The data sharing conundrum: revisiting established theory in the age of digital transformation Open Access

When considering the impact of digital transformation (DT), supply chain management (SCM) researchers and industry experts consider end-to-end real-time visibility and autonomous process integration as likely trajectories (Hendriksen, 2023; McKinsey & Company, 2021). The last decade has been characterized by rapid developments in technologies that facilitate interorganizational communication, advanced analytics and automation (Culot et al., 2020a; Frank et al., 2019). Electronic component miniaturization, wireless technologies and direct networking are now allowing the interconnection of processes, machines and products. These trends have spurred ideas of “hyper-connected,” “smart” and “self-thinking” supply chains (SCs), which integrate data and decisions across multiple tiers and extend to technology and service providers (e.g. Calatayud et al., 2018; Dolgui and Ivanov, 2022; Hartley and Sawaya, 2019; Pessot et al., 2022).

Amid the excitement for emerging opportunities, some studies have contested the immediate materialization of major disruptions (Klöckner et al., 2022; Sodhi et al., 2022). Expectations may have been inflated by technological breakthroughs that overlook the intricate interplay of technologies with the processes, communication channels and operational principles that characterize SC relationships. Alongside technological advancements, new interorganizational interactions and governance structures are needed (Kache and Seuring, 2017). However, early implementation cases rarely reported substantial changes in ongoing practices (e.g. Cannas et al., 2023; Franzè et al., 2024; Sauer et al., 2022). Moreover, technology adoption has been slower than anticipated, as companies face implementation challenges against existing SC structures and routines (e.g. Cecere, 2022; Hasija and Esper, 2022; McKinsey & Company, 2021). This raises the key question of whether DT is fundamentally altering the functioning of SCs as previously theorized.

Interorganizational data sharing is a central aspect that is crucial for advanced analytics and autonomous decision-making in SCs (e.g. Calatayud et al., 2018; Hendriksen, 2023; Zhou et al., 2022). Although technological opportunities exist (Müller et al., 2020; 2018; Münch et al., 2022), concerns persist about confidentiality, opportunistic behavior and value appropriation dynamics (de Prieëlle et al., 2020; Legenvre and Hameri, 2023; Mathivathanan et al., 2021). These echo past concerns about technologies such as electronic data interchange (EDI) and Web-based solutions (e.g. Fawcett et al., 2007; 2011; Harland et al., 2007; Kembro and Näslund, 2014). To fully understand the impact of DT on SCM, it is thus important to investigate whether previously theorized dynamics still hold true.

The aim of this study is to address the following main research question (RQ): How are the determinants of interorganizational data sharing changing in the age of DT? Although the literature has begun exposing the issue as problematic, the factors that drive or prevent firms from sharing data have not been properly clarified. This knowledge gap hinders both researchers and managers from framing challenges and opportunities accurately. Against prospects that are essentially based on the potential of new technologies, it is important to return to the core of SCM. By assessing the fit between established theoretical models and a changing empirical reality, extant knowledge can be leveraged against fashion waves and managerial hypes (Culot et al., 2020a; Hanelt et al., 2021).

As we investigate a long-lived topic, our approach is based on abductive reasoning (Dubois and Gibbert, 2010; Ketokivi, 2006; Ketokivi and Choi, 2014). Unlike the deductive theory-testing approach, this choice prevents us from being bound by pre-formulated hypotheses and allows us to capture novel elements. Further, abductive reasoning is more appropriate than a purely inductive study, as we could leverage prior knowledge to make sense of emerging dynamics without presenting “old wine in new bottles.” A review of research prior to DT led to a preliminary framework based on three theoretical lenses: the resource-based view (RBV) (Barney, 1991; Wernerfelt, 1984), resource dependence theory (RDT) (Pfeffer and Salancik, 1978) and the complex adaptive systems (CAS) perspective (Choi et al., 2001). A case study analysis of data sharing practices within an extended supply network allowed us to confront the factors existing before DT with emerging evidence and formulate plausible explanations.

We contribute to the literature by showing that although DT affects the mode and extent of data sharing, extant knowledge can still explain emerging dynamics. Alongside established factors motivating firms’ decisions, we also find new elements that refer, for example, to pooling data from different sources for joint analyses, digital platforms and legal setups. We discuss how these elements relate to RBV, RDT and CAS. These findings deepen our understanding of SCs’ future beyond speculations on the potential of new digital technologies, aiding managerial decision-making by more clearly highlighting roadblocks and enablers.

Interorganizational relationships always entail some sharing of data, information and knowledge (here, data describe the properties of objects and events, information results from data analysis provide descriptions and explanations, and knowledge refers to the use of information for specific purposes; Ackoff, 1989). The topic has been amply investigated across different research traditions with some kind of ambiguity (e.g. inventory and customer data have been examined in “information sharing” and “knowledge sharing” studies; Colicchia et al., 2019; Fawcett et al., 2007; Gast et al., 2019). For the sake of simplicity, in this overview, we refer only to data as the basic building block of any kind of interorganizational exchange.

In SCM, although buyers and suppliers share data for various purposes – from operational coordination to long-term strategy formulation (Wiengarten et al., 2010) – most of the interest has focused on material flow alignment, following the work of Forrester (1958) (e.g. Fawcett and Magnan, 2002; Kembro and Näslund, 2014). We thus complemented concepts germane to SCM with others originally elaborated on in strategic (e.g. Gulati et al., 2000; Grant, 1996; Grant and Baden‐Fuller, 2004) and innovation (e.g. Gast et al., 2019; Lawson et al., 2009) management. It would be beyond the scope of this study to go into these different research streams in detail. The scientific landscape is diverse in terms of the foci and disciplines involved; however, there is a common interest in the factors that motivate firms (not) to share data. To frame the issue in more general terms, we cluster the factors into three overarching dimensions. Three theoretical lenses are used to derive a logical explanation (Kembro et al., 2014).

