BDA is defined as “the application of statistical, processing, and analytics techniques to big data for advancing business” (Grover et al, 2018, p. 390), whereby big data can be delimited from ‘normal’ data by its volume, velocity, variety, and veracity (Abbasi et al, 2016). It is widely recognized as a cornerstone of contemporary digital transformation efforts of organizations and viewed as a key enabler and catalyst that allows companies to become more data-driven and agile in responding to market dynamics. BDA promises the effective use of the vast and ever-growing amount of data within organizations by detecting and recovering hidden patterns, trends, or customer preferences that inform strategic decisions. Companies implement BDA to enhance decision-making, optimize business processes, improve products and customer experiences, and ultimately drive business value (Côrte-Real et al, 2019; Grover et al, 2018; Günther et al, 2017; Mikalef et al, 2020; Newell and Marabelli, 2015; Sharma et al, 2014). For example, retail giants like Amazon use BDA to analyze customer behavior, optimize inventory management, and provide personalized product recommendations in real-time. Not least in the age of artificial intelligence, BDA frequently leverages machine learning and other AI algorithms to process data and generate predictive or prescriptive insights (Susanto et al, 2024).

The term ‘adoption’ already indicates the early stage of integrating these technologies within the organizational context (Cooper and Zmud, 1990), which initiates massive transformation efforts throughout the entire organization (Dremel et al, 2017). While the potential benefits of BDA are compelling, previous research has identified several challenges organizations face when implementing this technology, comprising both technical and organizational aspects. While on the technical side, issues such as data privacy and security concerns, data quality and integration issues, and infrastructure and scalability are highlighted, potential hurdles on the organizational level, such as talent and skills gaps or organizational culture, are also addressed (Sivarajah et al, 2017; Susanto et al, 2024). Previous research referring to organizational aspects has studied BDA adoption mainly concerning the overarching organizational level as well as the individual level (Sidorova et al, 2008). On the organizational level, scholars have explored the antecedents of organizational BDA adoption intention (Caesarius and Hohenthal, 2018), starting points for BDA adoption (Bremser, 2018), and drivers of postadoption usage (Obal, 2017). Furthermore, scholars have also explored the impact of BDA adoption on firm performance (Gunasekaran et al, 2017) and digital transformation (Dremel et al, 2017). In addition, on the individual level, challenges have been identified that impede individual adoption (Beath et al, 2012; Clarke, 2016; Sivarajah et al, 2017). Moreover, scholars have explored factors underlying individual adoption of BDA (Boldosova, 2019; Verma et al, 2018).

While these studies provide valuable insights into BDA adoption at both macro and micro levels in the organizational context, the group level—as the essential level for developing and delivering BDA applications (Carton and Cummings, 2012)—remains largely underexplored. The development of BDA solutions within organizations requires close collaboration within an interdisciplinary and cross-functional work setting. Usually, a BDA work team consists of business experts and data science experts (Boldosova, 2019; Gupta and George, 2016). Business experts (from here: BE) are managers from diverse business units, such as marketing or sales, who aim to use BDA to improve their operations and decision-making. However, their role is not limited to the use of the final application; rather, they are deeply involved in the conceptual development and functional design, providing requirements and feedback around the respective use cases. Data science experts (from here: DS) are, for example, data scientists and data engineers who contribute their analytical and technical understanding to extract knowledge from data (Michalczyk et al, 2021). Thus, their role is mainly the technical development and realization of the application in close exchange and collaboration with BE. Research suggests that the collaboration between BE and DS is uneasy (Barbour et al, 2018; Hagen and Hess, 2021). On the one hand, this tension arises from the different expertise domains and understandings, leading to different languages, incongruent mindsets, and unrealistic expectations (Hagen and Hess, 2021). For example, BE often aims for quick returns on investments, whereas DS aims for accurate results (Yamada and Peran, 2017). On the other hand, tension arises from the shaken power structure during BDA adoption. For example, BE may tend to hoard their data to maintain their influence in the organization, fearing that algorithms make their judgment obsolete (Barbour et al, 2018).

