Implementing digitally enabled collaborative innovation: A case study of online and offline interaction in the German automotive industry

In the context of implementing collaborative innovation, a range of digitally enabled infrastructures impact core organizational activities. Automotive manufacturing is one such industry where competitors now openly collaborate, facilitated through new technologies, in an effort to enhance collective innovation systems. We conducted a longitudinal case study of the first open innovation network in the German automotive industry to determine how online and offline channels interact to fuel firms' joint search for external ideas. Delving into the physical, virtual and cognitive enablers of collaborative innovation, our findings suggest that, while online platforms can help to facilitate knowledge sharing processes that promote collaborative innovation, firms implementing digitally enabled collaborative ideation need to develop additional mechanisms based on stronger offline interactions. As such, our findings contribute to a better understanding of how online technologies can facilitate knowledge sharing processes to enhance collaborative innovation.

In the context of implementing collaborative innovation, a range of digitally enabled infrastructures impact core organizational activities. Automotive manufacturing is one such industry where competitors now openly collaborate, facilitated through new technologies, in an effort to enhance collective innovation systems. We conducted a longitudinal case study of the first open innovation network in the German automotive industry to determine how online and offline channels interact to fuel firms' joint search for external ideas. Delving into the physical, virtual and cognitive enablers of collaborative innovation, our findings suggest that, while online platforms can help to facilitate knowledge sharing processes that promote collaborative innovation, firms implementing digitally enabled collaborative ideation need to develop additional mechanisms based on stronger offline interactions. As such, our findings contribute to a better understanding of how online technologies can facilitate knowledge sharing processes to enhance collaborative innovation.  (Heil & Bornemann, 2018;Najafi-Tavani et al., 2018).
Driven by advances in digital platforms, new ways to organize CI have emerged over the years that facilitate the seamless crossing of organizational boundaries. Enterprise social software solutions such as Yammer, Jive and Chatter enable new channels of interaction among employees, customers and management, while facilitating and even democratizing the decision-making process when it comes to collaborative idea scouting and selection. These developments have been accelerated by the COVID-19 pandemic. With the mandated switch to remote working, online platforms may be the sole mechanism for innovation partners to collaborate nowadays.
While interactions through online platforms for the purpose of CI have risen significantly, relatively little is known still about how firms-and their employees-adjust themselves to these changing digital ecosystems (Dahlander et al., 2021;Rangus & Černe, 2019). A tension between the online and the offline workplace may emerge when individuals cross virtual and organizational boundaries as part of their CI endeavours. Thus, further insight is needed on how modern technology facilitates emerging innovation ecosystems (e.g., Autio & Thomas, 2014;Ritala et al., 2013). Post COVID, employees will probably slowly return to working and interacting physically in shared office spaces, instead of solely relying on a digital environment for their functional and creative interactions. Stakeholders need to understand the limits of online interaction and how offline channels can enhance such initiatives, in order to advance future CI initiatives. In the spirit of recent calls in this journal to delve into the physical, virtual and cognitive enablers of CI (e.g., Leminen & Westerlund, 2019), the objective of our paper is to generate this knowledge. As such, our study is framed along the following research question: How does the combination of online and offline channels for inter-organizational knowledge sharing lead to successful searching, filtering and identification of ideas in collaborative innovation (CI)?
In terms of behavioural effects, little research examines how CI participants' interface. Our research provides for a deeper investigation on how such processes are shaped. We use longitudinal data from the German automotive industry where competitors now openly collaborate, facilitated through ICT, in an effort to collectively render competitive momentum for the future. This CI initiative was a nongovernmental affair that brought together major global carmakers, such as Daimler, Porsche and Opel, their suppliers, engineering services and consulting companies, as well as research institutes and private inventors to jointly scout and develop innovative knowledge and ideas.