The first dimension concerns the characteristics of the data. As firms regard data as a resource, the RBV provides a useful anchor. The RBV suggests that firms develop a sustainable competitive advantage if they possess resources that are valuable, rare, inimitable and/or non-substitutable, which should be protected (Barney, 1991; Peteraf, 1993; Wernerfelt, 1984). Since firms do not operate in isolation, resources may originate from interorganizational interactions (Dyer and Singh, 1998; Lavie, 2006). In general, RBV predicts that firms retain data that generate private benefits, whereas they share those that enable superior interorganizational processes (Bergh et al., 2019; Fawcett et al., 2011; Hernández-Espallardo et al., 2010). Based on the literature, two main characteristics motivate sharing decisions:

1. core-relatedness, namely, whether the data reveal a firm’s competitive advantage in its core business (Barthélemy and Quélin, 2006). Sharing is less likely with respect to customer details, production and innovation processes, sales and operational costs (Ritala et al., 2015; Soekijad and Andriessen, 2003); and

2. fungibility, which describes possible different use occasions (Choudhury and Sampler, 1997; Sampler, 1998). Business partners’ self-interest can lead to opportunistic behaviors on shared data (e.g. buyers might lose volume discounts after disclosing their demand patterns – Klein and Rai, 2009; business partners might improperly take advantage of shared data for private innovation – Estrada et al., 2016; Yigitbasioglu, 2010).

The second dimension describes the characteristics of the dyadic relationship between business partners. Relevant factors are examined through the RDT (Aldrich and Pfeffer, 1976; Mindlin and Aldrich, 1975; Pfeffer and Salancik, 1978). The basic idea is that patterns of dependence originate as organizations engage in exchanges to obtain resources (Davis and Cobb, 2010; Hillman et al., 2009; Wry et al., 2013). The level of dependence depends on the criticality of the resource and the availability of alternatives. Based on the RDT, partners’ motivations can be grouped into five factors:

1. the need for complementary resources, as data might not represent a resource per se but only in conjunction with other assets and capabilities that reside within firms and other organizations (Im and Rai, 2008; Lawson et al., 2009);

2. level of dependence, which is defined as the sum of the dependence between actors. The high stake of maintaining a smooth relationship prompts a stronger relational orientation and joint action (Lu and Shang, 2017; Gulati and Sytch, 2007). Thus, firms are more interested in broadening the scope of data sharing (e.g. Larson, 1992; Zaheer and Trkman, 2017);

3. dependence asymmetry, that is, the difference in actors’ dependencies on each other (Gulati and Sytch, 2007; Wry et al., 2013). An example is the case of a supplier that generates most of the revenue with a specific customer. More powerful business partners can exert influence (e.g. Bouncken and Kraus, 2013; Goodhue et al., 1992; Yang et al., 2017);

4. protection mechanisms, which are part of the tactics that firms use to manage relationships (Pfeffer and Salancik, 1978). These encompass clauses that regulate the ownership, allocation and use of data (e.g. Gast et al., 2019). Formal mechanisms include contracts, intellectual property rights and information security norms; informal mechanisms are routines and processes (Estrada et al., 2016; Ritala and Hurmelinna‐Laukkanen, 2013; Salvetat et al., 2013); and

5. benefit distribution mechanisms, namely, the upfront definition of the gains from sharing. This factor is relevant for both operational efficiency (Ganesh et al., 2013) and innovation (Ahuja et al., 2013; Sun and Zhai, 2018).

The last dimension concerns network-level dynamics. Interorganizational practices are embedded in the broader network of interfirm relationships and the external environment (Choi and Kim, 2008; Kembro et al., 2017). Factors related to this dimension can be analyzed through the CAS approach (Choi et al., 2001; Holland, 1995). CAS builds on general system theory, defining systems as groups of autonomous agents that are characterized by a common purpose, function as a whole and adapt to external changes (Katz and Kahn, 1978; Mele et al., 2010). CAS allows a simultaneous focus on the network as-is, external influences and coevolutionary trajectories. From the literature, we identified the following factors in this dimension:

• network governance structure, which refers to how the firms involved in the network coordinate their actions (Gulati et al., 2012). For example, smaller suppliers that provide complex products to large buyers are typically supported and controlled; thus, exchanges are more intense (Gereffi et al., 2005). The opposite is true for turn-key suppliers (ibid.). Overall, multi-tier flows depend on the responsibilities assigned to each party (Mena et al., 2013);

• environmental complexity and dynamism describe the business, geopolitical, legal and technological landscape. The more complex and dynamic the environment is, the more firms seek data in support of their decisions (Srinivasan and Swink, 2018); and

• assimilation of technological/communication standards, that is, the spread of technical specifications allowing exchanges between information systems (Sodero et al., 2013). The lack of standards has been highlighted as one of the key barriers to systems such as EDI (Frohlich, 2002; Hart and Saunders, 1997; Narasimhan and Kim, 2001). Their origins and diffusion are related to the actions of a few dominant firms that generate coevolutionary patterns (Zhao et al., 2007; Zhu et al., 2006).

Figure 1 summarizes the dimensions and factors outlined above, including key references.

