The digital grapevine: team dynamics, task prioritisation and social interaction in intra-organisational digital platforms

In numerous manufacturing and service settings, teams are frequently formed for brief collaborations before being disbanded and subsequently reconfigured (Argote et al., 2021; Kim et al., 2022). The concept of team familiarity, which denotes the degree to which members have previously collaborated and the extent of their mutual knowledge, is increasingly recognised as a critical factor that influences team performance (Reagans et al., 2005; Luciano et al., 2018). This familiarity facilitates the integration and application of diverse information within team actions and decision-making processes, thereby enhancing performance outcomes (Guimerà et al., 2005). However, the influence of team familiarity is not a straightforward, universally positive relation.

A considerable body of research has explored the ways in which task-related specialisation or exposure to variety impacts team performance (Madiedo et al., 2020; Akşin et al., 2021; Ching et al., 2021). Recently, some studies have broken new conceptual ground by examining the temporal aspects of the task-related experience, e.g. the same-day same-task experience vs the all-time same-task experience (Staats and Gino, 2012) or the experience of frequently-vs infrequently-performed tasks (Luciano et al., 2018). However, considering the existing body of literature, further task-specific characteristics, such as task prioritisation and its impact on the time to process a task, remain under-explored. Furthermore, most studies are concerned with the impact of team familiarity on shorter process times (Avgerinos and Gokpinar, 2017), shorter lengths of hospital stay (Patterson et al., 2015) and lower costs (Agha et al., 2019). However, the impact of team familiarity on the time to accept a task, e.g. on a digital platform with a broadcast mechanism, has received little attention in the literature.

In addition to the fact that task-specific characteristics, such as task prioritisation, have been under-explored, there is also a lack of consideration for social interactions and the traditional human factor in the context of the use of digital platforms that employ a broadcast mechanism. For example, the main advantages of using digital platforms, such as the reduction of information asymmetries resulting in increased information transparency, have already been explored (Culotta et al., 2024). Also, various theories related to digital platforms, including theories within the context of the “theory of the firm”, have been extensively studied (Sedera et al., 2016; Beske, 2012; Teece et al., 1997; Culotta et al., 2024). In this research, an intra-organisational digital platform in a logistical context was examined. The unique contribution of this study lies in addressing an empirical gap in the research on how digital platforms influence operational performance through intra-organisational team behaviours and interactions. While prior research has extensively explored the role of digital platforms and the significance of social theories, there is a lack of empirical evidence that connects these two domains in a concrete operational context. This gap is particularly evident in the absence of data or evidence needed to fully understand and explain the phenomenon of digital platforms mediating team dynamics and operational processes. Empirical gaps arise when essential data or evidence is missing to provide a complete understanding or explanation of phenomenon (Miles, 2014). In this case, the study bridges this gap by examining the interplay between human factors and digital platforms, thereby providing novel insights into operational impacts. By addressing this gap, the research not only enhances theoretical understanding but also contributes practical insights for optimising the use of digital platforms in complex intra-organisational environments.

In this study, we investigated multi-member teams operating at the same hierarchical level utilising a digital platform for task allocation and a selection based on a broadcast mechanism. Specifically, we focused on teams of drivers working simultaneously at the same yard. In the workplace, tasks for drivers are proposed for all team members through an intra-organisational digital platform. Team members can select which task to accept, which aligns with the principles of a broadcast mechanism; that is, the principles of a broadcast mechanism involve allowing drivers to select tasks from an available pool. The driver who chooses a task first is assigned that task (Xu et al., 2020). Additionally, the various tasks are prioritised. After a driver selects a task through an application running on a personal digital assistant (task-acceptance time), the execution of the driver assignment commences (task-processing time). The utilisation of a digital platform is intended to optimise communication among the involved parties while reducing the use of informal communication. The warehouse and yard employees, along with the yard drivers, were considered the involved parties in this study. The study also proceeded under the assumption that optimised and reduced informal communication between the involved parties can lead to standardised and more efficient processes (Kapkaeva et al., 2021).

In our study, the task-processing time essentially involved the transportation of semitrailers within the yard from one location to another. We thus examined the effects of team familiarity and task prioritisation on both task-acceptance and task-processing times. Therefore, our research question was “How and to what extent do team familiarity and task prioritisation impact the time to accept a task on a digital platform and the time to process a task in the yard in the presence of a broadcast mechanism?”

To answer the research question, we cooperated with a large German brick-and-mortar grocery retailer and obtained an archived dataset containing 146,446 routes performed by 878 unique yard driver teams. The dataset covered the time frame between January 2022 and April 2023. We used a linear mixed-effects model with two random intercepts for each team and each driver, thereby capturing team and driver differences as well as between-team and between-driver heterogeneity effects that were not accounted for when fixed effects were included.

Our study revealed the following four key findings. (1) The greater the team familiarity within a team, the shorter the task-processing time. Drivers in the yard who had had more experience working together in the past were likely to have more shared experiences that enabled them to, for example, understand each other’s strengths, weaknesses, expertise, backgrounds, personalities and habits. (2) The greater the same-month high-priority task specialisation within a team, the shorter the task-processing time. Frequent execution of the same task entailed task-related rhythms and processes, which helped to accelerate task execution. We expected that the logic of task-related specialisation and the resulting rhythms would also apply to high-priority task specialisation. (3) The greater the team familiarity within a team, the longer the task-acceptance time. With a positivist perspective on team members in a team, we predicted that the team members would not accept certain tasks because they would know that other team members were better suited for these tasks or more motivated to perform them. Taking a non-positivist perspective on members in a team, we argued that a social loafing or free-rider effect might appear on the team level. (4) The greater the same-day high-priority task specialisation within a team, the longer the task-acceptance time. Herein, we argue that a habitual effect might occur on the same-day and same-month levels. All our key findings directly address the identified empirical gap of this study. This gap pertained to the lack of evidence on how the human factor, influence the task-processing and task-acceptance time mediated by digital platforms. By providing the necessary empirical data to understand the interplay, our findings contribute to a deeper understanding of the relationship between digital platforms and intra-organisational team dynamics.

The remainder of this paper is structured as follows. In Section 2, we explain intra-organisational digital platforms and present our hypotheses in Section 3. In Section 4, we introduce the empirical setting of yard management and explain the variables as well as the assumptions of our econometric model. Our empirical results are presented in Section 5 and discussed in Section 6, while Section 7 offers our conclusions.

Digital platforms facilitate transactions and information exchange among participants in various settings, including business-to-business (B2B), business-to-consumer (B2C) and intra-organisational interactions. Such platforms mitigate the information asymmetries that can occur when one participant has information that other participants need, thereby creating value through information exchange (de Reuver et al., 2018). Notably, digital platforms can be two-sided or multi-sided and connect multiple participants, such as buyers, suppliers and complementary service providers, at the same time (Trabucchi and Buganza, 2020). The focus of this study is on intra-organisational platforms, particularly a two-sided one that connects warehouse and yard employees. The primary advantage of the platform in this context is the reduction of information asymmetries, thus ensuring that all employees have identical information about trailer locations, which reduces the idle time and cycle time of the trailers, creates more transparency about pending tasks and enhances productivity in operational warehouse and yard management processes.

Digital platforms for intra-organisational transactions are especially relevant in logistics and operations management because they are capable of connecting participants across different levels, from organisational units in inter-organisational supply chains to employees in a single warehouse. According to Gawer (2021) and Cusumano (2019), digital platforms can be categorised into transaction platforms, innovation platforms and hybrid platforms that encompass elements of both types. Transaction platforms facilitate the exchange of goods, services and information between different stakeholder. In a supply chain context, these platforms include Internet of Things (IoT) and visibility platforms that share supply chain data or depict digital twins, who are virtual representations of physical objects (Abideen et al., 2021; Kalaiarasan et al., 2022), supply chain finance platforms that enable automated payments (Gomber et al., 2017) and freight exchange platforms that connect forwarders and carriers (Miller et al., 2020; Herold and Fahimnia, 2023). Innovation platforms, on the other hand, create value by providing technological building blocks to enable a joint innovation process among different actors. The focus of this study is on transaction platforms due to their role in depicting the exchange or movement of trailers in the yard of a grocery retailer and providing real-time data to all participants in a single warehouse.