Over a 5-year period, we analyse how offline channels of interorganizational knowledge sharing interact with online channels in a manner that allows for both trust-enabled knowledge sharing (to overcome competitive tensions) and cognitive flexibility (to prevent crowding) along the different phases of CI. Particularly, we seek to deepen our understanding of the procedures through which parties simultaneously interact across more than one type of channel and how it affects their cognitive capacity to adopt and collaborate in the front-end stage of the innovation funnel, specifically the handover from idea scouting to idea filtering.
Our paper makes a number of important contributions on how the use of collaboration technology changes the way value is created and extracted within and across the boundary of the firm.
Contributing to the management literature on CI (Asplund et al., 2021;Baldwin & von Hippel, 2011;Heil & Bornemann, 2018;Najafi-Tavani et al., 2018), as well as the literature on digital technologies in a creativity context (Jarvenpaa & Välikangas, 2020;Pagani, 2013), we reflect upon the conditions under which digital technology can give rise to new opportunities for CI and idea filtering in particular. First, our findings suggest that while online technologies can facilitate knowledge sharing processes in the context of CI, these processes remain strongly dependent on offline interactions, particularly when competitors are involved. This perspective departs from the notion that different channels of interaction are beneficial at different points of the CI process. Instead, we suggest that channel multiplexity, that is, the extent to which two parties simultaneously interact across more than one type of channel with each other, has a substantial and qualitative different effect in comparison to the effects of either in isolation.
Second, we also incorporate a behavioural dimension into the literature on collaborative R&D and innovation, by highlighting the 'why' mechanisms that drive front-end CI performance. Whereas the 'where' of CI refers to online versus offline, the 'why' denotes the partial coopetitive nature of these interfaces and the cognitive flexibility for those assessing the value of what is being exchanged.
This insight is of relevance as cognition, and its interaction with the environmental structures that facilitate knowledge exchange, can provide fruitful grounds for new, innovative knowledge to develop (Aalbers et al., 2013;Hautala & Jauhiainen, 2014;Peschl & Fundneider, 2014).
The remainder of paper is organised as follows. In the following section, we provide the theoretical support that underpins this study, namely, digital CI and channel multiplexity. It reviews the different mechanisms at play in the CI process. We then provide details on the case study sites and the methods of data gathering and analysis as we contrast the pertinent aspects of the idea search and filtering phases of the CI process from an ideation juror point of view. In conclusion, we discuss our findings, as well as the limitations and future research avenues.

| Digital collaborative innovation
Research on CI and the broader literature on open innovation has linked online channels to enhanced access to external sources of ideation (Dahlander & Wallin, 2006;Füller et al., 2008;Leminen et al., 2015). Recent work contrasting the usefulness of online open ICT platforms for ideation with more traditional mediums report online users generating higher quality ideas than non-users (Parise et al., 2015;Poetz & Schreier, 2012). However, exposure to a vast amount of distant knowledge, largely as a result of the inclusions of new ICT-enabled channels of ideations, can easily result in 'crowding' (Bergendahl & Magnusson, 2015;Piezunka & Dahlander, 2014). Thus, appropriate knowledge filtering mechanisms must ensure that distant knowledge is not too easily discarded by the scouting organization, while preventing the organisation from overloading of irrelevant information, particularly those individuals tasked with the appraisal of multiple ideas (Whelan et al., 2011). Work on idea scouting that crosses organizational boundaries, for instance, shows that complex boundary spanning opportunities require supplementing structural social capital with a strong relational component and suggest that offline interaction, such as personal feedback, complements online idea initiation (Monteiro & Birkinshaw, 2016;Smits et al., 2015). Prior work on knowledge networks suggest that diverse knowledge can be most effectively assessed by actors that are part of open-specialized networks (Aalbers, 2020;Burt, 2004;Gargiulo & Benassi, 2000;Hansen & Haas, 2001). Such networks allow actors to build on similar knowledge domains, accompanied by shared interpretive schema (Ruger et al., 2021). The redundancies that exist between the information received and the receiver's prior information help actors to meaningfully interpret diverse and distant knowledge (Ter Wal et al., 2016).