The relationship between BE and DS bears similarities to the cooperation and coordination between business departments and information technology (IT) department members as functionally distinct groups. This issue has kept scholars on tenterhooks for several decades, revealing challenging group dynamics due to different perspectives and interests (Chan, 2008; Van den Hooff and De Winter, 2011). However, while the constellation between business and IT might be similar, it is not equal to the context of BDA adoption. In the former constellation, IT clearly assumes a supportive role, constituting an asymmetrical group constellation. IT’s primary focus is on providing and maintaining the technical infrastructure and ensuring the smooth operation of systems for business. In contrast, in the case of BDA adoption, DS, as a new professional group within the organization, play a pivotal role. Therefore, we propose that this relationship reflects a qualitatively different constellation, akin to the widely accepted distinction between IT-enabled organizational transformation and digital transformation (Besson and Rowe, 2012; Vial, 2019; Wessel et al, 2021), and thus warrants dedicated scholarly attention. The development of BDA applications and the leverage of the full potential of BDA calls for close and intensive interdisciplinary collaboration of DS and BE in a cross-functional work team, as the expertise and competencies of both expert groups are necessary. Unlike people in IT departments, data scientists are more actively engaged in the core business by leveraging data to drive strategic decisions, optimize processes, and foster innovation. Their work goes beyond mere support and service for business, as their task is to analyze complex datasets to uncover insights that directly influence business outcomes, product development, and customer strategies to enhance overall organizational performance. Close collaboration on eye level between BE and data scientists is essential for a deeper understanding of relevant data sources, ensuring the quality of data, and facilitating the successful implementation of insights into business operations. By working together, both groups jointly develop and iteratively refine models that ensure their applicability in real-world settings. To this end, a functioning BDA work team is the foundation to ensure that BDA solutions align with the company’s strategic goals and address practical needs. Thus, in contrast to the business-IT relationship, a more symmetric group constellation can be assumed in the case of BDA adoption.

These insights suggest that adopting BDA needs to be examined at the group level. Shedding light on social group practices between BE and DS and revealing underlying drivers for these practices contributes to a deeper understanding of emerging group dynamics in the light of digital technology-induced organizational change and the associated increasing role of DS, and thus provides indications for the success or failure of digital transformation initiatives.

To theoretically grasp group dynamics during BDA adoption, we combine SIT (Tajfel, 1978; Tajfel and Turner, 1979; Tajfel and Turner, 2004) with a practice theoretical perspective (Bourdieu, 1977; Giddens, 1984). While SIT serves as a theoretical lens for closely examining group dynamics in terms of a ‘middle-range theory’ (Merton, 1968), we refer to practice theory as an epistemological encompassing theoretical frame, understanding the group level in organizations as a mutual interplay between the social structures (organizational level) and social practice (individual level). The organizational group level, thus, displays the meso level, where the complex and mutual interdependence between social structure and social practice takes place. In general, social practices from this perspective are understood as patterns of human behavior or activity that are organized around collective understandings. Groups serve as the primary context within which social practices are developed, maintained, and transmitted. Through group interactions, individuals negotiate meanings, establish norms, and reinforce behaviors that constitute specific practices and, thus, social structures.

SIT draws our attention to the relationship within and between groups, focusing on inter- and intra-group practices based on group construction and identity, helping to trace group dynamics and group constructions. Whereas the practice theoretical lens enables us to unpack underlying collectively shared understandings and explicitly consider organizational structures. Within the broad approach of practice theory, we particularly refer to Bourdieu’s idea that social practice is neither purely individual nor wholly determined by social structures (Bourdieu, 1977; Bourdieu, 1990). Rather, it emerges from the interplay between habitual dispositions and the positioning within power relations. In particular, we utilize Bourdieu’s concepts of habitus and field to better understand why group dynamics emerge the way they do in the context of BDA adoption. Even though Bourdieu himself did not directly address group practices in the organizational context, following Emirbayer and Johnson (2008) as well as Sieweke (2014), we suggest that this perspective provides a suitable and promising theoretical lens to study organizational phenomena.