Transaction platforms are widely used in both industrial and logistics settings to reduce information asymmetries, increase transparency into pending tasks and available resources and, thereby, reduce process complexity. In the industrial context, the challenges of digital platforms include their high complexity, lack of standardisation, mistrust between the various stakeholders on the platform, legal provisions such as traceability of objects and a lack of digital compatibility and competencies among supply chain partners (Hein et al., 2019; Anderson et al., 2014; Cichosz et al., 2020; Winkelhaus and Grosse, 2020). For logistics settings, Helwing et al. (2023) examined the direct impact of digital platforms on the inclusion and exclusion of actors. Specifically, inter-modal transport and logistics systems achieved high performance due to the introduction of digital platforms (Dmitriev and Plastunyak, 2020). Our focus, however, was on the use of digital platforms in the logistics context, particularly in yard management, where the complexity related to spatiotemporal dynamics is a key concern. Complexity, defined by the quantity and variety of tasks, their relationships and the changeability of these elements and relationships, underscores the necessity of digital platforms in managing yard operations effectively (Meyer, 2007).

Platforms involve various players. Our study’s context involved transaction platforms within the intra-organisational domain, specifically a two-sided platform that connected warehouse and yard employees. Fundamentally, there are four main players in a platform system. The first player is the owner, who serves as the controller of the platform. The second player is the provider, who supplies the interface for the platform. The third player is assigned as the creator of the platform, known as the producer. Finally, the fourth player is the consumer, who acts as the buyer or user (Alstyne et al., 2016). When applying these four players to a transaction platform within the intra-organisational context of yard management in a single warehouse, the players can be defined as follows: the owner represents the company itself or an external platform company that leases the platform. The provider refers to mobile devices (such as smartphones) that facilitate platform interactions. The producer corresponds to the application responsible for managing and orchestrating platform activities. The consumer encompasses various stakeholders, including warehouse and yard employees as well as drivers who utilise the platform for their tasks.

The success factors for the use of digital platforms include the technology and the understanding of the users as to its use as well as having a suitable vision for new business models. That is, platform managers need a clear vision for solving problems for consumers, firms and ecosystem members. Moreover, success in platform business requires an integrated knowledge of both, technology and business (Evans and Gawer, 2016), and companies need to articulate mutually compatible business models for themselves and their ecosystem partners. The combination of the technological and business skills of individual employees is also considered a success factor for the use of digital platforms (Evans and Gawer, 2016). The development of appropriate technology and an understanding of this technology, along with employees to bridge the gap between technological and business skills, can be seen as success factors for transaction platforms within the intra-organisational context of yard management (Evans and Gawer, 2016).

When examining digital platforms, previous research drew on the meta-theory of “theory of the firm” with four theories that are frequently applied to logistics and operations management contexts: the resource-based view, transaction cost economics, internationalisation theory and dynamic capabilities (Culotta et al., 2024). The resource-based view is commonly applied to explain aspects of platform performance or the role of network effects. This view characterises specific resources and capabilities of digital platforms, such as technology, operational experience, training and market-specific resources (Wernerfelt, 1984; Barney, 1991; Lou et al., 2021; Sedera et al., 2016). In logistics and operations management, the resource-based view can support strategic decisions, such as supplier selection, supply chain collaboration and purchasing strategies (Halldórsson et al., 2015; Hitt, 2011). Transaction cost economics offers another perspective by organising labour via a digital infrastructure. That is, digital platforms enable firms to efficiently coordinate resources, disposal rights and labour without relying on traditional firm structures. Rather than outsourcing processes or purchasing external services and goods, firms consider how to organise activities through digital platforms. These platforms reduce firm boundaries, reshape supply chains and enable new business models. Managers must define their firm’s essence, delineate its boundaries and design supply chains accordingly. Specialisation may increase, while generic tasks find efficient organisation via digital platforms (Culotta et al., 2024; Oliver, 2007, 2009). Internalisation theory addresses the activities of multinational firms (Hymer, 1976; Narula et al., 2019). It assists in the understanding of global supply chain strategies, such as entry mode decisions and internal organisational design. Digital platforms facilitate firms’ entry into foreign markets, thereby making platform design and local network effects crucial for supply chain managers to optimise platform activities in these markets. The dynamic capabilities approach focuses on processes, competencies and path dependencies within a company (Teece et al., 1997; Eisenhardt and Martin, 2000) and explains how firms sustain competitive advantage in dynamic environments and enhance supply chain performance (Beske, 2012). Digital platforms, by providing transparency and real-time information, can optimise supply chain performance and network effects, thus benefiting the entire ecosystem (Culotta et al., 2024).

Our review revealed that social interaction and operational-level processes have received comparatively little attention in existing research. Yet, digital platforms have emerged as pivotal structures for orchestrating a diverse array of human corporate operations. These platforms not only facilitate economic transactions but also significantly influence social dynamics. Fundamentally, platforms serve as interfaces that facilitate interactions between different actors (Boudreau and Hagiu, 2009).

Digital platforms aim to connect various groups of providers and users (Broekhuizen et al., 2021). Furthermore, a characteristic of digital platforms is to shape social interactions (Braune and Dana, 2021). On the one hand, digital platforms aim to establish stability and homogeneity through standard components, while on the other hand, variability and heterogeneity are important to meet the requirements of stakeholders. The tension between collective and individual and control and autonomy as well as standard and variety therefore require further targeted examination (Wareham et al., 2014). The paradox theory provides a useful lens for understanding these tensions emphasising the coexistence of contradictory yet interdependent elements within organisational systems. Rather than resolving these tensions through trade-offs, the theory advocates for their simultaneous embrace and management, as both stability and variability are essential for an effective digital platform (Smith and Lewis, 2011). In the context of social interaction, this means balancing the structural control provided by platforms with the flexibility needed for meaningful and adaptive collaboration. By examining these paradoxes, this study sheds light on how digital platforms mediate and shape social interactions, ultimately influencing organisational performance and quality. Within this framework, the combination of human and machine learning can be viewed in terms of successful collaboration. While machine learning can reduce an organisations demand for human exploitative learning, adjustments by humans are also beneficial under certain conditions (Sturm et al., 2021). In addition, the mediating roles of design control, stimulus-response variety and interaction rules should be considered. This influence is especially noteworthy when product platform digitalisation leads to phase transitions (Sandebrg et al., 2020).

The primary advantage of digital platforms lies in their ability to reduce information asymmetries by centralising data flows and ensuring greater transparency in task execution (Culotta et al., 2024). However, digital platforms have also been criticised for their control mechanism, which can lead to a lack of trust among users, particularly concerning the monitoring and evaluation of work quality (Möhlmann et al., 2021). To address this paradox, organisations must invest in digital infrastructure that fosters trust among users and ensures a balance between control and autonomy (Elkaffas, 2023). An equally critical, yet often overlooked, factor influencing the success of digital platforms is the social interaction that occurs with them. Effective communication and collaboration among users are key drivers for quality in platform-mediated workflows (Thies et al., 2016). Despite its significance, the role of social interaction as a determination of digital platform performance has received little attention in prior research. This empirical gap underscores the need for the investigation into how team behaviours and interactions mediated by digital platforms influence the operational performance.

Most intra-organisational digital platforms either adopt a broadcast mechanism, in which workers have the autonomy to select tasks, or a dispatch mechanism, in which the platform assigns tasks to workers. If a platform employs a dispatch mechanism, tasks are assigned to workers, e.g. directly by their managers (Dai et al., 2022). In this context, workers have no opportunity to influence which tasks are allocated to them. Tasks can be either randomly allocated or assigned by using algorithms including a set of rules designed for this specific purpose (Helling, 2021). For instance, the allocation of ambulance crews to an incident may be based on a dispatch mechanism. In this scenario, allocation is carried out by a central dispatch centre, which takes into consideration the following three factors: (1) distance from the incident, (2) priority of the incident and (3) availability of the crew (Bavafa and Jónasson, 2021).