However, organizations with similar knowledge domains are very likely to be competitors, leading to a paradox: knowledge resources of competitors are both similar and complementary, which increases the potential for learning compared with traditional alliances (Bouncken & Fredrich, 2016;Fernandez & Chiambaretto, 2016). In the context of CI, similar knowledge resources can enable competitors to share their existing knowledge and meaningfully assess diverse new knowledge. At the same time, knowledge-sharing in a competitive CI environment creates tensions between rivals (Bouncken et al., 2018;Naqshbandi & Tabche, 2018;Zobel & Hagedoorn, 2020 ). Thus, in order to make digitally supported CI successful, supporting filtering mechanisms need to be introduced that allow for knowledge sharing among actors from similar knowledge domains. This is where we see the importance of offline interactions which complement online channels for knowledge sharing in the filtering process. Yet little research has focused on the longitudinal dynamics of such processes, that is, if and how online and offline interactions complement each other and which organizational mechanisms facilitate innovative knowledge exchange and evaluation in such a setting (Bouncken et al., 2018;Fernandez & Chiambaretto, 2016).

| Channel multiplexity
While not broadly portrayed in the creativity and innovation literature, the combination of the offline with the online in the corporate ideation process has received some scholarly attention, particularly in the information systems (IS) domain (Ding et al., 2019;Filiposka et al., 2017;Jarvenpaa & Välikangas, 2020;Mesch & Talmud, 2006). This stream of literature has linked channel multiplexity, that is, the degree to which individuals simultaneously interact across more than one type of channel with each other, to the innovative capacity of organisations (Cross et al., 2001;Wang et al., 2020;Zhang & Venkatesh, 2013). Relationships that are maintained through various media tend to create greater obligation (Ho & Levesque, 2005), and more and/or higher quality information tends to be exchanged (Sias & Cahill, 1998). The implications of multiplex relations that include nonhuman technological elements still remain largely underexplored nonetheless (Contractor et al., 2011;Wang et al., 2020). As communication within and between organisations has increasingly become digitised, online interaction adds an additional dimension to the layering of an organisation's functional relations. Network theory, and its notion of relational multiplexity in particular, helps to better understand how and why the combination of online and offline channels of inter-organizational knowledge-sharing leads to successful CI. The theoretical argumentation for this is twofold.
First, characterizing the interplay between online and offline communication networks, prior scholarly work suggests the combination of both to allow for complementing resources. Drawing on a field study at a large telecommunication company, Zhang and Venkatesh (2013) report that the combination of online and offline workplace communication networks fuels complementary resources and enhances individual job performance. In an ideation context, being able to shift back and forth between online and offline environments to probe for contextual information should improve idea attention and idea visibility for evaluators. Extending prior work that outlines the complementary effect of combined online and offline workplace communication to the context of CI, we argue that the resulting ease of accessibility to and control over alternative resources of idea screening positively affects innovative outcome. Online and offline interactions thus are both relevant for CI-related knowledge sharing, particularly in the front end of the innovation trajectory.
While both the IS and management literatures do examine how online and offline channels of interaction influence innovation activities, these investigations tend to focus on each channel separately, rather than considering how channels co-evolve and influence each other (Spagnoletti et al., 2015;Tortoriello et al., 2012).
Second, prior work on the combination of online and offline workplace communication informs us of the positive effect when it comes to employee job performance. After a review of 83 studies on how ICT affects individual employees, Wang et al. (2020) purport that ICT use should be considered as an interaction between intensity and function. ICT use is more likely to influence job demands and decision-making when it is applied to the technical or task aspects of work (e.g., searching for ideas) and influences relational work design when applied to the social aspects of work (e.g., evaluating those ideas). Where online information access allows for the inflow of more ideas, mere exposure to online channels of interaction are likely to result in increased exhaustion via information overload (Wang et al., 2020;Yu et al., 2018). The option to redirect to offline interaction, next to online interaction, deescalates such individual work pressure, subsequently improving creative idea assessment. Based on this, we argue that while online platforms can help to facilitate knowledge sharing processes in the context of CI, those working in a digitally enabled collaborative ideation context will benefit from additional mechanisms based on stronger offline interactions. Research still has to open up the black box of ICT-enabled CI. Existing studies are confined to the spread of highly specific information within limited populations (e.g., Aral & Walker, 2011;Bakshy et al., 2012). Additionally, the role of the open innovation jurors has received limited research attention. As a consequence, our knowledge of how to implement a digitally enabled CI strategy and how collaborative choices by the jurors in such inter-organizational arrangement are made is rather limited.