The concepts of social identity and group habitus obviously share similarities. Both refer to commonly shared beliefs, values, presuppositions, and affective dispositions among group members and provide explanations for group dynamics. However, two slight differences in focus and conceptualization are to be considered: On the one hand, social identity is primarily related to an individual’s self-categorization and thus implies a stronger focus on consciousness, whereas the concept of habitus emphasizes the pre-consciousness of dispositions, becoming internalized in the course of processes of socialization and experience. On the other hand, as Carter and Spence argue, SIT “tell[s] us much about the ‘here and now’ but are far less sensitized to the historical development of identity or of the context that sustains an identity” (Carter and Spence, 2020, p. 272). In this vein, we suggest that both theoretical approaches can cross-fertilize each other. While the SIT perspective helps to trace inter- and intra-group practices during the BDA project, shedding light on in- and out-group constructions, the Bourdieusian perspective enriches this analysis by pointing to the underlying habitual dispositions that constitute the group’s social identity, on the one hand, and by integrating the organizational context and considering broader field dynamics and power relations, on the other, in order to gain deeper understanding of the driving forces for the observed group practices.

For our study, we, therefore, regard the company as a social field. The adoption of BDA initiates a digital technology-induced organizational change, potentially leading to shifts within the field structure. Since, in the course of BDA adoption, DS gain importance as a new organizational group, the established field order might be questioned and become subject to re-negotiation. BDA adoption is realized through BDA projects, where BE and DS are unified into a BDA work team, constituting a new interdisciplinary group within the organization. We expect that within the BDA work team, emerging group dynamics are anchored in different group identities and habitus, as well as in their respective position within the organization as a field. Thus, to gain a comprehensive understanding of emerging group dynamics, group construction, and identity, as well as underlying group habitus and group positioning, must be taken into consideration.

We selected ToolCo for our study, as it fitted our selection criteria. These were (1) the organization must adopt BDA, (2) the organization has at least one BDA project in the backlog that we can accompany throughout its entire lifecycle, and (3) the project requires collaboration between BE and DS. ToolCo is an international retailer and manufacturer specializing in precision tools and workplace equipment for various industries (e.g., drills, cutting tools, measuring and testing equipment, personal protection equipment). They offer a combination of premium tool brands, their own house brands, and expert consulting services. The company is headquartered in Europe, has almost 4000 employees across 50 global locations, and generates annual revenue of more than 1 billion Euros. ToolCo integrates e-commerce, catalogue sales, and personalized customer service. In the course of its history, the sales department was the powerhouse of ToolCo, with a strong position within the organization. About three years ago, since several executives recognized the potential of BDA, a centrally anchored data science team was built, aiming at delivering case-based BDA prototypes for the business units, for example, for sales forecasting. The members of the data science team are not assigned to the IT department but belong to the core business department.

We selected a project called “content automation tool” (CAT), which is a standard BDA process improvement project (Grover et al, 2018). The project goal is the development of a supervised machine learning tool to accelerate the onboarding of supplier content in ToolCo’s online shop. The content onboarding process comprises, e.g., the integration of product data, including descriptions, specifications, images, and categorization, into the company’s e-commerce platform. The project goal is driven by the business problem that the onboarding of supplier content into ToolCo’s structure systems (e.g., product information management system) goes along with a high manual classification effort. Classification is necessary to position the supplier’s products within ToolCo’s portfolio and to properly market them online. Prior to CAT, an editor needed 3.9 minutes to classify one article. Given that ToolCo’s product portfolio includes more than 300,000 articles, this process requires a considerable amount of time and should be automated via machine learning.