In contrast, platforms that employ a broadcast mechanism afford workers full autonomy and flexibility to choose which tasks they wish to conduct. Workers select a task from a variety of available assignments. In this context, the prior experience learning effect is significant for improving employee performance (Chu et al., 2018). Limited guidance and restrictions imposed by the company may potentially constrain the free selection of tasks (Dai et al., 2022). When a platform utilises a mobile application, task allocation based on a broadcast mechanism may take the following steps: (1) workers select a task from the available assignments and accept it by clicking the “accept” button in the application, (2) workers complete the selected task and (3) upon completing the task, workers confirm the accomplishment in the application (Dai et al., 2022).

An example of a platform with a broadcast mechanism is food delivery services with various retailers (e.g. Uber Eats). Occasional workers serving as delivery drivers might use a broadcast mechanism on a mobile application to select tasks for delivery. Initially, the workers can decide, within the application, whether to start or stop working. Once the worker begins performing, they can view currently available tasks and additional information, such as weight or pickup and delivery locations. The worker who first accepts the task by clicking within the application is assigned that task. The tasks are displayed to all workers within a specific distance from the pickup location (Xu et al., 2020). This example underlines that within a broadcast mechanism, several individuals can view the tasks and choose one out of the various available assignments. The worker who first accepts the task within the platform is assigned the task.

In the context of this study, a digital platform that employ a broadcast mechanism is examined. While prior literature on digital platforms employing broadcast mechanisms has mostly focused on the individual level Urrea and Yoo (2023), our study was dedicated to the team level. We were especially interested in intra-organisational digital platforms in the context of yard management tasks. Our special contribution is that we decomposed the total time to perform a task in terms of task-acceptance time and task-performance time. This allowed us to study the impact on team familiarity in connection to social behaviour (e.g. social loafing in task acceptance) and individual behaviour (e.g. learning effects in task performance). We continue by describing our hypotheses.

Team familiarity is defined as the extent to which team members have been working with one another and the level of knowledge team members have about each other (Muskat et al., 2022). It includes each team member’s amount of time of shared work experience as well as the team relationships, cooperation and collaboration between members and team communication (Narayanan et al., 2009). Therefore, team familiarity differs from an individually perceived proximity when working together, as working next to each other might not necessarily be a joint experience or include shared knowledge and interactions.

A well-documented empirical result from prior literature is that team familiarity is positively associated with performance (Argote et al., 2021). Existing research has shown team familiarity to be associated with shorter processing time (Avgerinos and Gokpinar, 2017), shorter lengths of hospital stay (Patterson et al., 2015) and lower costs (Agha et al., 2019). High levels of team familiarity enhance team members’ knowledge about each other’s strengths, weaknesses, personalities and habits. The general concept of team members’ knowledge considering knowledge transfer and social learning is grounded in the transactive memory systems theory (Wegner, 1987), which posits that team members collect, encode, store and retrieve knowledge which leads to positive team outcomes (Brandon and Hollingshead, 2004; Kanawattanachai and Yoo, 2007).

Consistent with this extensive prior literature, we expected that increasing team familiarity would reduce task-processing time and, therefore, increase performance. Drivers in the yard who have had more experience working together in the past are likely to have more shared experiences that would enable them to (1) feel psychologically safe and better assess mutual behaviour (Bradley et al., 2012), (2) understand each other’s strengths, weaknesses, expertise, backgrounds, personalities and habits (Staats, 2012) and (3) learn from each other and acquire explicit knowledge (Taylor and Fearne, 2006). One operational example related to our context is a truck driver who might adapt the route according to experience with the mutual behaviour of colleagues. While this might be associated with waiting when, e.g. giving colleagues priority in intersections, we would expect that the overall effect would still be a reduction in task-processing time. Therefore, we hypothesised:

Task specialisation within a team spawns a positive impact on task performance and reduces task-processing time (Espinosa et al., 2007). The literature discusses this relationship under the concept of a learning curve or an experience curve (Wright, 1936; Dutton and Thomas, 1984). There are several reasons why teams’ and individuals’ task-processing times benefit from specialisation. First, frequent task execution of the same task entrains task-related rhythms and processes, which help to accelerate task execution. Second, frequent execution of the same task enables the development of a streamlined workflow and enhances expertise in that task (Argote et al., 2021). One explanatory approach is the application of learning from the past, which can lead to further optimisation in the current execution of the task with the aim of increasing productivity. One study that provided empirical evidence for learning effects in repeatedly performed tasks was presented by Batt and Gallino (2019), who found significant performance improvement when the specialisation within an order-picking task increased.

While the positive effect of task specialisation on performance has been well replicated in the empirical literature, the effect of task prioritisation specialisation remains under-explored (Cui et al., 2020; Narayanan et al., 2009; Powell, 2000). Task prioritisation is implemented to nudge workers and predictably modify behaviour without restricting any options or substantially altering their economic incentives. Nudge theory posits that choice architecture can be employed to shape human actions, and researchers across various domains have conducted experiments to understand decision-making processes in the context of nudging (Hummel and Maedche, 2019; Peer et al., 2020; Balconi et al., 2023; Kalnikaitė et al., 2013). With the advancement of technology, the concept of digital nudging has emerged, and various definitions have been proposed based on existing nudges in the physical world (Schneider et al., 2018). Digital nudges may involve design elements within user interfaces to steer individual decision-making processes in digital choice environments (Boskovic-Pavkovic et al., 2019; Weinmann et al., 2016). An example of a digital nudge definition can be found in Lembcke et al. (2019), who defined nudges as an intervention element that is goal-oriented and intended to influence individuals’ judgments, choices or behaviours in digital or blended environments.

We expected that the logic of task-related specialisation and the resulting rhythms would also apply to high-priority task specialisation. That is, we expected that drivers in the yard who had repeatedly executed the tasks with the same high priority would be likely to expand their task-related knowledge with the characteristics of task priorities. This might include (1) route optimisation and (2) docking and undocking processes at the gate in collaboration with the warehouse employees. We hypothesised that for same-month high-priority task specialisation:

The literature concerning team familiarity provides scant empirical evidence for negative effects on team performance. One potential counterproductive effect of team familiarity might exist in innovative and creative contexts. That is, with increasing team familiarity, team members might converge to group thinking (Janis, 1972; Hart, 1991). This would potentially reduce the willingness to discuss and challenge existing ideas. Teams with high levels of team familiarity might generate less innovative ideas as team members’ knowledge tends to homogenise and, in contrast, more diverse ideas and innovative perspectives may be likely to evolve when creative teams have lower team familiarity (Guimerà et al., 2005). While our context does not include innovative or creative tasks, the aspect of group thinking might apply to our setting.

We proposed that group thinking in our context of the broadcasting mechanism applied to digital platforms might have a negative effect on the time to accept a task by increasing this time. In other words, we expected that drivers in the yard who had more experience working together would be likely to have more shared experiences that would enable them to understand each other’s strengths, weaknesses, expertise, backgrounds, personalities and habits. With a positivist perspective on members in a team, we predicted that the members would not accept certain tasks because they would know that other team members would be better suited for these tasks or more motivated to perform them. This mechanism is grounded in self-determination theory, which proposes that individuals are motivated to engage in activities that they find intrinsically rewarding. We predicted that knowing about team members’ habits would also include knowledge about task performance and about what motivates team members. Leaving tasks for other team members might, therefore, increase the task-acceptance time.