Drawing from the arguments above, Table 1a provides a conceptual framework that contrasts the pertinent aspects of the idea search and filtering phases of the CI process.
We now consider how this conceptual framework guides the gathering and analysis of data.

| Research setting
To study the underlying dynamics of online-offline interactions, we adopt a single case study design. We gathered data from the first publicly visible CI initiative in the German automotive industry, referred to as the automotive innovation network (AIN). The AIN represents a unique case (Yin, 2009) to us as it offers the opportunity to study how different carmakers collaborate with the joint aim of finding new ideas beyond the boundaries of their firms and industries, supported by the use of ICT.
The AIN was founded in 2009 and represented a loosely coupled, project-based network of over 60 official member firms, although a much larger number of companies became active in AIN projects. The AIN aimed to bring together firms, institutions, entrepreneurs and private inventors interested in developing automotive innovations. To become a member of the network, firms need to pay an annual membership fee. The network's founder and manager loosely initiated and coordinated activities into different working groups. Rather than studying the network as a whole, we focus on the most active working group in terms of regular meetings and number of participants from an early stage of our research. The so-called 'innovation scouting' working group, which we refer to as the CI working group, was tasked with finding new ideas and technologies from outside the boundaries of the automotive industry. The CI working group maintained a healthy emergent culture due to the fact that all individual members participate on a voluntary basis. Both face-to-face and conference call meetings were a regular feature of the CI working group over the study lifespan.
The CI working group was characterized by heavy fluctuation in memberships, but our constant interaction with the group allowed us to identify a core group of five carmakers. Except for one carmaker who joined later in 2010, the circle of carmaker representatives remained stable and consistently active (based on email exchange and meeting participation).

| (1) Email data
We were granted access to the complete email correspondence of one of the core members of the CI working group, an innovation manager (coded as CAR4) from one of the car manufacturers. The data contained over 1500 emails with other members of the CI working group. These emails included the recipients' lists and, in most cases, the whole conversation history as well as attachments such as meeting minutes, strategy papers and presentations. We complemented the email data with data from the web-portal hosted for the open collaborative innovation competition (OCIC). These data were transformed by the organisers of the OCIC into a database and then exported to an excel file that contained detailed information on the idea submitters, a brief description of their idea, technical specifications and proof of product validation. This information was also available to the jury. In addition, the email data contained the initial evaluations of all jury members that formed the basis for discussion within the jury before a final decision on the winners of the competition was made.

| (2) Interviews
We conducted a series of telephone and onsite interviews with key actors of the CI working group at two stages. Table 1b provides an overview. A semi-structured instrument guided the interviews, ensuring that all topics of interest were covered. Depending on the background and position of a particular interviewee in the network, we asked for the evolution of the CI group over time, CI practices, the involvement of particular actors in projects, and perceived outcomes.
Additional interviews with innovation managers of two carmakers were conducted in 2016 to clarify some final questions that evolved during the revision process of this paper. The interviews-17 in totaltypically lasted 60 min and were taped and transcribed afterwards.
Informal talks with experts, as well as with key informants from the CI group, helped us to increase the validity of our data, including a series of such interactions as we observed the various working group meetings reported in Table 1b.

| (3) Field observations
From the beginning of the AIN initiative, we were included in the general mailing list and received invitations for all meetings. Meetings attended were documented by our team and field notes were writtenup within 24 h of the meetings.

| Data analysis
The three sources of data informed our analysis in different ways: the most extensive source was the email data. Based on this set, we first drafted an extensive case report on the formation and development of the working group and the CI community over the period of 5 years.