DT is defined as “[…] organizational change that is triggered and shaped by the widespread diffusion of digital technologies” (Hanelt et al., 2021, p. 1160). These include “[…] the combination and connectivity of innumerable, dispersed information, communication and computing technologies” (ibid.). Similar concepts include Industry 4.0 and the idea of a new industrial revolution (Frank et al., 2019). Timewise, there is no clear-cut date marking its beginning. The technologies underpinning the phenomenon have been around for a long time, even though they are not mature for industrial applications (OECD, 2017). Conventionally, the early 2010s are regarded as the start years due to an acceleration in development, falling costs and industrial policies fostering technology adoption (Culot et al., 2020a).

The list of possible front-end applications now available for firms is extensive. On their basis, few enabling technologies are combined for specific purposes (Battaglia et al., 2023; Frank et al., 2019). In SCM, the most relevant ones are those related to connectivity, data-driven decision-making and automation (Calatayud et al., 2018; Dolgui and Ivanov, 2022). These include cloud computing (i.e. hardware and software computing services delivered on-demand through the network; Marston et al., 2011), cyber-physical systems and the Internet of Things (i.e. sensors and connectivity applied to products and machines; Lee and Lee, 2015), blockchain technology (i.e. distributed, immutable and encrypted transactional databases; Babich and Hilary, 2020) and advanced data processing techniques based on artificial intelligence, machine learning and big data analytics (Hendriksen, 2023; Waller and Fawcett, 2013).

These technologies mark a step change in how data can flow and be used across organizational boundaries (Culot et al., 2020a; Paolucci et al., 2021). On the one hand, sensor-generated data can be accessed beyond the ownership of the physical asset (Kiel et al., 2017; Münch et al., 2022). For example, machines can connect with industrial technology providers for smart maintenance and optimization services (Cannas et al., 2023; Cepa, 2021; Dalenogare et al., 2022). Similarly, data from connected products can be analyzed by manufacturers to improve their design (Franzè et al., 2024) and activate services (Chen et al., 2021; Holmström et al., 2019). On the other hand, information systems can be more easily integrated so that data are shared vertically across the different levels and functions of the same firm, as well as horizontally between firms (Müller at al., 2020; 2018). This potentially allows real-time multi-tier visibility, enabling predictive analyses and autonomous configurations (e.g. Calatayud et al., 2018; Dolgui and Ivanov, 2022).

Although the literature describes some successful implementation cases (e.g. Cannas et al., 2023; Sauer et al., 2022), there is evidence of ongoing challenges connected to interorganizational data sharing (Legenvre and Hameri, 2023; McKinsey & Company, 2021; OECD, 2019; WEF, 2020). These prevent technology diffusion and lead to suboptimal utilization (Kouhizadeh et al., 2021; Sodhi et al., 2022). Some sparse arguments have emerged in the current research, resonating with the dynamics discussed prior to the advent of new digital technologies (e.g. Bechtsis et al., 2021; Kache and Seuring, 2017; Ye et al., 2022). Given the potentially disruptive impact of DT, these factors should be further explored. To this end, we detail the main RQ (i.e. How are the determinants of interorganizational data sharing changing in the age of DT) into three specific research questions (sRQs):

Case research was selected because it offered the opportunity to investigate data sharing within the context of DT in depth and with latitude (Eriksson and Kovalainen, 2008). As our study started from an already theorized landscape, the approach was based on abductive reasoning (Ketokivi and Choi, 2014). Case research is apt for theoretical elaboration in SCM, as it allows the gaining of a comprehensive understanding of complex phenomena in the real world, which is important in investigating how questions (Meredith, 1998; Yin, 2018). Moreover, faced with the relative novelty of DT and the hype around emerging technologies (Sodhi et al., 2022), a qualitative empirical approach was appropriate for developing broad and nuanced explanations (Voss et al., 2002).

Under this premise, we investigated data sharing within an extended supply network, defined as the set of multi-tier linkages related to the flow of goods, technology and service provision (Carter et al., 2015). Consistent with this empirical setting, we developed a single embedded case study analysis that involved 16 firm subcases (Yin, 2018). This choice is in line with the methodological recommendations of Dubois and Gadde (2002) and Halinen and Törnroos (2005), as interorganizational practices need a common frame to control for extraneous variables.

The unit of observation was an extended supply network in the automotive industry. Previous research has broadly investigated the sector (e.g. Iskandar et al., 2001; Lockström et al., 2010); thus, the practices prior to new technologies in the industry remain largely known. Moreover, the extended supply networks in the automotive industry are deeply affected by DT. On the one hand, product connectivity is becoming a standard in the industry (i.e. the “connected car”) (Bohnsack et al., 2021). On the other hand, the automotive industry has rapidly adopted new digital technologies in operations and SCM, ranking among the first industries for investment size (Kamble et al., 2020). DT is already considered the norm among original equipment manufacturers (OEMs), whereas suppliers are under pressure to upgrade their capabilities (Arcidiacono et al., 2022; Franzè et al., 2024).

In designing our study, we acknowledged the presence of large first-tier suppliers serving multiple OEMs, usually on a continental basis (Huang et al., 2020; Mohamad and Songthaveephol, 2020). We thus adopted a stratified approach to account for the tiered structure and selected firms headquartered in Europe (Patton, 2002). Europe houses some of the largest and most renowned players (Arcidiacono et al., 2022) while being quite advanced on DT due to policy action (Culot et al., 2020a).

We started with OEMs and their primary interfaces. This led to the identification of an initial sample of 11 subcases. The preliminary analysis highlighted the need to consider other actors (Halinen and Törnroos, 2005). Two industrial technology providers and three digital technology providers were included.