Taking a non-positivist perspective on members in a team, one could argue that there is a tendency for individuals to expend less effort when working in a group than when working alone. This effect might increase with increasing team familiarity. In a team setting, some individuals might assume that others will pick up the slack, especially for tasks they personally find less appealing or challenging. This is still grounded in self-determination theory for the individual level (Deci, 1985; Ryan and Deci, 2000), but a social loafing or free-rider effect might appear on the team level. Social loafing refers to the phenomenon wherein individuals exert less effort in a group task than when working alone (Harkins, 1987; Latané et al., 1979; George, 1992). In addition, the free-rider effect can be defined as the absence of an individual’s contribution towards completing a group task, even though they would benefit from the group task (Andreoni, 1988; Marwell and Ames, 1981). Given that a positivist and non-positivist view on team behaviour yields the same relationship of team familiarity and task-acceptance time, we hypothesised:

In contrast to the previous hypothesis, we predicted that greater high-priority task specialisation would lead to higher task-acceptance time. Nudges signalling high priority are initially applied to reduce the task-acceptance time (Thaler and Sunstein, 2008; Johnson et al., 2012). However, when high priority tasks are increasingly accepted on the platform, users might become habituated to the nudges. This habitual effect is grounded in the stimulus response theory, which proposes that behaviour is a function of environmental stimuli and rewarding (Skinner, 1953). With specialisation in identical nudges, e.g. high-priority task specialisation, individuals might learn that the stimulus is neither rewarding nor relevant (Toates, 1997; Sugden, 2009; Gravert and Olsson Collentine, 2021).

Consistent with stimulus response theory, we predicted that greater high-priority task specialisation would lead to higher task-acceptance time. That is, we expected that drivers in the yard who had repeatedly carried out high-priority tasks within a day may have become accustomed to high-priority tasks. Consequently, the incentive that was supposed to be triggered by high-priority tasks would diminish. Therefore, we hypothesised for same-day high-priority tasks:

The matrix in Table 1 illustrates the relationship between the dependent variables, task-processing and task-acceptance times as well as the two independent variables of team familiarity and task specialisation. Additionally, the proposed directions of the hypothesised effects are represented by a plus or minus symbol.

We tested our hypotheses using a dataset of all yard transports at a location of a large German brick-and-mortar grocery retailing company over the course of one year and four months. Our dataset comprised a yard with two rectangular warehouses, totalling 211 gates. The yard spans a total area of 200,000 square metres. The shifts are distributed over the period from Sunday 8:00 PM to Saturday 3:00 PM.

Yard management is considered an excellent setting for our research for the following reasons. First, yard management links transportation with intra-logistics at warehouse sites. Yard operations play a crucial role in the handling of semitrailers, vehicles and swap bodies outside the warehouse. Their influence extends significantly to the efficiency of internal warehouse operations. Achieving optimal performance necessitates minimising the time required for changing transportation units during the unloading and loading processes (Clausen and Goedicke, 2012). Second, both researchers and practitioners have paid limited attention to yard management. Particularly in practice, the focus is on optimisation of transports or intra-logistics, which necessitates investments in warehouse or fleet management systems (De Muynck and Tunstall, 2019). Finally, we anticipated a difference in shunting operations, utilising a digital platform with a broadcast mechanism, between the task-acceptance and task-processing times considering team familiarity and task prioritisation, which required investigation.

There are various ways to organise the yard transports and the associated shunting process. One option is for vehicle drivers to bring semitrailers directly to the gates, with the further transport of these semitrailers being carried out by shunting vehicles. Before leaving the yard, the vehicle drivers pick up a semitrailer either directly from the gate or from another designated area. Another option is for vehicle drivers to set down semitrailers in designated areas, and these are only transported to the gate by shunting vehicles. Before leaving the yard, the vehicle drivers pick up a semitrailer only from a designated area, where it has been placed by a shunting vehicle (Clausen and Goedicke, 2012). The yard processes in our research were grounded in the second option. Therefore, incoming drivers place semitrailers in a designated area. For further processing, these semitrailers are picked up by internal drivers and may be driven to unloading gates, where they are set down. If the semitrailers are prepared for further transport, internal drivers place them in designated areas. Drivers then retrieve them for onward transportation. Consequently, any movement of semitrailers not related to incoming or outgoing transports is carried out by internal drivers. In addition, there is a one-way street regulation within the yard that explicitly exempts internal drivers. In our context, we exclusively considered the internal drivers, hereinafter referred to simply as “driver” and also known by the term “shunter”.

Before the actual shunting process, however, comes the acceptance or assignment of the shunting task. Therefore, we decomposed the shunting operation into task-acceptance and task-processing times. The combined execution of both sub processes in our context corresponds to the performance of the drivers. The task-acceptance time can be defined as the time between task creation and task acceptance. Consequently, it is crucial to consider in which platform the drivers can receive and accept the task. In our research setting, we employed an intra-organisational digital platform with a broadcast mechanism. Thus, after the task is created by, for example, warehouse employees, the task is displayed to all available drivers. In the application, drivers can view the starting and ending points of the task, the timestamp of the task generation and the task prioritisation. The task prioritisation is categorised into A, B or C, with A representing the highest priority level and C the lowest. Task prioritisation is based on guidelines regarding the task type embedded within the app, ensuring that the process is executed automatically. The generated but not yet accepted tasks are structured in descending order based on prioritisation and, within each priority category, in ascending order based on the timestamp of the task creation. If drivers decide to accept a task, they confirm it within the application, which marks the end of the task-acceptance time. The figure in  Appendix 1 illustrates the visualisation within the application.

After the task has been accepted, the task-processing time, also referred to as execution time, commences. The main objective of the task-processing time is to couple semitrailers, subsequently transport them to the destination and safely uncouple the semitrailers (Clausen and Goedicke, 2012). The task-processing time begins with the task acceptance and concludes with the confirmation that the task has been completed by the driver. After accepting the task, the driver proceeds to the current location of the semitrailer to be transported. The travel time to the semitrailer is influenced by the positions of the driver and the semitrailer at the time of task acceptance. Upon arriving at the semitrailer’s location, the driver lifts up the semitrailer. However, before coupling, wheel chocks securing the trailer against rolling must be removed (Verkehr, 2020). Following the coupling, the driver transports the semitrailer to the designated destination. Because shunting vehicles are exempt from the one-way street regulation, the transportation route is based on the assumption of the shortest route. In contrast to conventional vehicles, shunting vehicles, also known as terminal trucks, have better manoeuvrability and can quickly pick up semitrailers and transport them with pulled out fifth wheel (Clausen and Goedicke, 2012). Upon reaching the destination, the driver safely uncouples the semitrailer and secures it with wheel chocks to prevent rolling (Program, 2013). Subsequently, the driver confirms the completion of the task in the digital platform, marking the end of the task-processing time.

Throughout their entire working hours, the drivers can communicate with each other and with warehouse employees using radio devices. Our dataset consisted of archival data of all yard shunting transports during the period from 1 January 2022 to 30 April 2023 and included 146,446 transports performed by 878 unique driver teams. The number of drivers operating simultaneously, and therefore the team size, varied between two and five drivers during the observation period. For each transport, the dataset included, among other elements, information about the starting and ending points of the transport, type of task, task prioritisation and the driver who executed the task. Additionally, it listed when the task was generated, accepted and completed. The driver teams were temporary teams aggregated at a five-minute level, which means that we analyse the team compositions at a five-minute interval level. In addition to the loss of various influencing factors when aggregating teams at a longer interval, the fact that the average task-processing time is shorter than five minutes also supports this aggregation level. The drivers operated at the same hierarchical level.

Our primary focus was on researching the impact of team familiarity and high-priority task specialisation on two dependent variables, including task-processing time and task-acceptance time. First, we measured task-processing time as the time between the two timestamps: task accepted and task completed. Both timestamps are set by the drivers through corresponding features in the digital platform. Between these two timestamps, the core process of physical shunting by the drivers takes place. It involves picking up semitrailers at a gate visible on the interface of the digital platform (after task acceptance), changing the physical location of the semitrailers by transporting them in the yard to a target location and accomplishing the final documentation when completing the task on the interface of the digital platform. We selected task-processing time as a dependent variable because it has been widely applied when researching digital platforms (Agha et al., 2019; Dunn et al., 2023; Esposito De Falco et al., 2017).