For each year, a timeline was compiled with all major events that were important to understand the major dynamics around the OCIC. For each major event, we coded the channels of interaction as online (i.e., emails and web) or offline (i.e., telephone conference and face-to-face). This allowed us to observe an evolution of channel multiplexity from mainly online interactions to an increased complementarity with offline interactions. Once we had a better understanding of the history of the OCIC, we used the interview data to provide a more interpretative account of the major events from the perspectives of the core actors. Additionally, our own meeting minutes served to validate some of the interpretations made by our interviewees and provided important background information. We coded the interviews and minutes by grouping phrases, sentences or paragraphs into codes and categories in an inductive fashion. This first phase of coding was followed by axial coding where we generated more abstract codes, deleted and merged codes (Strauss & Corbin, 1990). During this stage, we started to connect our inductive codes to established constructs from our preliminary theoretical framework. In this phase, the initial open codes were translated to specific themes. The axial codes that emerged formed the overarching categories at a higher level of abstraction (Strauss & Corbin, 1990).
The axial coding was conducted by two members of the research team and discussed with the third author until any disagreements would be resolved. To validate our findings on the underlying mechanism in the CI process, the results of our case study were presented to and discussed with the full assembly of members of the CI working group in 2008 and later on also with the core group of five carmakers in 2012. While most of our interpretations were confirmed by our interviewees, the personal discussions also helped to resolve remaining misunderstandings about the case. Finally, we used the additional interviews conducted in 2016 as an opportunity to validate our final interpretations and obtain retrospective reflections from two core participants on why the integration of ideas obtained from the OCIC proved to be so difficult for the carmakers.
The following section provides a more detailed case description of the complementary roles of offline and online channels along the CI process. Unpacking the searching phase in the CI process, we review the role of both online and offline channels at the initial stage of the OCIC, examining if and when online and offline channels become more interlinked and result in channel multiplexity as grounds for idea search. We provide exemplar quotes to support our interpretation of events and the underlying mechanism that drives multiplex use of channels in collaborative innovation, particularly related to new idea search. We describe the evolution of the CI idea search trajectory  Clusters 1, 2, and 3 will remain the same, a new cluster 4 is to be set-up that is explicitly addressing research institutes and universities, and cluster 5 on radical innovation will be discarded and postponed to 2010.   The jury agreed on using general evaluation criteria (consumer value, breadth of applicability in the vehicle, maturity of the innovation in its current field of application, expected product life duration of the innovation, sustainability and customer acceptance) instead of more specific ones to allow for more openness for radical and unconventional ideas. All of the 20 jury members were requested to evaluate each of the 150 submissions independently by filling in a score based on predefined evaluation criteria. At first, jury members sent around their scoring results by email in excel files, which did not turn out to be an effective way to perform the evaluations: Our use of Excel-files was rather mechanical. We were too polite to question each other's evaluations. If someone thought an idea was great and I did not, then this could be due to different reasons that were unknown to me. Maybe someone liked the idea because he or she thought it fits well in their product portfolio. Thus, we did not really try to convince that person that this idea might not be good but we simply put all our evaluations in the list and compiled an overall ranking. (interview with CAR4).
As a result, the overall evaluation results of the jury were any-

| Refinement (Years 2-3)-Growing offline interactions to complement online channels
After the experience of the first year of the COIC, three theme clus- The process of pooling firm-specific expertise helped to make sense of ideas that were hitherto foreign to the automotive con- what was often referred to as 'freeriding' behaviour of members who just wanted to stay informed without making their own contributions.
Frustrated by this experience, the innovation manager from carmaker 3 started an email exchange with his colleagues from CAR1, 2, 4, and 5 under the heading 'Entre Nous' where he suggested an informal meeting in the evening before the next official group meeting to discuss 'the future of the innovation scouting group'. After this meeting, the carmaker representatives decided to stay involved in the official group and continue to carry out their organizing activities for the competition, but the 'Entre Nous' meeting became the start of a series of informal meetings that institutionalized as the 'Innovation Roundtable'. In these additional meetings, the five carmakers discussed any issues regarding the OCIC with each other first before raising them in the official meeting with the CI scouting group. These informal meetings helped the five carmaker representatives to establish a leading position inside the network.