For CAT, two relevant use cases have been created: Use case 1 includes articles with labeled supplier data, which have been classified before. Here, CAT is supposed to identify new data based on historical data and classify it accordingly. The second use case includes the classification of unlabeled supplier data based on patterns. This distinction of labeled versus unlabeled data turned out to play a key role in the project (see results). CAT followed the typical phases of analytics projects in large parts (Mariscal et al, 2010; Phillips-Wren et al, 2015), as shown in Table 1. However, throughout the project, group dynamics within the BDA work team emerged that impeded collaboration and recurring misunderstandings, which resulted in suboptimal project progress. While the primary objective was to implement both use cases, by the end of the project period, the project stood behind its expectations, and only the first use case was realized and implemented.

To investigate the group dynamics during the BDA project, a combination of ethnographic research methods was used (Ybema et al, 2009). This approach is particularly suitable for understanding group dynamics and practices from the perspective of participants, as it enables the tracing of observable as well as informal aspects of group dynamics. For data collection, one of the authors accompanied the BDA work team as a non-participatory observer throughout the entire project trajectory. One exception to the non-participatory observer role was the moderation of the group discussions, as outlined afterward. throughout the entire project trajectory. Data collection took place over 18 months (March 2020 until August 2021). Data analysis and interpretation were conducted by the whole research team, consisting of three researchers in total. The research objective was thus investigated from both the inside and outside (Evered and Louis, 1981). The two major subgroups involved in CAT were employees from the content management team, representing BE, and the DS team. Table 2 gives an overview of the BDA work team and the interaction frequency during data collection. The identification codes in Table 2 are used as quotation sources in the Results section.

The data used to study the group dynamics were obtained from diverse sources listed in Table 3 below. Primary sources were formal and informal conversations and meeting observations (Myers and Newman, 2007). To triangulate this data and to stimulate group discussion, the BDA work team members additionally answered a standardized satisfaction survey three times during the project (see Appendix Table 7). The results were discussed in group meetings, which enabled the observation of group dynamics beyond the daily project business. Since the BDA work team mainly worked remotely and ToolCo’s operations took place virtually via Microsoft Teams, the authors had the chance to observe the meetings online. The research interactions were recorded and transcribed verbatim, except for three informal conversations, which were reported from memory in terms of field notes. Data was generated from all stakeholders listed in Table 2, but, depending on their role, at different frequencies and depths. Furthermore, the field researcher had access to the Confluence tool (an integral digital workspace and collaboration platform for communication, document sharing, and collaboration) and CAT. Moreover, press releases were studied to establish a broader understanding of ToolCo as an organization. For data triangulation and insights regarding secondary socialization, social media profiles (LinkedIn) of the work team were analyzed.

We used thematic analysis to analyze our data (Braun and Clarke, 2021; Guest et al, 2012). Thematic analysis both mirrors reality and unravels the surface of reality. This resonates well with our objective to identify group dynamics in terms of observable practices as well as latent, more preconscious patterns of perception, thought, and appreciation. As an approach for analyzing “both implicit and explicit ideas within the data” (Guest et al, 2012, p. 9), Thematic analysis has been widely used to study group identities and habitus (e.g., Kerr and Macaskill, 2020). In our analysis, themes pinpoint repeated patterns of observable practices and (pre)conscious dispositions of the subgroups. For this, we differentiated between semantic themes and latent themes (Boyatzis, 1998). Semantic themes tie to the observable practices (“what”), representing the “explicit or surface meanings of the data” (Braun and Clarke, 2006, p. 84). Latent themes tie to the dispositional patterns (“why”) and require interpretation to go beyond what a participant has said. The coding of recurrent themes was first done deductively, using our research question as a start. However, since we searched for practices and dispositional patterns as grounded in the data, we mainly employed inductive coding. Axial coding was employed to systematically develop themes and subthemes and to organize them according to relationships between codes. To ensure stability and intercoder reliability, the researchers did the coding procedure individually (Krippendorff, 2013). During data analysis, we adhered to the fundamental concept of the hermeneutic cycle (Klein and Myers, 1999), constantly iterating between the interpretation of sequences and the situation as a whole. We developed a thematic map based on 210 quotes (Braun and Clarke, 2006), including the observable practices (semantic themes) and the habitual dispositions (latent themes). Table 4 provides coding examples.