Second, we measured the task-acceptance time as the time between the timestamps of task generation and task acceptance. Both timestamps are marked on the interface of the digital platform. Upon task generation, it is displayed to all available drivers. Once a driver decides to accept the task, they select it through a corresponding feature in the application running on a personal digital assistant. The process and, therefore, the task-acceptance time ends with selection of the task in the application. Task-acceptance time has not been comprehensively examined in the team familiarity and task specialisation literature to date. The same applies for the behavioural aspect of teams working with digital platforms.

Our research includes two major independent variables of interest. First, team familiarity quantifies the amount of shared experience that team members have accumulated while working together on shunting tasks (Kim et al., 2021). For each pair of individuals on a team, we calculated the number of times that the pair had collaborated for a time period of five minutes in the yard, not including the current time period measured. This count variable was denoted as K_ij. We summed K_ij across pairs on the team and divided by the possible number of pairs to define the level of relationship-specific knowledge available to the team (Reagans et al., 2005):

Herein, N is the team size and K_ij is the number of five-minute intervals that a driver i has operated within a team j. The variable indicates the average amount of experience that team members had while working with each other (Staats, 2012). We scaled our team familiarity variable with 1/1,000, equalling 5,000 min, which is equivalent to a total of two working days with eight hours of work (5,000 min/60 min per hour/8 h per workday = 10 workdays). Therefore, the coefficients for team familiarity are interpreted as the changes in the dependent variable when the amount of shared experience increased by ten workdays or two weeks of work.

Second, high-priority task specialisation quantifies the amount of shared experience that team members had accumulated while working together on high-priority tasks. The prioritisation is essentially based on a predefined logic depending on the type of task. There are three categories of prioritisation: A, B and C, where A represents the highest and C the lowest prioritisation level. To test H2 and H4, we divided task specialisation into two separate measures to capture whether the prior task experience of teams occurred when the teams processed high-priority or non-high-priority tasks. This approach resembles what has been applied in prior studies to divide one cumulative task experience variable into two variables representing different categories, e.g. same-location or different-location team familiarity (Staats, 2012) or firm-specific and non-firm-specific task experience (Huckman and Pisano, 2006). For high-priority task specialisation, we counted the number of prior high-priority tasks a team worked on together prior to the current task, not including the current task. We summed these values across all unique pairings on the team for (1) each month to calculate same-month high-priority task specialisation and (2) each day to calculate same-day high-priority task specialisation.

Next, we describe the control variables that were included in our empirical analysis to control for various effects that have previously been found to impact team performance. Based on the recommendations outlined in Bernerth and Aguinis (2016), our selection of control variables is grounded in a theoretical foundation, a selective approach, data analysis and transparency. Specifically, the theoretical foundation ensures that the relevance of control variables can be theoretically justified. The selective approach involves including only those variables that provide an alternative explanation for the observed effects or contribute to establishing discriminant validity. The data analysis criterion entails assessing whether a variable correlated with the dependent variable before incorporating it into the model. Transparency is achieved by explicitly describing and documenting why a variable is included, and what impact it has on the results (Bernerth and Aguinis, 2016).

First, team size reflected the number of drivers operating in the yard and utilising the digital platform. Controlling team size is necessary because it can significantly influence performance. On the one hand, larger teams may encounter greater challenges in coordinating their actions, potentially impairing their performance (Kim et al., 2021). On the other hand, larger teams benefit from access to more resources, which may positively enhance their performance (Avgerinos et al., 2020). Therefore, the control variable team size is considered a variable that provides an alternative explanation for the observed effects according to Bernerth and Aguinis (2016). For this study, teams were aggregated at a five-minutes level, and observations included teams comprising between two and five members during the operation period.

Second, we included task prioritization as a three-stage dummy-coded variable to control for non-high-priority tasks, which we excluded in our quantification of high-priority task specialization:

Third, we controlled for remaining same-day workload because several studies have recently turned to understanding the impact of workload as an integral environmental factor on team performance. Kc and Terwiesch (2009) found that workers speed up as workload increases and that this positive effect may be diminished after long periods of a high workload. Anand et al. (2011) found that service providers slow down as customer intensity increases, which causes the equilibrium service value to increase. As a result, they suggested that servers may become slower when the number of competing servers increases. We expected that the acceptance of tasks could be influenced by the workload. Hence, the control variable remaining same-day workload provides an alternative explanation for the observed effects, while also contributing to establishing discriminant validity by ensuring that the main effects are not confounded by variations in workload (Bernerth and Aguinis, 2016). Therefore, we included the remaining workload on a daily basis as a control variable. This was measured as the total number of tasks remaining per team and per day.

Fourth, we controlled for driving distance in the yard. In logistics and operational settings, driving or travel distance can significantly impact efficiency, performance and resource utilisation. Extended driving distance often lead to increased fatigue, which can impair driver performance, reduce time efficiency and increase resource consumption, such as fuel or energy (De Santis et al., 2018). The driving distance thus provides an alternative explanation for the observed effect, as longer distance can impact efficiency and productivity by increasing resource consumption and time expenditure. Within the research context of task-processing time, drivers transport semitrailers from a starting to a destination point within the yard. As this time is highly dependent on the distance covered, we considered the driving distance a control variable. In this regard, we examined the distance between the starting and destination points of the semitrailer based on the assigned task in metres. The distance covered by the driver to reach the starting point was not considered. To measure the distance between the starting and destination points, we initially measured the distance between the centre of all warehouse sides to another. This means, for example, that we measured the distance from the south side to the north side of the warehouse. Subsequently, we analysed individual transports for their transitions between warehouse sides and recorded the previously measured average distance as the driving distance. Therefore, we operationalised the driving distance using a kind of distance matrix. If both the starting and destination points were on the same warehouse side, we did not record any distance. Due to the underlying one-way traffic rule, except for internal drivers, this was considered minimal and could be incorporated into the analysis.

Finally, during the course of a day, typically as a result of warehouse demands, the existing task requirements change, and performance can depend on task characteristics. We categorised each task type with a task fixed effect to account for different task characteristics. This categorisation was also determined by the task creator and was based on, among other factors, the type of transported goods and the associated task for the warehouse. Drivers do not have visibility to the task type in the digital platform, but they can infer the task type based on the starting and ending points of the semitrailer. Therefore, we controlled the effects for task types, referred to as task identifiers in our context. Different task types could significantly impact performance outcomes because they could require varying levels of time, resources and coordination (Rossiter Hofer and Knemeyer, 2009). As such, task types can be considered as a control variable, as it helps explain variations in performance. Table 2 summarises the descriptive statistics of our variables.

We applied a mixed-effects model to our longitudinal data because the archival dataset provided by our partner firm included multiple drivers observed repeatedly over time. Applying standard ordinary least squares (OLS) regression models to our data would violate the assumption of independence (Awaysheh et al., 2021). The mixed-effects model allowed us to estimate the impact of our predictor variables on service performance more than once without artificially inflating our estimates due to the dependency structure in the data (Bliese et al., 2018). Mixed-effects models were particularly useful in our case because we were dealing with a nested data structure (Brown, 2021).

We found no noteworthy multicollinearity issues in our dataset as the variance inflation factors (VIFs) were below the threshold value of 5.00. VIFs measure how much the variance of an estimated regression coefficient increases if the predictors are correlated. If VIF is high, it indicates that predictor variables are highly correlated. We calculated the VIFs for each variable by $VIF(Xj)=11−Rj2$⁠. A VIF value of 1 indicated no multicollinearity, while values greater than 10 suggested increased levels of multicollinearity. All VIF values in our data were well below the commonly used threshold of 5.00 (Miles, 2014).

Furthermore, we needed to account for between-driver variability through random slopes. Random slopes address the dependence between observations by clustering observation-level data from the same high-level subjects. We summarised the ICC values for the team and driver level as well as all dependent variables in Table 3.