A case study protocol (see the  Appendix) including semi-structured interview questions was developed and refined through three pilot interviews. Consistent with the aims of the study, the interviews were meant to gather information about technology-enabled data sharing practices that refer to the last wave of digitalization (i.e. since the early 2010s, which are conventionally assumed as the onset of the phenomenon). The questions specified data sharing practices along the SC (i.e. with buyers and suppliers) as well as with technology providers. To better understand the factors driving or preventing such practices, we asked about the motivations (not) to pursue them as well as relevant enabling factors. Data collection started in autumn 2020 and continued into the first months of 2022. Given the broad scope of data sharing, we looked for executives with expertise in SCM or purchasing; production or logistics; DT or information systems (e.g. information and communication technologies [ICT] director, chief information officer [CIO]); research and development [R&D], innovation or connected vehicles; general management (e.g. chief executive officer [CEO]); project management [PM]; or customer-facing roles. The informants were selected based on their job titles and seniorities. They were identified within the contacts of the researchers and their institutions and by searching the Web and professional social networks (i.e. LinkedIn). Their expertise was verified through preliminary conversations. The informants were asked to share further contacts within their firms. A total of 32 informants were involved. Each interview lasted between 50 and 150 min, and the respondents disclosed specific dashboards on their information systems. All interviews were recorded, fully transcribed and stored in a database, which also included  Appendix. Although for some companies, we could not find respondents for all areas, the abundance and variety of material allowed triangulation. Table 1 presents an overview of the subcase companies and informants. The companies are referred to by generic monikers.

The materials were analyzed through open coding to identify key terms and concepts (Miles and Huberman, 1994). NVivo software was used to simplify the association of categories and text passages, their storage and retrieval. Consistent with the aims of the study, we focused on data sharing enabled by new digital technologies. Pre-existing technologies were considered if complemented by new ones (e.g. supplier portals enhanced by cloud computing).

With respect to the interview protocol, the data were analyzed as follows. The first section concerned the company background, existing interorganizational relationships and the role of the respondent. Together with information from secondary sources, this was used to clarify the contextual conditions surrounding the phenomenon of interest. The second section supported the identification of technology-enabled data sharing practices that were pursued or dismissed by the firm. The data were analyzed inductively to cluster emerging practices based on DT opportunities, area and purpose, flow and assimilation level. Publicly available material was also examined. The results are presented in subsection 4.1. The third and fourth sections of the protocol were meant to gain insights into the motivations (i.e. drivers and barriers) underlying the firm’s pursuit or dismissal of such practices. We also asked about the implementation challenges that arose during the process. The arguments were analyzed for consistency with the factors already expounded in prior literature. The results are presented in subsection 4.2. The interview protocol included additional questions on the firm’s performance and supply network configuration to better understand the breadth and depth of the data sharing practices.

The results of the coding process were examined by all involved researchers. Any issue was solved through discussion. The presence of multiple companies in each category enabled a pattern-matching logic. Following an abductive approach (Ketokivi, 2006), the factors behind data sharing were compared with those derived from the literature (Figure 1). The plausibility of theoretical explanations was carefully addressed through multiple rounds of discussion.

To ensure rigor in the development of the analysis, we adopted several approaches (Gibbert and Ruigrok, 2010). In terms of construct validity, we triangulated data sources, explicitly indicated the collection and analysis procedures and involved colleagues in reviewing our analysis. Internal validity (i.e. “logical validity”; Yin, 2018) was ensured by deriving the research framework from the literature as well as through pattern matching, both comparing responses of different firms and the outcomes of previous studies (Eisenhardt, 1989b). External validity was pursued by matching empirical observations with theory (Yin, 2018). Finally, reliability was met using a case study protocol, recording and transcribing all interviews, setting up a case study database and involving multiple coders (Silverman, 2005).

Several data sharing practices related to DT were undertaken within the automotive extended supply network. These are summarized in ²Table 2and illustrated in the following paragraphs to contextualize the findings of the sRQs.Three overarching categories of data sharing practices emerge: process connectivity, asset/equipment connectivity and product connectivity. In Table 2, each data sharing practice has been attributed an alphanumeric code (e.g. A1, A2). These are consistently used in the text (in bold) to allow the reader to follow our reasoning while confronting more easily exemplary quotes and further details.

In our sample, firms increasingly interconnected and granted external access to their information systems, facilitated by cloud computing and information system integration, minimizing the cybersecurity risks associated with data stored on private servers. Various data sharing practices were identified in production planning. In practice A1, firms shared real-time production planning, work-in-progress and inventory data, empowering suppliers to autonomously decide when and how to build stock. Similarly, A2 involved suppliers integrating their information systems with buyers. Digital platforms interconnected players along the SC in A3.

Logistics practices included A4, in which timely data access aligned suppliers and operators with inbound material requirements, and A5, which leveraged delivery status visibility to enable manufacturers to adjust schedules. Cost reduction in externalized activities was a driver for A6, as Supplier 1 integrated information systems with a global logistic service provider to optimize inbound routes. Blockchain technology in A7 expedited traceability, accelerating the identification of quality issues and recall procedures. In new product development, cloud computing provided external design centers with methods, tools and data while maintaining control over proprietary information (A8). Additionally, in A9, increased storage space and secure data transfer solutions facilitated OEMs in requesting extensive data from first-tier suppliers, such as simulation parameters and results.

Regarding assimilation levels, OEMs extensively shared production planning data (A1) and delivery requests (A4). In terms of data sharing from suppliers to buyers, ample sharing occurred in support of inbound logistics (A5) and component/system development (A9). For other practices, instances were limited, primarily consisting of pilot projects and preliminary conversations.