To analyze group dynamics more closely, in the first step, we traced observable group practices. We reconstruct group practices that appeared between the two subgroups (BE and DS), referred to as intra-subgroup practices, as well as practices that emerged within the BDA work team as a whole, labeled intra-workgroup practices. Fig. 1 provides an overview of the identified group practices.

With respect to inter-subgroup practices, we categorized four distinct practices that appeared throughout the whole project period, which displayed group dynamics between BE and DS that turned out to be an obstacle to the project’s progress: (1) both subgroups refused the responsibility and ownership for the project. Each subgroup held the other responsible. (2) The subgroups were not able or not willing to really constitute a team, but stuck to their functional silos. (3) While both subgroups recognized a mutual lack of knowledge, they did not mitigate it properly. Instead of providing the other subgroup with explanations and information, both complained about the poor understanding of the other subgroup. (4) We frequently could observe that the subgroups talk past each other during their meetings. Instead of addressing latent conflicts such as their mutual unfamiliarity, they avoid discussing them openly. This, not least, led to a massive misunderstanding regarding the use case and development goal for almost 8 months.

These identified inter-subgroup practices obviously showcase distinctive practices occurring between both subgroups. Interestingly, we also found two practices towards unification and alliance formation at the work team level (intra-workgroup practices). This is, on the one hand, their alliance against internal organizational complexity. The work team members aligned to justify a lack of project contribution and progress by referring to complexity, either in terms of the complex organizational structures and processes or in terms of the complex data structure. On the other hand, they alliance against requirements from top management. For example, they stood together when it came to defending against the changing priorities or misplaced expectations of management.

Subsequently, we detail the observed practices. The described practices in subsections 4.1.1 to 4.1.4 refer to inter-subgroup practices between BE and DS, while the subsections 4.1.5 and 4.1.6 detail intra-workgroup practices at the BDA work team level.

Having outlined the observable group practices we could reveal during the BDA adoption project at ToolCo, in a second step, we trace our empirical material for more subtle themes that could help explain these group practices. Thus, this section focuses on underlying dispositions, providing potential indications and explanations for the described group practices. It became evident that both subgroups ascribe fundamental differences to each other, as aptly expressed in this statement: “We’re just two worlds colliding. It is inevitable that there are different views” (B3). Throughout the material, we could reveal four latent themes, with each subgroup displaying distinct, to some extent, opposing dispositions. These (pre)conscious themes relate to (1) value orientation, (2) understanding of teamwork, (3) working focus, and (4) mode of thought and expression. We interpret these latent themes as specific subgroup identities and habitual schemes that underlie and guide the observable group practices. In addition to these habitual dispositions toward identity and boundary work, we also revealed differences in symbolic positioning that reflected latent power asymmetries. These asymmetries were not formally assigned, but enacted through the ways each subgroup framed expertise, justified decisions, and claimed legitimacy during moments of conflict or ambiguity (see subsection 4.2.5).

Our observation clearly revealed that within the BDA work team, both BE and DS exhibit group practices that suggest they perceive and identify themselves as distinct subgroups. Moreover, these subgroup practices led to an in-group – out-group constellation within the BDA work team, as described, e.g., by Ashforth and Meal (1989). These dynamics, not least, impeded successful collaboration in the BDA project. The BDA work team members stuck to their subgroup by maintaining their functional silos and mutual gatekeeping of information (see subsections 4.1.1 and 4.1.2). Instead of constituting the BDA working team as one superordinate group by jointly taking over responsibility for BDA adoption in general and the CAT project in particular, a rather competitive relationship emerged, leading to an ‘us-versus-them’ mentality (Ashforth and Mael, 1989; Sidanius, 1999). Both subgroups frequently complain about the deficiencies and ignorance of the others, thereby implicitly devaluing the respective out-group. This behavioral pattern is, for example, reflected in their mutual blaming rather than addressing knowledge gaps or fostering a commonly shared understanding (see subsections 4.1.3 and 4.1.4).