Hence, random effects accounted for between-team variances that might potentially arise from team heterogeneity (Bresman, 2010). To test whether random effects should be applied, we calculated the intra-class correlation coefficient (ICC) for one no-predictor model and tested whether within- and between-subject variance existed for drivers. The ICC values of the no-predictor models can be interpreted as the total amount of variance in the dependent variable service time originating from the driver-related variations. These variations are attributed to the between-driver rather than the within-driver variations over time. Higher values also indicate a nontrivial degree of observation non-independence where traditional regression approaches are inappropriate (Bell et al., 2019). We summarise the ICC values for the team and driver level as well as all dependent variables in Table 3.

The ICC value for the task-acceptance time was 0.048 for the team level, indicating that 4.8% of the variance in the task-acceptance time was attributable to between-team variability, while the rest can be explained by within-team variability over time. The values in Table 3 indicate that time differed between the drivers and teams, suggesting that estimating more complex models with hierarchies and temporal change was warranted.

We began our analysis by examining the impact of team familiarity on our first dependent variable, task-processing time, to answer H1. We used a linear mixed-effects model with two random intercepts for each driver (α_0i) and each team (δ_0i) capturing between-team and between-driver differences that were not accounted for when including fixed effects. Our first model for task-processing time is:

The vector λ contains variables to control for time-invariant heterogeneity within tasks. The vector τ is used to account for time-varying heterogeneity, including month, day of the week (Monday to Saturday) and the date. Our results are summarised in Table 4.

In H1, we predicted that the greater the team familiarity within a team, the shorter the task-processing time would be. We draw the reader’s attention to model A1 in Table 4. Herein, we find a negative 0.020 and significant p < 0.01 coefficient for the effect of team familiarity on task-processing time. This implies that, all other variables being constant, a 1,000-unit increase in team familiarity was associated with a decrease of 2% in task-processing time. Therefore, we can confirm H1.

In the second model for task-processing time, we also applied a linear mixed-effects model with two random intercepts for each driver (δ_0i) and team (α_0j).

In H2, we predicted that the greater the same-month high-priority task specialisation within a team, the shorter the time would be to process a task. We now draw the reader’s attention to model A2 in Table 4. Herein, we find a negative 0.175 and significant p < 0.01 coefficient for the effect of same-month high-priority task specialisation on task-processing time. This implies that, all other variables being constant, a 1,000-unit increase in same-month high-priority task specialisation was associated with a decrease of 17.5% in the task-processing time. We can therefore confirm H2.

We continue our analysis by examining the impact of team familiarity on the second dependent variable, task-acceptance time, to address H3. We used a linear mixed-effects model with two random intercepts for each driver (α_0i) and each team (δ_0j), thereby capturing team and driver heterogeneity effects that were not accounted for when including fixed effects. Our first model for task-acceptance time is:

The vector λ contains variables to control for time-invariant heterogeneity within tasks. The vector τ accounts for time-varying heterogeneity, including month, day of the week (Monday to Saturday) and the date. Our results are summarised in Table 5.

In H3, we predicted that the greater the team familiarity within a team, the longer the task-acceptance time would be. In model B1 in Table 5, we find a positive 0.148 and significant p < 0.01 coefficient for the effect of team familiarity on the task-acceptance time. This implies that, all other variables being constant, a 1,000-unit increase in team familiarity was associated with an increase of 14.8% in task-acceptance time. Therefore, we can confirm H3.

Another interesting finding resulted from our random intercepts and the quantification of variances with τ. When we compared the τ_drivers, we found the highest value for the task-processing time. The value 0.11 in model B1 is a measure of variance, which is the average of the squared time differences from the mean. In the case of a mixed-effects model, this is the variance of the intercepts across the different groups or levels in the model (Bliese et al., 2018). Therefore, we can derive that between-driver heterogeneity was especially relevant for task processing. However, the contrary was the case for the team level: when we compared the τ_team, we found the highest value for the task-acceptance time (0.08 in model B1). This suggests that between-team heterogeneity was especially relevant for task acceptance.

Finally, we examined the impact of specialisation in high-priority tasks on task-acceptance time to address H4. We used another linear mixed-effects model with two random intercepts for each driver (α_0i) and each team (δ_0j) and captured between-team and between-driver differences that were not accounted for when fixed effects were included. Our model is denoted as follows:

The vector λ contains variables to control for time-invariant heterogeneity within tasks. The vector τ accounts for time-varying heterogeneity, including month, day of the week (Monday to Saturday) and the date. Our results are summarised in Table 5.

In H4, we predicted that the greater the same-day high-priority task specialisation within a team, the longer the task-acceptance time would be. Model B2 in Table 5 shows a positive 0.180 and significant p < 0.05 coefficient for the effect of same-month high-priority task specialisation on task-processing time. This implies that, all other variables being constant, a 1,000-unit increase in same-month high-priority task specialisation was associated with an increase of 18% in task-acceptance time. Additionally, we found a positive 0.010 and significant p < 0.01 coefficient for the effect of same-day high-priority task specialisation on task-acceptance time. This suggests that, all other variables being constant, a one-unit increase in same-day high-priority task specialisation was associated with an increase of 1% in task-acceptance time.

We also calculated the average marginal effects (AME) to allow for a relative comparison of the effect sizes. We were primarily motivated by the question of whether the habitual effect was stronger for same-day or same-month high-priority task specialisation within a team. AMEs are calculated by averaging the partial effects (the marginal effects) of a predictor variable on the dependent variable across all observations. This approach ensures that the effect size is directly comparable between all variables in the model. Our AME for one unit of same-day high-priority task specialisation was 0.00962, while the AME for one unit of same-month high-priority task specialisation was 0.00003. This suggests that the habitual effect exists in both temporal dimensions but is stronger for same-day high-priority task specialisation.

A key questions for any empirical paper is whether different modeling choices might generate different results. Therefore, we repeat our main analyses from Model (A1) and (A2) using a nested random slope component for drivers instead of separated random slope components for drivers and teams. While our application of separated random slope components is grounded in the assumption that drivers’ and teams’ average performance is independent of each other, one might argue that average driver performance depends on the average team performance. Therefore we incorporate a nested random slope component where drivers are nested in teams. For the model with our dependent variable time-to-accept, we drop α_0i = γ₀₀ + υ_0i as defined in Formula (2) and Formula (5) as well as δ_0j = γ₀₀ + υ_0j as defined in Formula (3) and Formula (6), which we replace (αδ)_0i/j = γ₀₀ + υ_0i/j. The random intercept component αδ)_0i/j is defined by the grand intercept of the population γ₀₀ and the deviations of driver-team combinations’ slope υ_0i/j from the grand intercept γ₀₀. This robustness check allows us to assess the sensitivity of our results to the replacement of separated random intercepts by nested random intercepts, helping to determine if the model’s findings are consistent across different specifications (Wooldridge, 2020). Model (R-A1) and (R-A2) in Table 6 shows that the directions of the main effects remain robust when comparing to the initial results of Model (A1) and (A2).

For the model with our dependent variable task-acceptance time, we drop α_0i = γ₀₀ + υ_0i as defined in Formula (8) and Formula (11) as well as δ_0j = γ₀₀ + υ_0j as defined in Formula (9) and Formula (12), which we replace (αδ)_0i/j = γ₀₀ + υ_0i/j. Again, the random intercept component αδ)_0i/j is defined by the grand intercept of the population γ₀₀ and the deviations of driver-team combinations’ slope υ_0i/j from the grand intercept γ₀₀. Model (R-B1) and (R-B2) in Table 7 shows that the directions of the main effects remain robust when comparing to the initial results of Model (A1) and (A2).