The data sharing practices revealed from the interview data revolved around sensor-generated data from manufacturing plants (e.g. connected machines, gates, working stations, test cells) and logistics operations (e.g. connected trucks, containers, parcels). These were shared either through direct connectivity (i.e. the Internet of Things) or by interconnecting information systems (i.e. through cloud computing, enhanced Web interfaces and secure data transfer). Several of these practices affected interorganizational logistics. Especially in just-in-time or just-in-sequence (JIT/JIS) approaches, suppliers synchronized deliveries based on real-time data from systems of sensors and gates on the assembly line (B1). Similarly, truck geo-localization allowed higher accuracy in estimating materials’ expected time of arrival so that the production sequence could be adjusted (B2). Sensors and intelligent labels were also used to automate data acquisition for inbound materials (B3) and to monitor shipments’ ambient conditions (B4).

Asset/equipment connectivity also influenced traceability. By accessing suppliers’ machine data, buyers could verify the number of pieces produced and process parameters (B5). With respect to quality, these practices allowed buyers to adjust to what occurred at suppliers (B6). Suppliers improved their processes by accessing connected test cells’ data (B7). All manufacturers collaborated with industrial technology providers, engaging in data sharing practices aimed at process optimization (B8) and maintenance (B9), both derived from analytics on machine data.

Except for B1, we found only pilot projects and a limited number of instances.

Various data sharing practices concerned connected vehicles, using the Internet of Things for data gathering and subsequently sharing data through information system integration, cloud computing, teleservice systems and data transfer solutions. For individual vehicle data, OEMs operated upon owner authorization (C1). Sharing these data was essential for service access/provision, as many vehicle functionalities needed remote data processing (e.g. localization). First-tier suppliers also implemented component/system connectivity, and dealers connected data ports with central control units (C2).

Thereafter, sharing individual-level vehicle data enabled external engagement in service provision (e.g. maintenance and insurance) (C3). Partially different dynamics emerged for in-vehicle digital services, including infotainment and advanced driver-assistance systems (ADAS), such as automated parking and accident prevention (C4). Whenever OEMs leveraged external collaborations (e.g. with first-tier suppliers and digital companies), data sharing occurred in bulk for the development, testing and training of algorithms. Interestingly, connected vehicle data were shared through third-party digital platforms, often backed by OEMs’ direct investments. Vehicle-specific data were shared for multi-brand fleet management services (C5). Moreover, these data aggregators were used to sell and buy anonymized data in bulk (C6). For quality and new product development purposes, data could be shared with suppliers to enable the analysis and evaluation of the performance of their components (C7).

In terms of assimilation levels, OEMs were at the center of data governance (C1, C3), while dealers and first-tier suppliers could not directly access the data (C2). Data-driven collaborations (C4) and digital platforms were common (C5, C6). In terms of quality and new product development, data were shared with suppliers only in the case of issues (C7).

As described in subsection 4.1, data sharing practices presented different levels of assimilation. The informants’ comments were compared with the initial list of factors (Figure 1). Tables 3, 4 and 5 contain illustrative quotes. The main findings are described in the following paragraphs, highlighting the specificities of DT.

Our evidence confirmed that data were shared based on their core-relatedness and fungibility, in line with the RBV (Table 3). In terms of core-relatedness, both OEMs and first-tier suppliers hesitated to share data about innovation and core manufacturing processes. Data related to typically outsourced activities were easily shared with greater timeliness and granularity (e.g. logistics and maintenance; A4, A5, A6, B9). Regarding connected vehicle data, OEMs controlled data flows to engage partners offering non-differentiating services (C3).

In the face of DT, companies redefined their core competences. This was determined by servitization opportunities stemming from data analysis of connected processes (e.g. planning and logistics optimization; A3, A6), assets/machines (e.g. maintenance and process optimization; B9; B8) and products (e.g. service access, development and provision; C1, C2, C3, C4, C5 and C6). Against these opportunities, however, firms’ decisions regarding the extent of data sharing were still determined by core-relatedness.

Considering fungibility, the comments confirmed reluctancy (e.g. with respect to data that could be used in price negotiations and exposing confidential information; A2, B8 and C7). There seemed to be a higher likelihood of sharing data exclusively related to the activity performed together with the business partner (e.g. specific deliveries; A5, B2, B3; machines/production steps; B5, B6) and from a limited timeframe (e.g. B8, C7). According to the informants, DT technologies could increase control over data fungibility. Cloud computing simplified the sharing of supplier-specific process data (A1). Similarly, the Internet of Things enabled the connectivity of customer-specific machines and deliveries (e.g. B2, B3, B5). This entailed some complexities; in the case of connected vehicles, it was too costly to isolate component data (e.g. C7).

Exemplary quotes for relational preconditions are presented in Table 4 according to the factors that pertain to the RDT. Regarding the need for complementary capabilities, data were shared whenever an external party was required to make sense of the data. For example, suppliers were engaged in analyzing data from defective components (e.g. C7). Conversely, the presence of internal capabilities reduced the extent of sharing (even for data not perceived as risky): smart maintenance was mostly performed in-house for simple machines (B9), fleet operators did not leverage OEMs’ services when they had in-house analytics (C1), and firms did not engage external parties in production planning (e.g. A3). Data were not shared whenever the problem-solving capabilities of the firm owning the data exceeded those of its business partners (e.g. B6).