The regular physical meetings also helped to build the necessary trust between the participants and allowed them to openly share their own views on a technology that was up for evaluation and actively engage with the views of others to overcome competitive friction: We Intensifying interactions in terms of both frequency and channel multiplexity-email, telephone conferences and physical meetings became fully intertwined at this stage-supported the process of increasing cognitive flexibility: The challenge was that we were dealing with ideas in a very premature stage, that we could not fully grasp and where you can easily end up with different opinions regarding their value. I always found it extremely enlightening if one of us would defend an idea and say: 'I understand this differently, you can actually use this for this, I could imagine that this will influence that ….' And then suddenly I realized, ok, I have not really thought of this before. So that was really an interesting dynamic going on. (interview with CAR 2) Once the five carmaker representatives exchanged their views on the evaluation of the submissions and felt more confident in the quality of their assessments, they would discuss their assessment in a separate meeting with the remaining members of the jury: The meeting with all jury members in one room were really useful as everyone had a different approach.
From a firm perspective, you always run into danger to think too one-dimensionally because of branding issues. In the jury, you had different suppliers, carmakers, all together who were all thinking from different angles. Thus, the chances of a more differentiated perspective became much higher. (interview with

CAR4)
This process helped to increase the number of new ideas that were filtered from the OCIC: There was an increase in quality of the competition from year 1 to 2, and from year 2 to 3. We did find a lot of interesting ideas that we did not expect to find.
This is not only because the competition attracted more submissions from different countries but also because the quality of our assessment improved over the years. (interview with CAR3) Thus, the multiplexity in interaction not only helped to improve sense-making of diverse ideas but also to increase cognitive flexibility among those that were evaluating the ideas. This led to successful filtering and identification of the most promising ideas from beyond the boundaries of the automotive industry. This is also reflected in a higher share of winners from outside the industry (i.e., private inventors and universities), as opposed to firms from the automotive industry that were dominating the competition in the first 2 years.
In graphical summary of the collaborative ideation trajectory outlined above, Figure 2 summarizes the CI idea generation and filtering process over time in relation to the preferred interaction channels, illustrating the role of multiplexity in the CI process as it matured over time.
Although some fluctuations took place over time, the core group has remained stable until today. These ties were largely rooted, however not restricted, to the personal level. In one case, one representative changed his position within the firm, yet his successor was able to seamlessly take over the role and effectively participate in the further group meetings.
The ongoing offline discussion among the OEMs also soon made clear, that there was a general consensus to stop the engagement in the wider innovation scouting group, but to continue regular meetings in the smaller circle. The group decided to rename itself the 'Innovation Roundtable' to stress its independence from the former context. insights, particularly during the filtering phase of the CI process. Idea appraisal-illustrated throughout the filtering phase in our study-is a process that benefits from the combined use of online and offline channels of interaction in particular. As ideas progressed through the filtering phase, a gradual move from online channels (email) to offline channels (telephone, face-to-face meetings) became apparent. Those evaluating the ideas were better able to overcome the initial cognitive rigidities that characterized their evaluation processes at the search stage, gaining a new perspective on ideas that were generated online (e.g., Perry-Smith & Mannucci, 2017).