Previous research on digital platforms has frequently drawn on the resource-based view (Wernerfelt, 1984), transaction cost economics (Baronian, 2020), internationalisation theory (Hymer, 1976) and dynamic capabilities (Teece et al., 1997; Eisenhardt and Martin, 2000) when examining platforms in logistics and operations management contexts. However, these perspectives largely overlook the significance of social interaction and operational-level processes, particularly in team settings. We observed a two-sided, intra-organisational platform involving both warehouse and yard employees. We identified the contributions of these digital platforms to blue-collar productivity, employer–employee relations and social interaction.

Our findings contribute to the existing literature by demonstrating that increased team familiarity reduced task-processing time but increased task-acceptance time within a digital platform setting, using the example of yard management transport tasks. Similarly, specialisation in high-priority tasks within the same month shortened task-processing time, while specialisation on the same day lengthened task-acceptance time. These insights highlight the critical impact of social and operational dynamics within teams on performance outcomes.

Previous research has made a strong case for the human factors that impact warehouse performance, especially in manual material handling (Grosse et al., 2017; Sgarbossa et al., 2020). We contribute to the research on human factors through our empirical investigation of yard management tasks. Human factor issues have been identified as relevant factors in manual material handling and include physical, psychological and social factors (Grosse et al., 2017; Sgarbossa et al., 2020). Social factors include interactions among various actors. These factors lead to performance differences that are observable between humans and within humans (Loch and Wu, 2008; Urda and Loch, 2013). Thus, we observed that the experience of working together plays an important role in, among other things, the efficiency of sub processes. This establishes a connection to the transactive memory system theory, which proposes that individuals within collectives learn, store, utilise and coordinate their knowledge to achieve personal, group and organisational objectives (Wegner, 1987; Hollingshead et al., 2011). What we observed can be considered evidence that it exists for blue-collar workers sharing the same working environment and the same tasks organised with a broadcast mechanism.

In terms of the employer and employee relationship, the observed digital platform made the following contributions. First, the digital platform utilised task prioritisation, with tasks being created by employers and prioritised based on their task type. Therefore, employers use this as a management tool because prioritization based on nudge theory aims to initially apply nudges that signal high priority, with the goal of reducing task-acceptance time (Thaler and Sunstein, 2008; Johnson et al., 2012). Yet, a high number of highly prioritised tasks can lead to employees becoming accustomed to them, resulting in a habitual effect (Skinner, 1953). This implies that for the employer-employee relationship, the management of tasks and their prioritisation becomes transparent through the digital platform. Depending on the management approach, this can have varying impacts on factors such as task-acceptance time.

A main advantage of the examined platform type is the reduction of information asymmetries, which increases blue-collar productivity (de Reuver et al., 2018). Specifically, the use of the digital platform in this research ensured that all stakeholders had identical information about trailer locations. Consequently, this reduced idle times and the cycle time of the trailers. Additionally, it provided more transparency about pending tasks and increased productivity in operational warehouse and yard management processes.

Essentially, digital platforms seek to greatly restrict informal communication (Culotta et al., 2024). However, the above discussion demonstrates that informal communication still occurs and can influence both task-acceptance and task-processing times. This highlights the contribution of the digital platform to social interaction by showing how it shapes the nature and impact of informal communication within the yard.

Particularly with respect to the hypotheses concerning task-acceptance time, various explanatory approaches from the field of social research, such as self-determination theory, social loafing and stimulus-response theory, were explored. The hypotheses regarding task-acceptance time, in combination with greater team familiarity and same-day high-priority task specialisation, were confirmed. However, the underlying factors merit a more nuanced discussion.

One factor that can influence task-acceptance time is the level of experience of individual internal drivers (Kolb, 1984). In the empirical setting under consideration, with the exception of one person, all of the internal drivers had previously been employed as warehouse employees or drivers for transport operations outside the yard. Consequently, these individuals can be credited with a high degree of expertise in operational processes. An internal driver can therefore prioritise a task higher or lower independently of the priority assigned to the task, based on their level of experience, which can have immediate effects on task-acceptance time.

In addition to the level of experience, the formal and potential informal feedback systems for internal drivers should be considered. It must be noted that there is no formal penalty system in place that rewards or evaluates the number of accepted tasks. However, it should be noted that on an informal level, there may be a form of feedback through praise and criticism from the team leader. The team leader, in addition to the internal drivers, is responsible for an area in the warehouse. The team leader can communicate with the internal drivers during the shift via radio devices.

Not only can the team leader communicate with the internal drivers during the shift via radio devices, but selected warehouse employees can also establish contact with the internal drivers through these devices. The communication between warehouse employees and internal drivers, combined with the experience level of each internal driver, can amplify the informal social influence on task-acceptance time. For instance, warehouse employees may exert emotional pressure on internal drivers to accept and perform tasks more quickly, regardless of the assigned prioritisation of the task. Additionally, the social relationship between the warehouse employee and the internal driver may also serve as an influencing factor. Furthermore, the internal drivers can also communicate with each other using the radio devices. Just as warehouse employees can exert emotional pressure on internal drivers through informal communication channels, this effect can also occur among the internal drivers themselves. They can exert a form of social pressure on each other, which can influence both task-acceptance and task-processing time.

The nature of communication greatly depends on the individuals involved. The communication via radio devices in general is not represented in the provided dataset and inherently presents a point for discussion. In essence, digital platforms aim to strongly limit informal communication (Kapkaeva et al., 2021). However, the above discussions demonstrate that informal communication still occurs and can influence task-acceptance and task-processing times. Therefore, the interplay between digital platforms and informal communication should be examined as its occurrence can be reduced but not entirely prevented by the introduction of digital platforms. As derived from this, a question for further research could be how social relationships and informal communication influence the efficiency of digital platforms. One could argue that social relationships and informal communication are considered crucial success factors for digital platforms, more than the technical design of the platforms.

To date, researchers and practitioners have paid limited attention to yard management at warehouse sites (De Muynck and Tunstall, 2019). Therefore, yard management at warehouse sites is often referred to as the “black hole” of the supply chain (McCrea, 2020). It links transportation with intra-logistics at warehouse sites and plays a crucial role in handling operations, such as those involving semitrailers (Clausen and Goedicke, 2012). Previous research has focused on, for example, the field of digitalisation and yard management systems, with an emphasis on transparency regarding the location of trailers. Another focus is the integration of transport, yard and warehouse management systems (Tunstall et al., 2022). However, a consideration of yard management systems in combination with human factors is missing. Other research has focused on optimising the shunting process based on simulations with an emphasis on identifying a suitable shunting strategy (Clausen and Goedicke, 2012) and on the identification and evaluation of critical sub processes in the yard. In this context, the shunting process was identified as a critical yard process (Müller et al., 2022). However, this study did not consider human factors in relation to a digital platform for managing the shunting process. Instead, we contribute an empirical investigation of social interaction and real-world operational-level processes for driver teams focused on shunting transport tasks at a single location of a German brick-and-mortar grocery retailing company, encompassing 146,446 yard transports completed between January 2022 and April 2023. Aside from the empirical analysis of the extensive data related to the critical shunting process in the yard, the analysis focuses on the interplay with digital platforms from a human factor perspective. This analysis is dedicated to the team level rather than the individual level.

In addition to the theory contributions, our study offers management implications divided according to task-acceptance and task-processing times. Regarding task-acceptance time, it is recommended for management to implement a minimum number of accepted tasks per shift to counteract the social loafing effect (Kyngäs and Nurmi, 2021). A corresponding feedback system and formal incentive scheme ensure the achievement of a minimum number of accepted tasks per shift (Ågren et al., 2022; Ribeiro and Alves, 2023; Ghazizadeh et al., 2012). Furthermore, a responsible approach to prioritisation assignment is recommended to counteract the habitual effect (Sharif et al., 2023; Mancuso et al., 2014).

Regarding task-processing time, two main implications for management emerge. The more frequent increase in the deployment of same-driver teams in line with prior research on team familiarity and can lead to reduced task-processing time. Through the more frequent deployment of same-driver teams, the concept of the learning curve is applied, thereby reducing task-processing time (Luciano et al., 2018). Another potential approach is an implementation of a fixed allocation of drivers to yard areas with the aim of reducing the driving time to the receiving semitrailer, and, therefore, the time to process a task. This approach represents an adaptation of the zoning concept from warehouses (Azadeh et al., 2023; Yu and Ramanathan, 2009).