Interestingly, several OEMs and first-tier suppliers developed data management and analytics capabilities to keep pace with DT (B8). Nevertheless, it was still common for manufacturers to seek external help, especially when initially embracing DT and new technologies (e.g. machine vision). Data were shared to adjust collection and analysis procedures (e.g. B8, B9 and C4). Large manufacturers looked for ad hoc collaborations, whereas smaller companies were likely to have their data wholly managed. Similarly, it is important to note that data were shared either continuously (e.g. the algorithm for processing connected car data was operated by the supplier/technology provider – C4) or for the time needed to set up a specific project (e.g. when connecting production machines and building their digital models – B8). Another peculiarity was cases of sharing due to complementarity among data sets owned by different firms, especially when using artificial intelligence and machine learning approaches. Examples related to matching product/process data for quality improvements (e.g. B6) and large-volume logistics and connected vehicle data (e.g. A6 and C6).

Concerning the level of dependence and dependence asymmetry, our evidence corroborates previous views, with limited new elements. The case findings confirmed the link between relational orientation and data sharing. The presence of strongly entwined business processes – as in the case of JIS/JIT models (B8) – determined more widespread practices, especially those related to real-time integration. A high level of dependence was also conducive to easier adoption of common data formats and interfaces, for example, in logistics (e.g. A4, A5) and with respect to the blockchain (A7). Similarly, dependence asymmetry was a consistently relevant factor, as large and powerful buyers more easily requested data from suppliers and logistic operators and had them adopt their technological solutions (e.g. A2, A4, A5, A7 and B5).

In terms of protection mechanisms, we found the application of usual contractual clauses, although a slow update of corporate policies was identified as a roadblock. Specific to DT, the use of new technologies was associated with higher data security and confidentiality (e.g. A8, C5 and C6). Moreover, as intermediaries within the SC, digital platforms were indicated as a solution to benefit from visibility while avoiding connected risks (e.g. A3). New legal arrangements were also discipling the use and ownership of shared data, for example, data pooling with competitors for joint analytics and sharing with third parties (e.g. B8, B9, C1, C3 and C4).

Regarding benefit distribution mechanisms, we did not find agreements besides joint ventures. Many comments indicated that data were shared whenever the benefits were mainly appropriated by the sharing firm (e.g. B8, B9, C1, C5 and C6).

We framed evidence of network-level dynamics according to CAS (Table 5). On a general level, data sharing practices developed according to the existing network governance structure. The relatively low spread of data sharing practices concerning upstream process visibility, traceability and quality control (e.g. A2, B5 and B6) were often attributed to a limited interest by OEMs in improving suppliers’ efficiency, whereas the automotive tiered structure indicated a strong role of first-tier suppliers in upstream SCM.

The impact of DT was significant for the connected product data. On the one hand, the increasing prevalence of electronic components meant higher integration efforts; thus, OEMs required data from first-tier suppliers (C7). On the other hand, OEMs orchestrated connected vehicle data flows for service provision, thus capturing the value generated.Whereas data were shared to enable partners’ service provision, OEMs maintained ownership and control (C1, C2).

Factors related to environmental complexity and dynamism were also reported. Higher demand unpredictability determined by product customization and the risk of disruptions (e.g. in the aftermath of the COVID-19 outbreak) prompted firms to focus on upstream SC visibility (A2, A3). Similarly, traceability initiatives were called forth by higher product complexity (A7). The effects of DT appeared related to the pace of innovation; thus, manufacturers increasingly shared data with technology providers (e.g. B8, B9 and C4).

Finally, several comments pointed to the assimilation of technological and communication standards. The lack of standards caused frictions, as each player would rather use their own system. Although some standardization efforts were ongoing, the informants perceived that they would not cover all possible data and parameters. The evidence showed that some challenges could be partially overcome by implementing solutions specific to DT. Concerning connected assets/equipment, industrial technology providers became system integrators (B8, B9), while digital platforms and blockchains offered new ways to share data (e.g. C5, C6, A7 and, to a limited extent, A3).

One key question that arises in the context of DT is the alignment between theoretical models and the dynamics unfolding in this evolving landscape (Hanelt et al., 2021; Culot et al., 2020a). In the face of today’s profound transformations, there is a call for potential new theoretical perspectives (Hendriksen, 2023; Paolucci et al., 2021). However, it is essential to commence the journey by assessing compatibility with established models to prevent hypes and misunderstandings. Starting from a comprehensive review of the relevant SCM literature, supplemented by concepts from adjacent managerial disciplines, we developed an initial conceptual framework (Figure 1). This served as the lens through which we interpreted the empirical evidence gathered from 16 firms within an extended supply network in the automotive industry.

The findings outlined in subsection 4.1 reveal that the untapped potential of new digital technologies for interorganizational data sharing in SCs persists. Except for the cost of technology and the limited digital maturity of firms, informant comments suggest that overcoming these challenges may not merely be a matter of time. In line with theoretical predictions, the interviews underscored that data sharing is influenced by factors related to their characteristics (RBV), relational preconditions at the dyadic level (RDT) and network-level dynamics (CAS). Subsection 4.2 delves into the specifics of these dynamics in the context of DT.

The results are synthesized in Figure 2, which illustrates the interpretation of the novel elements based on the three theoretical perspectives underpinning this study. Adhering to Busse et al.’s (2017) methodological recommendations to handle contextual idiosyncrasies as boundary conditions (Bacharach, 1989; Dubin, 1969), we accommodate the impact of DT by revising relationships and constructs. Relationships are refined through moderators (mod) (describing a different shape and intensity of causal relationships; Bamberger, 2008) and mediators (med) (accounting for different causal pathways; Busse et al., 2017). Constructs are reformulated to capture changes in their meanings (Suddaby, 2010).