Interestingly, this in-group – out-group relationship is carried out mainly within the BDA work team level. When the BDA work team is confronted with organizational pressure in terms of demands and expectations, the subgroups unite symbolically into one superordinate group. This pattern became visible when the BDA project team as a whole was confronted with top management demands. In such situations, the subgroups aligned symbolically and temporarily merged into a superordinate identity (see subsections 4.1.5 and 4.1.6). This reflects what Turner describes as contextual identity salience (Turner et al, 1987), where the level of self-definition shifts in response to situational cues. From a SIT perspective, such identity realignment can be interpreted as a functional response to external threat, in line with the “common ingroup identity model” (Gaertner et al, 1993), which suggests that intergroup boundaries may soften when a shared group identity becomes more salient.

The observed intra-workgroup practices can be interpreted as impeding practices with regard to the project, as these practices mainly aim at excusing and concealing a lack in the project’s progress. To a certain degree, the subgroups’ allied practices, i.e., allying against the top management and ToolCo’s internal complexity, are aimed at covering up the lack of project advancement and safeguarding against criticism. Although these practices cannot be described as supportive in the sense of project success, they indicate that there is a common denominator between the subgroups that becomes particularly salient in moments of group-external pressure.

Referring to Bourdieu’s concepts of group habitus and field, we can uncover potential driving forces that provide explanatory avenues to a deeper understanding of why the observable group practices emerge in the way they do. Our findings suggest that BE and DS show different social practices driven by distinct subgroup habitual dispositions. We suppose that these different attitudes, described in section 4.2, are rooted in past work-related experiences and their current organizational position. During education and work experiences, context-specific taken-for-granted assumptions become increasingly internalized and thus part of the group habitus and identity. Specifically, we presume that value orientation, working focus, and mode of thought and expression (see section 4.2) are considerably shaped by socialization processes within the respective work domains. In the case of DS, we suppose that their similar educational background has a strong imprinting effect on their habitual disposition. In contrast, while BE members have different educational backgrounds, we assume that their habitus is strongly affected by the organizational identity in general, due to their strong position within the company. This could explain why BE is more impact-oriented and often utilizes an instrumental focus, while DS has a more scientific orientation and is, therefore, more focused on methodical rigor.