Based on these insights, our contribution is twofold. First, in line with recent work on the organizational antecedents behind coopetitive collaboration for innovation (cf. Bouncken et al., 2018;Fernandez et al., 2018;Mention, 2011), we considered the knowledge sharing preferences amongst those involved. In reply to recent appeals to further explore the organizational and collaborative mechanisms facilitating innovative knowledge exchange and knowledge evaluation in coopetitive settings (Bouncken et al., 2018;Fernandez et al., 2018) our findings suggest channel multiplexity-the extent to which two parties simultaneously interact across more than one type of channel with each other-has a substantially different effect on CI evaluation outcomes in comparison with the effects of either in isolation. In pursuit of a better understanding of how online technologies can facilitate knowledge sharing processes in the context of CI, we identify ICTenabled knowledge exchange as a complementing factor, not a substituting one, to the offline process of inter-firm idea appraisal. We show that complementary offline interactions are necessary to overcome coopetition amongst those involved in the collective appraisal of CI rendered ideas. While physical and virtual environments may provide for a fruitful basis for collaborative R&D (Leminen & Westerlund, 2019), innovation research commonly focuses either on ICT-enabled interaction or traditional offline interactions, but has largely ignored their combined effect. We believe our study is the first to apply the multiplexity lens to the online/offline interaction processes in an innovation context (Aalbers et al., 2014;Phelps et al., 2012). Thus, we add to both the creativity and CI literatures as we provide for a more finegrained understanding of how and when online communities facilitate to collaborative open innovation initiatives.
Second, in an attempt to bring behavioural explanations to our understanding of the CI process, our case study illustrates how cognitive flexibility can be facilitated through channel multiplexity.
Cognitive flexibility, defined as the ability to shift schemas and cognitive categories, has been identified as a prime mechanism to build trust-enabled knowledge sharing (Amabile, 1983;Mednick, 1962).
Idea generation depends upon divergent thinking and novel associations (Berg, 2016;Perry-Smith & Mannucci, 2017 A deep understanding of individual-level network dynamics is critical for implementing strategy and organizational change (Hung, 2002;Lynch & Mors, 2019;Vogel, 2005). Yet strategizing in a digital world frequently commences without much concern for the offline. Simultaneous and consistent offline interaction enabled the closer circle of jurors to improve the filtering of truly valuable of ideas, suggesting the effectiveness of a technology platform to rests on more than just the technical specifications (cf. Denyer et al., 2011). Such close offline interactions should be central to any digital strategy initiatives.

| LIMITATIONS AND FUTURE RESEARCH AVENUES
Our study is not without some limitations, which also presents opportunities for future research to advance our work. First, participants in our study used email almost exclusively to collaborate and share knowledge. As different communication platforms offer differing affordances, it may not be possible to generalize our findings to networks communicating through social media platforms such as wikis, social networking sites, blogs, forums or instant messaging. For example, Leonardi and Treem (2012) theorize four social affordances-visibility, persistence, editability and association-represented by social media. They also note that email only enables some of these affordances. Email has high editability (users can carefully craft messages prior to sending), persistence (users who can save, store and search through their own messages), but does not easily enable association (creating ties with other users) or visibility (viewing the communications of others). Thus, future studies should use the affordances lens to better understand how different online communication mechanisms influence knowledge exploration and exploitation in organizations.
Second, our study was strictly in the search and filtering process of new and distant ideas and less on firm-internal integration success of these ideas. The integration of innovation in a mass-produced product like automotive is not trivial and requires extensive application engineering and testing to comply with strict passenger safety standards. It remained unclear until the end of our study as to what extent ideas scouted through the online competition were actually integrated in new models of the member carmakers. We echo Dong and Wu (2015) in that firms need to develop implementation capabilities to filter and exploit the voluminous ideas afforded by online innovation platforms. In the current study, we did not focus on objective measures of 'success' of the CI initiative, but instead relied on subjective statements of interviewees regarding their personal satisfaction with the outcome of the competition and the quality of the evaluation process. Nevertheless, it would be desirable if future studies would make use of existing, more refined measures to assess the quality of ideas submitted to online idea competitions (e.g., Blohm et al., 2011).
As a third and final limitation our study is based on a single case of an CI network in the automotive industry. Innovation processes for automotive are structured in quite a unique way. We argue, however, that the problems we address in our study-such as the necessity to deal with the difficulties of filtering distant ideas (Piezunka & Dahlander, 2014)-are universal to all CI initiatives. Nevertheless, we highlight the need to extent the study of CI to more industry contexts to further increase the external validity of the concept.