Regarding core-relatedness, in many ways, DT is simplifying data sharing practices that were pursued even before the advent of new technology, although less efficiently, accurately and timely. Growing servitization trends are driving a redefinition of what represents core business in the automotive industry alongside many other industries (Culot et al., 2020b; Bohnsack et al., 2021; Peerally et al., 2022). As highlighted in the literature (e.g. Chen et al., 2021; Dalenogare et al., 2022; Porter and Heppelmann, 2015), servitization implies that data are shared for service orchestration and provision, which can be absolved by the initial firm receiving the data (e.g. OEMs) and/or dementated to third parties (e.g. service centers, insurers, technology providers). Servitization opportunities are not limited to connected products but also affect processes (e.g. outsourcing of production and logistics planning for data-driven optimization) and assets/machines with process optimization and maintenance (Dalenogare et al., 2022; Cepa, 2021). Against these opportunities, however, the principle of core-relatedness holds, as firms hesitate to share data closely tied to their core business (e.g. production parameters), regardless of potential benefits. Similar considerations arise when considering services that firms are offering/developing internally or rather through external parties. In Figure 2, we propose a causal pathway in which firms decide on DT-related servitization opportunities based on the assigned value of the data to be shared.

Considering data fungibility, concerns persist about sharing data, potentially leading to opportunistic behavior in buyer–supplier negotiations and new product development. We also noted that firms can now share data with greater granularity and specificity than in the past. This aspect is peculiar to direct connectivity (Culot et al., 2020a; Münch et al., 2022) and allows, for example, suppliers to share data generated by assets/machines operated exclusively or predominantly for one customer without disclosing other information. A similar effect is obtained through cloud computing, as it enables access to a specific data environment while preventing the unauthorized use of data. We view DT as amplifying firms’ control over data fungibility, consequently increasing data sharing opportunities. In Figure 2, we introduce a moderating effect.

Data sharing still hinges on firms’ need for complementary resources to interpret and use their data. Importantly, DT is raising the bar for these capabilities, now encompassing expertise beyond the scope of many manufacturing firms (e.g. data science and system integration) and scarce on the market (Cepa, 2021; Münch et al., 2022). In Figure 2, we thus introduce a moderating effect of DT, suggesting that with increased adoption of technologies involving data analytics, firms are more likely to engage in data sharing with specialized external parties.

In addition, we identified other peculiarities of DT that can be reconciled with the general logic of RDT through construct refinement. We propose the introduction of amendment 1 in the independent variable to consider complementarity between data sets with similar (e.g. connected vehicle data within data aggregation platforms) or different scopes (e.g. product quality and suppliers’ process data). Technologically, this is motivated by the network effects of artificial intelligence, whose value is enhanced by more data (Gregory et al., 2021; 2022). Furthermore, we advocate for construct refinement 2 to distinguish between data sharing and data management externalization facilitated by cloud computing (Novais et al., 2019). This occurs when firms, especially smaller ones, lack the necessary data infrastructure and have digital platforms that serve as data aggregators. Moreover, in amendment 3, we incorporate a temporal dimension to account for varying degrees of data sharing. For instance, during the implementation of asset/machine and product connectivity, manufacturers can either share data in the ramp-up phase or continuously (i.e. algorithms are operated by the technology provider).

Concerning the level of dependence and dependence asymmetry, our study confirms the relevance of these factors while not highlighting any major discontinuity in DT. Despite the potential for end-to-end SC visibility (e.g. Calatayud et al., 2018; Dolgui and Ivanov, 2022) and the advent of “trustless” systems (Babich and Hilary, 2020), data sharing still mostly occurs between business partners linked by a direct relationship. Even the decision to adopt multi-tier data sharing technologies (such as the blockchain) is motivated by prior SC relationships.

Regarding protection mechanisms, technologies enhance data security and confidentiality, resulting in increased efficacy in cybersecurity solutions (Corallo et al., 2020). We propose a moderating effect hypothesis to account for this. Additionally, DT is linked to the rise of digital platforms acting as “middlemen” for sharing sensitive data (Legenvre and Hameri, 2023). To address this, construct adjustment 4 is suggested to incorporate protection mechanisms for both dyadic data sharing and digital platforms as an emerging organizational solution. Finally, construct refinement 5 highlights the importance of legal arrangements specifying ownership of shared data and acknowledging potential variations in risks and benefits for the sharing firm.

Finally, for the last factor pertaining to RDT, we could not find any major implications of DT in terms of benefit distribution mechanisms. This aspect, despite being amply advocated in the literature (Ganesh et al., 2013), is still mostly neglected in practice.

The network governance structure is normally understood in manufacturing as the distribution of coordination responsibilities (Choi et al., 2001; Gereffi et al., 2005). Connected products and related servitization call for greater attention to aspects such as customer relationships (i.e. who signs in for data access with the final user) and value capture (i.e. how to protect the margins related to services). According to the literature (e.g. Hanelt et al., 2021; Porter and Heppelmann, 2015), data sharing is inherently higher when implementing connected products, and it is important to consider the overall governance of data flows. In Figure 2, we suggest different pathways, depending on whether DT is implemented in support of manufacturing processes or for servitization opportunities.