To this end, the identified group practices are also expressions of the (sub)groups’ position within the organizational hierarchy, emphasizing the power dimension and the reciprocal relationship between the individual and organizational levels. As described in subsection 4.2.5, BE claim a strong power position within the company. Consequently, BE tried to defend its power position during ToolCo’s effort to transform into a data-driven organization. Their positioning is not least inscribed in BE’s dispositions that are tightly linked to a comprehensive business-centered attitude. Thereby, BE implicitly claims its perspective as the legitimate overarching perspective and thereby also subordinates DS under its interpretive predominance. Conversely, DS, as a newly established expert group within the company, held an inferior power position thus far. During the BDA project, they acted (pre-)consciously driven by their group habitus, which was, to a certain extent, more pronounced than that of the BE subgroup. However, their group practices can also be interpreted as an attempt to secure and enhance their own position and to gain legitimacy within the organization. To protect themselves against potential denials and appropriations from BE (e.g., lacking acceptance of their solutions and not recognizing them as equal partners), they built a kind of a ‘social cocoon’ (Anand et al, 2004; Ashforth and Anand, 2003; Greil and Rudy, 1984; Pratt, 2000) around their subgroup (even though not that severe or ideologically loaded as described in the literature). This ‘cocooning’ strategy by DS can be interpreted as an attempt to defend their group identity and increase the subgroups’ own power position. Our findings reveal that everyday practices such as cocooning or the assertion of interpretive authority can be understood as expressions of group-level dispositions and symbolic power struggles. These practices are not merely strategic or rationalized, but reflect deeper, habituated orientations shaped by organizational positioning. As Sieweke (2014) points out, this type of latent dispositions becomes empirically visible through patterned practices that reveal how actors navigate symbolic boundaries and legitimacy claims. In line with his interpretation, our study shows how symbolic struggles unfold through routinized interactional moves rather than explicit conflict. For instance, technical gatekeeping or withdrawal from negotiation can be read as practical enactments of group habitus and field-specific positioning. The respective group practices reflecting different (more or less conscious) strategies in the ongoing power struggles implied in the BDA adoption process, not least, led to unsatisfying alignment and collaboration during the CAT project. Consequently, both subgroups were only partially able to converge, living in two different worlds, unable to really understand the other group’s perception and approach. As such, both subgroups remained in their functional silos during their interaction. Rather than viewing group tensions as mere communication breakdowns or conflicting work styles, we derive that the structural positioning within the organizational hierarchy influences group dynamics. BE as the established group, seeks to maintain its dominance while the emerging group of DS employs legitimizing and protective strategies, such as information gatekeeping or functional siloing, to navigate their precarious status.

Fig. 2 illustrates our compiled findings on group dynamics during the BDA project, by visualizing the relationship between latent group dispositions, power positions, and observed group practices (upper part of the figure). Inter-subgroup practices can be understood as emerging from two underlying forces: the respective group habitus and the symbolic power positions of BE and DS within the organization. While the subgroups’ power position is associated with different strategies—BE aims at defending and preserving its power position, while DS aims at gaining legitimacy and relies on ‘social cocooning’ as a protection strategy—the divergent group habitus shape the concrete practices between both subgroups. At the same time, external pressure, particularly from top management and organizational complexity, triggers a symbolic unification (lower part of the figure). Both subgroups temporarily ally, exhibiting joint intra-workgroup practices to deflect criticism or conceal insufficient project progress. Together, these group dynamics illustrate how data-driven collaboration is shaped not only by functional differentiation but also by deeply ingrained dispositions and symbolic asymmetries, making group habitus and power relations key drivers of group dynamics in the BDA adoption setting.

This study contributes to research on group dynamics in organizations by shedding light on the intricate interplay between group identities, habitual dispositions, and power relations within interdisciplinary and cross-functional teams. While prior research has explored group dynamics primarily in the context of leadership (Barnett and Weidenfeller, 2016; Hoyt et al, 2006; Wolman, 1956), team performance (Kozlowski and Ilgen, 2006; Oyefusi, 2022; Priesemuth et al, 2013), and interdepartmental collaboration (Aime et al, 2014; Holland et al, 2000; Vad Baunsgaard and Clegg, 2013), our study extends this discourse by emphasizing the role of group habitus and power struggles in shaping intra-organizational relationships.

By integrating SIT and Bourdieu’s theory of practice, we offer a comprehensive perspective that captures both the conscious dimension of group dynamics with respect to group construction and group practices and the latent dimension of the group’s habitual dispositions. Our findings demonstrate that organizational subgroups exhibit distinctive group habitus, which not only informs their perception of tasks and responsibilities but also influences their interaction patterns and willingness to collaborate. These habitual dispositions shape how groups communicate, align their objectives, and negotiate their positions within the organization.

Furthermore, our study highlights how power dynamics within organizations manifest at the group level. By identifying habitual distinctions and power struggles as key determinants of group behavior, this study advances our theoretical understanding of how group dynamics contribute to (or impede) organizational change and innovation. This perspective not only enhances existing theories of group dynamics but also provides a socio-structural lens to understand interdisciplinary and cross-functional collaboration in complex organizational environments (Sieweke, 2014).

