Electronic copy available at: https://ssrn.com/abstract=3058630 The second extension examines the counter-factual case of a supplier unwilling and/or incapable to re-embed in our case study. We observed in fact that the supplier was initially skeptical to re-embed, but eventually joined Nissan's re-embedding efforts. Therefore we devised a model where the supplier stays de-embedded even when the buyer re-embeds, that is,  Figure 7 illustrates that a supplier unwilling to re-embed imposes a deadweight loss on supply chain performance when compared to our base model. Two factors drive this disadvantage: (i) the supplier's re-embedding is required to guarantee compatibility of bilateral search efforts, and (ii) future de-embedding will be less impactful since the supplier stays de-embedded. A way out of this dilemma, one could conjecture, is that the buyer compensates the buyer to re-embed, but more field and model-based research is needed to examine that approach. and dS2, respectively, while the buyer still owns N/2 decisions dB). Building on that first variation, we also devised a second variation where one of the two suppliers may be replaced when deembedding occurs. That means that one of the supplier's configuration dS1 or dS2 (with equal chances) may be entirely changed after de-embedding. Both variations with two suppliers offer the same inverse U-shaped patterns as regards to supply chain performance (see Figure 7).
To summarize, we examined three relevant what-if scenarios that generalize the contingencies of our case study. While these extensions give rise to intricate performance nuances of de-embedding under new scenarios, they all reinforce our key finding regarding the benefits of temporary de-embedding, and so lend further robustness to our theory building.

General Discussion
This research investigates the phenomenon of temporary de-embedding as a dynamic strategy to manage supply chain relationships, and analyzes how temporary de-embedding may influence supply chain performance. A longitudinal case study of Nissan's relationship with a strategic first-tier supplier showed that temporary de-embedding reinvigorated both parties' improvement and innovation initiatives to the betterment of the relationship's performance.
Evidence from this buyer-supplier case indicates that prior close embedding led to a deadlocked performance that neither partner's innovative efforts could dispel. De-embedding stirred the relationship, enabling both Nissan and the supplier to shape their own improvement initiatives, which enabled the supply chain to break free of its deadlock. Later efforts to re-embed the supply chain enhanced coordination that restored the compatibility of distributed initiatives. To validate our case findings and augment our theory building, we forged a parsimonious simulation model using complexity theory and its fundamental notion of search (Levinthal, 1997). The simulation experiments generalize our main finding that benefits arise from temporarily relaxing embeddedness between buyers and suppliers (Proposition 1). The simulations augment this finding in a threefold way. First, de-embedding harms supply chain performance when pursued too often (Proposition 2). Second, de-embedding benefits supply chain performance especially in environments subject to high technological complexity (Proposition 3). Third, short but intense de-embedding is most effective for enhancing supply chain performance (Proposition 4).

Theoretical Implications
Taken together, the case and simulation results feature three contributions to supply chain management research. First, we offer new theoretical insights on how to manage buyer-supplier relationships-a research topic hotly debated. For example, Cachon and Lariviere (2005) highlight the supply chain-wide benefits of aligning buyer-supplier aims through revenue-sharing contracts.
In a similar vein, Dyer and colleagues argue that embedded relationships comprised the main factor driving the competitive advantage of Japanese manufacturers over their American counterparts (Dyer, 1997, Dyer and Singh, 1998, Dyer and Chu, 2000, Dyer, 1996. Likewise, research on supply chain innovation advocates close, prolonged involvement of suppliers in new product development processes (e.g., Clark, 1989, Yan and Dooley, 2013. In contrast, our case and simulation findings suggest that an overly embedded relationship can impede supply chain innovation and performance. This finding is in line with later research that has documented excessive embeddedness as a liability (e.g., Gargiulo and Benassi, 2000, Swink and Zsidisin, 2006, Villena et al., 2011. This latter stream of research concludes that firms should avoid overly embedded ties and strive for a balanced degree of embeddedness. Yet, it has remained unclear how firms can achieve such a balanced level of embeddedness. We contribute to this discussion by building knowledge on how firms, by temporary de-embedding their relationships, can dynamically achieve an effective balance. Our study thus challenges the implicit rigidity assumption that has plagued SCM and so opens up a new research avenue that conceptualizes buyer-supplier embeddedness as a dynamic phenomenon. Second, we link our notion of dynamic embedding to the concept of search from complexity theory. We do so by demonstrating how temporary de-embedding influences trajectories and outcomes of supply chain-wide search in our case and simulation. Prior research on the dark side of embedded relationships proposes excessive opportunistic behavior as an explanation for the detrimental effects of over-embeddedness (Gargiulo andBenassi, 2000, Villena et al., 2011). In contrast, our search perspective suggests that even when supply chain members attend to each other's needs perfectly free of opportunism, embeddedness beyond a certain threshold locks both organizations into improvement initiatives that offer mutually acceptable, yet only incremental, performance improvements. Our search perspective also qualifies prior research that seeks to optimize supply chain relationships through formal models (e.g., Cachon and Lariviere, 2005). Due to the complexity of supply chain decision-making, it is oftentimes difficult-if not impossible-to formulate 'optimal' masterplans for buyer-supplier relationships.
Instead, the search perspective acknowledges this complexity and conceptualizes supply chains as a setting where imperfectly rational decision-makers jointly and iteratively search for improvements and innovations (e.g., Chandrasekaran et al., 2015, Giannoccaro, 2011, Kim et al., 2015, Sting and Loch, 2016.  Temporary relaxation of buyer-supplier embeddedness can aid overall supply chain performance via reinvigorated improvement and innovation initiatives.
Managers do not have to view supply chain embeddedness as a one-time choice but as a dynamic and malleable process over which they have agency.
Temporary de-embedding should be adopted as a strategic tool of buyer-supplier relationship management.
There is an inverted U-shape relationship between the inclination to de-embed and supply chain performance where a small inclination is most effective for supply chain performance.
Temporary de-embedding can offer supply chain wide innovation gains, but too much deembedding causes supply chain incompatibilities.
Temporary de-embedding is more beneficial to supply chain performance when technology landscapes are more complex.
Despite the received wisdom that "more complexity calls for more coordination," close coordination under complexity might actually "freeze" the relationship in a deadlock. Deembedding (and so temporarily reduced coordination) can break this.
Short and intense periods of de-embedding offer higher performance gains than prolonged or less intense periods of de-embedding.
De-embedding is most effective when executed in a short and intense way.