In terms of environmental complexity and dynamism, our results indicate that a high pace of innovation requires collaboration based on data sharing, thus suggesting a moderating effect. The assimilation of technological/communication standards also encompasses interoperability initiatives and the creation of data repositories. Technologies such as blockchains build interoperability layers between different firms’ information systems (Nandi et al., 2020; Sauer et al., 2022). Moreover, although the presence of intermediaries in SC relationships is not new (e.g. Gast et al., 2019; Mena et al., 2013), digital platforms can substitute or partially compensate for the lack of industry-wide standards and should be accounted for in the independent variable (construct refinement 6).

Returning to our main RQ, DT is changing some aspects of interorganizational data sharing, as new technologies allow the sharing of data with greater granularity, volume and timeliness, which better align to the needs of SCs in fast-changing environments and provide new avenues for value generation. Nevertheless, managerial decisions remain rooted in common business sense and consolidated SCM practices. Overall, innovative aspects are emerging, although they are still categorized within the overarching framework of grand theories. Namely, DT can positively moderate the relationship between certain factors and data sharing (e.g. by prompting companies to look for data-driven collaborations and by offering new ways of sharing and protecting data). In other cases, DT can be seen as a mediator (e.g. when data are shared consistent with a redefinition of core competences). More nuances can then also be captured by working on constructs (e.g. by specifying different forms of data sharing, such as the externalization of data management).

This study clarifies the factors that drive or prevent interorganizational data sharing in the context of DT. Building on the key learnings from prior research, we developed a case study analysis within an extended supply network in the automotive industry with the aim of theory elaboration (Dubois and Gibbert, 2010; Ketokivi and Choi, 2014). The findings indicate that interorganizational data sharing today can still be explained through established theories (i.e. RBV, RDT and CAS). However, there are some situations peculiar to the new contextual conditions. These were examined to provide middle-range theoretical insights into the phenomenon (Busse et al., 2017; Merton, 1957; Mintzberg, 1977). An integrative framework (Figure 2) illustrates how DT partially changes the meaning of previous constructs and their relationships. We account for these through the refinement of definitions and by positing new moderators and mediators (Bamberger, 2008; Suddaby, 2010).

This study makes two distinct contributions to the academic debate on DT in extended supply networks. First, we draw attention to the issue of interorganizational data sharing, which up until now has not been treated with all the due breadth and depth. By making explicit the key dimensions and factors, our study provides a structured perspective that is applicable to academic research across the range of technological opportunities at hand.

Second, we integrate three theories central to the study of interorganizational relationships (i.e. RBV, RDT and CAS). The application of these theories to our empirical evidence showed that data sharing is inherently a multi-faceted phenomenon that ought to be examined from different points of view. This is also relevant in the light of recent calls for a better understanding of the fit between established models and emerging trajectories (e.g. Hanelt et al., 2021). We present some approaches to increasing their accuracy and applicability in today’s business environment. Despite the alleged disruptive impact of DT and the allure of new theoretical perspectives, we show that researchers can still resort to grand theories to make sense of managerial decisions. However, we also acknowledge that some elements represented a higher degree of novelty with respect to already theorized dynamics. These refer to opportunities to share data in different forms (i.e. at the individual level, in real time, in bulk and anonymized) as well as to fully outsource data management. We accounted for this by arguing for a refinement of the construct “data sharing” based on the specific context and situations. These differences should be made explicit in future studies investigating this phenomenon.

From a managerial perspective, the main contribution of this study is that it clarifies what firms’ decision makers should consider when approaching the data sharing conundrum. Until now, the discussion has been rather polarized. On the one hand, managers have been perceiving data as “the new oil.” On the other hand, emerging narratives of digitally integrated SCs and ecosystems have often downplayed potential risks. Hence, the explication of key dimensions and factors can support a more structured and analytical approach to the issue. This can also be useful in deciding whether to invest in new interorganizational technologies that might not be welcomed by the firm’s business partners. Practically, managers can take our final framework (Figure 2) as an analytical tool to decide whether to adopt a specific technology as well as to assess the risks of taking part in a data sharing initiative. Moreover, by reporting the experiences of 16 firms, we provided several practical examples.

Further empirical studies are needed to improve and refine our elaborations. Our decision to focus on a single industry, although methodologically justified to keep extraneous variables under control, has the drawback of not allowing a comparison between contexts. Similarly, we should acknowledge the possible effects of the COVID-19 pandemic, which contributed to creating a sense of urgency around SC visibility. Due to the use of theory, the analysis minimizes these drawbacks; however, it can be appropriate to investigate other settings. In the study, we pursued analytical generalization, that is, the process of generalizing from empirical observations to theory rather than to a population (Gibbert and Ruigrok, 2010; Yin, 2018). In this respect, our final framework (Figure 2) provides an updated view of the factors identified by prior literature in the context of DT. Constructs and relationships can extend across industries; however, there might be a need to verify specific dimensions, mostly those related to network structure and governance, and to consider DT in industries where product connectivity is not extensively pursued. We thus suggest cross-sectional studies of data sharing practices related to specific areas of interorganizational collaboration (e.g. logistics, planning, service provision). Additional insights might emerge by leveraging other theoretical underpinnings, such as transaction cost economics (Coase, 1937; Williamson, 1985), agency theory (Eisenhardt, 1989a; Mitnick, 1975; Shapiro, 2005), the information processing view (Galbraith, 1974; Tushman and Nadler, 1978) and institutional theory (DiMaggio and Powell, 1983). Moreover, further theoretical insights are possible at the interface between RBV and dynamic capabilities (Teece et al., 1997) when investigating the reasons leading firms to establish different kinds of external collaborations and share data in the context of DT. Finally, the still limited adoption of some new technologies calls for further updates. Extensive testing of results to overcome the limitations peculiar to qualitative studies is also a clear avenue for future research.

Table A1