Prerequisites for a beneficial knowledge transfer between manufacturing plants

The paper aims at exploring the prerequisites for a beneficial knowledge transfer between manufacturing plants of multinational companies (MNCs), by taking the characteristics of the knowledge sending and knowledge receiving plant into consideration. This research seeks to understand how efforts undertaken by manufacturing plants, and how collaborative tools and coordination mechanisms influence a successful knowledge transfer. The study includes thirteen case studies conducted in manufacturing plants from four different European countries (i.e., Switzerland, Romania, Albania, and Macedonia). Given the exploratory nature of this study, the authors used a qualitative research approach. The main method of data collection involved multiple semi-structured interviews at manufacturing plants, uniformly applied in each country in order to observe general patterns across different cases. Their results show that the personal interaction between knowledge sending and receiving plants is more important for a successful knowledge transfer than information systems or prior related knowledge.

K nowledge sharing enables a firm to develop itself further and to become a learning organisation (Shi & Gregory, 1998). Since many firms do not consist of one manufacturing plant only but are actually dispersed networks of plants scattered around the globe, knowledge sharing and its challenges have gained even more attention over the last couple of years (Dunning, 2006).
In this process, critique arose that studies analysing knowledge flow in manufacturing networks stay at the aggregated network level (Foss et al., 2010). Some authors argue that it is not possible to fully understand the flow of knowledge based on organisational-level analysis alone (Argote & Ingram, 2000;Foss et al., 2010;Gupta et al., 2007). They claim that the plant level, and especially the underlying processes and relationships, need to be analysed as well. It is assumed that they mediate the variables at network level (Abell et al., 2008;Foss et al., 2010).
The objective to analyse both the network and the plant level is also supported by the mixed results concerning knowledge transfer benefits. Studies analysing the achieved benefits by internal knowledge transfer demonstrate positive (Ding et al., 2013), negative (Ambos et al., 2006), mixed (Szász et al., 2016) and curvilinear effects (Erden et al., 2014). To explain these differences, some authors suggest that the benefit of knowledge transfer depends on context similarities, adaptation cost, the cost of knowledge transfer, or the competences of the knowledge receiver. In addition, mechanisms, prerequisites, and motivation to engage in knowledge transfer also need to be considered (Tran et al., 2010).
Given the lack of research in this field, our research is primarily exploratory in its nature, and aims at exploring the prerequisites for a beneficial knowledge transfer between manufacturing plants within MNCs at network and plant levels. On the plant level we aim to analyse knowledge transfer activities by taking the characteristics of the knowledge sending and the knowledge receiving plant simultaneously into consideration (RQ1). On the network level, we take into consideration the coordination mechanisms that influence knowledge transfer activities between plants belonging to the same network (RQ2). Thus, in order to gain deeper understanding in when knowledge transfer is beneficial within a manufacturing network, the study at hand follows to answer two research questions: RQ1: "What are prerequisites of a beneficial knowledge transfer at the sending and receiving plant?" a) "How do efforts undertaken by the knowledge sending and receiving plant influence knowledge transfer success?" b) "How do collaborative tools influence a successful knowledge transfer?" c) "How do similarities between the knowledge sending and receiving plant influence knowledge transfer benefits?" RQ2: "How do coordination mechanisms influence a successful knowledge transfer?" We argue that the research questions are relevant from both theoretical and practical perspectives: it aims to contribute to filling a theoretical gap by enhancing our understanding of the conditions of beneficial knowledge transfer within MNCs, and from a practical perspective it aims to offer useful guidance for plant managers to facilitate successful knowledge transfer projects. The paper therefore builds on an exploratory research, and is structured as follows. We start with the literature review, followed by the description of the case research methodology used. In the last two parts we present our results and findings followed by the conclusions and discussions in the light of our research questions.