Moreover, with our study, we contribute to research on BDA adoption in particular (Boldosova, 2019; Dremel et al, 2020; Gunasekaran et al, 2017) by revealing social occurrences within the BDA work team that may impede BDA adoption and thus business value. We extend our understanding of how organizations, from a social perspective, should proceed/not proceed to extract value from BDA (Côrte-Real et al, 2019), thus addressing related calls for a stronger consideration of social aspects (Jones, 2019; Markus, 2017; Mikalef et al, 2020). Moreover, we suggest that the social occurrences must be considered in light of the organizational surroundings (Galbraith, 2014), as the above-mentioned power relations and the organizational structure are catalysts for habitus-driven practices. Our study indicates that the organizational setting reinforces the group habitus of DS as new organizational actors and, thus, the distinguishing practices between the subgroups. In our case, this organizational setting built an additional trench and fueled the DS’ cocooning tendency. We, therefore, put recommendations regarding a central anchorage of data science teams (Dremel et al, 2017; Schüritz et al, 2017) to the test. Hybrid approaches (Grossman and Siegel, 2014) where coordination is done within a central unit and business-related data science activities within the business units, promise to diminish the distinction, at least from a structural perspective.

We are convinced that our insights are helpful for organizations that adopt BDA, as our insights provide direct points of contact for strategic and operational interventions. Transferring our findings from the group level to practice, it becomes obvious that interdisciplinary and cross-functional groups need to be moderated carefully to diminish dysfunctional practices with respect to the success of BDA adoption projects. Therefore, it is important to consider the respective power positions of the subgroups within the organization to downsize struggles for power and position. One possible way would be to place emphasis on the uniting organizational identity that strengthens similarities and reduces differences between the subgroups. The organizational identity may serve as a shared basis for interpretation and sensemaking, and misunderstandings can be diminished. As to concrete measures, a change management program is needed that not only considers the different profiles of the groups but also addresses their individual needs and concerns. Specifically, enabling interventions may help foster mutual understanding. One way forward is mutual competency acquisition: Basic data science competencies can be developed by business through training (i.e., which questions can be answered, what are typical applications, methodological requirements, limits, etc.), and DS need to be enabled toward business, including deeper knowledge regarding operational processes and value drivers, for example, via job shadowing. Moreover, collaboration structures, such as joint processes and collaboration tools, must be provided by the organization and not left to chance.

We acknowledge that our data is derived from only one case, which limits the generalizability of our findings. Especially, in contrast to companies facing the challenge of digital transformation (Vial, 2019; Wessel et al, 2021) we expect different practices within born-digital organizations, where a higher level of data-related competencies can be expected, making business and data science less distinct. Here, a deduction of best practices would be fruitful. Furthermore, incorporating the effects of different organizational settings on the practices may be promising. Moreover, by mainly focusing our analyses on potentially challenging group practices, we may have paid little attention to supporting practices that may occur during BDA adoption. Finally, while focusing on group habitus, individual differences in personality and behavior, and intra-subgroup dynamics are potentially underestimated influences. Consequently, we encourage future studies that deal with the social aspects of digital transformation and technology adoption.

In closing, we found that BDA adoption appears as a complex interplay between subgroups’ habitual dispositions and their position within the organization’s power relations, which potentially sparks conflict-laden group dynamics. To derive value from BDA, organizations must address not only data-related, procedural, or management challenges during adoption but also social aspects. That is, the distinctive group identities and habitual dispositions of DS and BE, and the potentially resulting tension-laden interaction practices, have to be addressed. Furthermore, those involved must anticipate that change-induced renegotiations of established organizational hierarchies and power relations might occur. Moreover, organizations must understand how their organizational arrangement interrelates with these social aspects, which may act as a catalyst for social conflicts.