Managerial Implications
Our study offers important insights for managerial practice. These insights follow directly from the propositions put forth in this paper, as shown in Table 2. First, supply chain managers do not have to view embeddedness as a one-time choice. Rather, they can revise the embeddedness of buyer-supplier relationships dynamically. Our study notably shows how temporary deembedding can reinvigorate capability improvements which, in turn, can benefit supply chain performance. As illustrated in our case study, firms can de-embed by selling their equity holdings in the supplier; dedicating purchasers to components rather than to suppliers; diversifying business across several alternative suppliers; curbing close friendships with suppliers, and; clarifying standards that govern the relationship, among many other measures. Firms need to reverse such structural and relational efforts when re-embedding. However, re-embedding in the cognitive dimension can be particularly difficult. This dimension encompasses shared meaning, defined rules and set roles in the relationship. When these are altered drastically as part of intense deembedding, resumed realignment may not be straightforward.
Second, our investigation of the duration, intensity and frequency of temporary deembedding can also inform supply chain decision-making. Intuition calls for more coordination between supply chain members in more complex environments. This is due to increased technological interdependence among decisions of buyers and suppliers. Our results qualify this intuition by adding a dynamic perspective. Particularly in complex environments, the supply chain's search for innovation is more likely to get stuck because there are more technological interdependencies to consider, which further paralyze search. De-embedding the relationship can thus offer additional performance benefits due to its ability to unleash innovation by all supply chain members. By temporarily de-embedding interdependent efforts at the buyer and supplier, supply chain management can strike an effective balance between the coordination advantages of tight collaboration (in an embedded relationship) and the innovation advantages of distributed initiatives (in a temporary de-embedded relationship). Yet, engaging in de-embedding too frequently or staying de-embedded for too long leads to incompatible search outcomes, which in turn can hurt supply chain performance.

Limitations
The main limitation of our study is that our inductive case study relied on one specific buyer-supplier relationship. In order to triangulate how embeddedness evolved over time, and to strengthen internal validity of our case findings, we incorporated secondary data (e.g., equity ownership in the supplier and patent data). These secondary data show patterns consistent to what we find in our case. Furthermore, we confirmed with our Nissan interviewees that our observations were not specific to the particular component we studied. Most importantly, our research design combines an inductive case study with agent-based simulations in order to generalize and augment our understanding of the de-embedding phenomenon and its mechanisms.
Moreover, at first glance it seemed that Nissan reacted to poor supply chain performance by de-embedding the relationship with the supplier. However, our analysis suggests that a timely response was not the case. In fact, supply chain performance had been ailing for several years model, consequently, offers the generalizable and robust insight that short but intense temporary de-embedding improves supply chain performance even under imperfect monitoring and response.
In our model extension, we observe that performance-contingent de-embedding would boost performance even further. A repeated probing of relationships gives rise to a dynamic pattern of punctuated equilibria; where periods of standstill are trailed by periods of reinvigorated search.
That said, more field and modeling research is needed to fully understand whether, and if so, how, the timing of de-embedding can be optimized.
Finally, absent broader cross-industry and cross-country analyses that empirically test and further generalize our conclusions, our advice on de-embedding should be considered with caution.
Still, we are confident that our findings pave the way for scholars to develop further insights to support the dynamic and malleable view of buyer-supplier relationships that we propose here. t=t+1 end for end if Supply chain search oscillates between embedded (default) and de-embedded states