Presented empirical results do encompass limitations: First, data originate from a single depot and yard, limiting generalisability. Second, the applied dataset does represent activities for 16 months only – calling for long-term analyses across several years within future research. Third, temporary team configuration was defined using five-minute intervals, in line with average processing times experienced. This element could be varied in future research to gain further insights into team composition. As the analytical focus was dedicated to team familiarity, individual performance was not explicitly examined. Drivers in larger depots operate at the same hierarchical level, share physical and digital workspaces and are connected through communication tools and joint task objectives. While these aspects provide compelling insights, future research could explore additional dimensions beyond team perspective. Despite these limitations, the study offers valuable contributions to operations management research and practise.

Our research examines for the first time the effects of team familiarity and task specialisation on task-acceptance and task-processing times in an empirical setting by focusing on yard management at a warehouse site while considering a digital platform that utilises a broadcast mechanism for task distribution. The results from this specific empirical setting are transferable to other empirical settings.

Task specialisation is highly relevant for pickup and delivery operations in last-mile distributions, which refers to the supply chain process where goods are transported from a warehouse to an end customer. This is grounded in the effect that having customer-specific knowledge can help reduce service time. That is, customer-specific knowledge, such as time windows, unloading point or responsibilities can decrease the likelihood of unproductive waiting and searching times. One example is delivery of merchandise from retail stores, for which store personnel are scheduled according to specific time windows. Being unable to accomplish deliveries before these time windows might cause customer-related waiting times when no store personnel are available. Moreover, task specialisation and the related customer-specific knowledge might also support outsourcing decisions, e.g. in parcel deliveries for hospitals. Customer demands to depalletise shipments at hospitals and distribute elements of shipments within a hospital might be too specialised for logistics providers that standardise their operations for the distribution of palettes to one customer location. This in turn underlines the fact that the task specialisation construct is not solely relevant for yard management operations but also applies to pickup and delivery in last-mile distribution.

In addition to the transferability of the results and analysis to other areas of logistics or other industries, approaches related to team familiarity and task specialisation are found in the healthcare setting. For example, for ambulance teams, the total time to perform a task can be divided into the time spent for patient pickup and the patient handover process. While task distribution is based on a dispatch mechanism, the results for the time to process a task can be transferred to our research (Akşin et al., 2021). Thus, the overall process of ambulance teams can be divided into a task-acceptance and a task-processing time. Due to the task distribution through a dispatch mechanism, this sub process is not transferable to the results from the yard management setting. However, the task-processing time is transferable, particularly with regard to the two variables of team familiarity and task specialisation. This is due to the fact that the concept of team familiarity is applicable to ambulance teams because they work together with varying degrees of mutual knowledge (Reagans et al., 2005). Additionally, the diversity and different priorities of the tasks suggest the presence of task specialisation (Akşin et al., 2021). In addition, the study results are also applicable to the area of operating room procedures, which is analogous to the ambulance setting. While task allocation in this case is assumed to stem from a dispatch mechanism, the results of the independent variable team familiarity can be transferred to the task-processing time, which in this case is the surgery duration in minutes (Avgerinos and Gokpinar, 2017).

In summary, the discussions across various empirical settings, such as last-mile distribution, the food delivery sector and the healthcare sector exemplify that the results of the present research are generalisable but still must be proven in these particular contexts. Fundamentally, however, it can be assumed that the results are transferable to other industries, particularly with regard to task-processing time and the variables of team familiarity and task specialisation. Concerning the use of digital platforms in our research, it can be stated that the generalisability of the results is ensured by the fact that nearly all companies in the industrial context rely on and use digital platforms.

Future research on social aspects of digital platforms might also address how individual workers can learn from and with their peers. Previous research underscores the significant impact of social learning on the performance of group members (Liu et al., 2024). There is general consensus that individuals learn by observing and modelling the behaviour of fellow group members (Argote et al., 2021) and that social learning is primarily motivated by the need to address environmental uncertainty, which refers to the unpredictability in the external environment that impacts an organization’s decision-making and goal achievement (Peteraf and Shanley, 1997). There are three principal sources of experience for social learning in groups. First, learners accumulate vicarious experience by observing group members’ responses to situations that occur in the same environment. Vicarious experience enriches learners’ behavioural repertoires, enabling them to respond effectively to similar situations and to enhance their own performance (Martin and Cuypers, 2024). Second, learners gain direct experience by observing fellow members’ responses in interactions, enhancing the anticipation and response repertoire for efficient group interactions (Avgerinos et al., 2020). Third, reinforcing and inhibited effects arise from observing the consequences of group members’ responses to situations in the same environment. A reinforcing experience supports learners in evaluating their anticipation and response repertoire based on performance outcomes (Bosmans et al., 2020). Digital platforms might be an option to access all three sources of observational learning.

Future research could place an emphasis on the role of individuals at the micro-level and their potential impact within organisational settings and operational performance. A deeper look into understanding how employees adapt and innovate processes or digital platforms by observing peer responses to external pressures is particularly warranted. For instance, observing how team members modify processes or utilise digital platforms in response to fluctuating volumes of goods to be transported or external requests from warehouse employees could provide valuable insights into adaptive behaviours. Investigating how such micro-level dynamics vary across industries and are shaped by organisational factors or digital platforms could enrich the research on social learning impacting logistics operations issues connected to platforms and beyond.

Future research on yard management could benefit from enhancing simulation models. Specifically, incorporating both “time to accept” and “time to process” tasks into simulation models with assumptions regarding the distribution of these times would be valuable. Typically, these distributions are presumed to follow a normal distribution, however, this assumption often diverges from real-world functional forms. By analysing the dataset and utilising regression weights, one could establish functional relationships for “time to accept” and “time to process” tasks that more accurately reflect real-world settings. Such an approach exemplifies empirically grounded analytics, an area currently gaining significant attention in operations management research (de Treville et al., 2023).

Complementing such a quantitative approach, a qualitative research approach could analyse a more holistic view of sustainable concepts in yard management. As the social sustainability dimension is already present here with the involved human actors and related theory insights, this could be extended towards issues of fair work, data protection or work safety as well as towards green (environmental) yard management sustainability issues regarding resource use, emissions or the handling of toxic and dangerous goods.

Although organisations and research increasingly rely on digital platforms and temporary teams to perform tasks, our understanding of task allocation via a broadcast mechanism utilising a digital platform in the context of team familiarity and task specialisation is limited. Therefore, in this study, we observed the impact of team familiarity and task prioritisation. We offer new findings of how team familiarity and task specialisation influence task-acceptance and task-processing times.

Most of the previous literature in operations and management has focused on the influence of task-related specialisation on team performance (Narayanan et al., 2009; Huckman and Staats, 2011). In addition, there are initial approaches where temporal aspects of task-related experience are examined (Staats, 2012). Approaches regarding the influence of other task characteristics, such as task prioritisation and its impact on the time to accept or process a task, are still underrepresented in the research. Consequently, the present study enhances our understanding of team familiarity by examining team familiarity and task prioritisation with regard to task-acceptance and task-processing time in the context of a digital platform that employs a broadcast mechanism for task distribution.

Our key findings indicate that greater team familiarity within a team reduces task-processing time on the one hand and increases task-acceptance time on the other hand. Regarding a greater same-month high-priority task specialisation within a team, we observed a decrease in processing time. In terms of task-acceptance time, we found a longer task-acceptance time with a greater same-day high-priority task specialisation within a team. Altogether, this paper outlines the theoretical and management contributions of the use of digital platforms in intra-organisational task completion and performance with the example of driving tasks in a yard at warehouse sites and contribute to a deeper understanding of the relationship between digital platforms and intra-organisational team dynamics.