STUDIES AND ARTICLES
Knowledge flow within the manufacturing network describes the transferring process between the knowledgesending and knowledge-receiving plant (Tseng, 2015;Gupta & Govindarajan, 2000;Minbaeva, 2007). Three different flows of knowledge exist: (1) forward, from headquarters to a plant, (2) reverse, from a plant to headquarters, and (3) lateral, between peer plants (Ambos et al., 2006).
Since we are interested in the knowledge transfer between peer plants, we concentrate in this paper on the lateral knowledge flow. In lateral knowledge flows the knowledge-sending plant needs to be willing to transfer knowledge and needs to have transferring capabilities in order for the knowledge transfer to be successful (Wang et al., 2004;Szulanski, 1996;Mahnke et al., 2005;Szász et al., 2019). The knowledge-receiving plant, on the other hand, needs to have absorptive capacities to be able to internalise the provided knowledge (Tsai & Ghoshal, 1998;Foss & Pedersen, 2002). The knowledge-receiving plant furthermore needs motivation to accept and use the provided knowledge, otherwise, the recipient may reject the implementation or feign acceptance of the provided knowledge (Hayes & Clark, 1985). According to Demeter and Losonci (2016) there are several reasons why such a knowledge transfer can become less effective, such as: • lack of motivation at both plants, especially at the knowledge-receiving, but also at the knowledgesending plant, • low level of similarity between the technological and/or geographical attributes of the sending and receiving plants (Rosenkopf & Almeida, 2003), • competitive relationship between peer plants (Dyer & Nobeoka, 2000), • causal obscurity, when we do not know how the transferred knowledge affected the performance of the knowledge-receiving plant, • low absorptive capacity of the knowledge-receiving plant, due to lack of experience and previous knowledge (Cohen & Levinthal, 1990).
Manufacturing network coordination, and especially mechanisms to coordinate the flow of knowledge within one network have been recognised as essential to combine the dispersed knowledge in the network (Ferdows, 2006;Rudberg & West, 2008;Vereecke et al., 2006). Even though many coordination mechanisms have been discussed in literature, the answer is still lacking which of these mechanisms serve as prerequisites to coordinate the knowledge flow within a network. The main coordination mechanisms are: the degree of standardisation (Maritan et al., 2004;Rudberg & West, 2008;Scherrer-Rathje & Deflorin, 2017), centralisation (Feldmann & Olhager, 2011;Netland & Aspelund, 2014;Scherrer-Rathje & Deflorin, 2017) and autonomy, split into strategic and operational decision making autonomy (Golini et al., 2016;Kawai & Strange, 2014). Standardisation is the degree of similarity of products, processes, or systems throughout the network (Maritan et al., 2004). Centralisation and autonomy are linked, as decentralisation of decision making is the main indicator of a site's autonomy (Maritan et al., 2004). Based on the presented literature, for the paper at hand, we take the dimensions of standardisation, centralisation and autonomy into consideration when analysing prerequisites of a beneficial knowledge transfer at network level (RQ2).

Knowledge flow on plant level
Knowledge flow within the manufacturing network describes the transfer process between the knowledgesending and knowledge-receiving plant (Gupta & Govindarajan, 2000;Minbaeva, 2007;Tseng, 2015). To analyse the knowledge transfer on plant level, we follow the suggestion of Szulanski (2000) and take the basic elements of knowledge transfer into consideration: the source, the recipient and the channel.
(1) The source of knowledge, in our case the knowledge-sending plant, needs to be willing and needs to have transferring capabilities in order for the knowledge transfer to be successful (Mahnke et al., 2005;Szulanski, 1996;Wang et al., 2004;Szász et al., 2019).
(2) The recipient, in our case the knowledgereceiving plant, needs to have absorptive capacities to be able to internalise the provided knowledge (Foss & Pedersen, 2002;Tsai & Ghoshal, 1998). Absorptive capacity depends on the pre-existing stock of knowledge. Literature discusses causal ambiguity to be an essential prerequisite to internalise provided knowledge (Gupta & Govindarajan, 2000;Phelps et al., 2012;Szulanski, 2000).
(3) The channel links the knowledge sending and receiving plant with each other. These channels can be formal or informal. Formal channels are those that are coordinated centrally, whereas informal channels are self-established friendships or other social ties within the manufacturing network (Adenfelt & Lagerström, 2008;Bell & Zaheer, 2007;Foss et al., 2010;Song, 2014). Thus, most of these informal channels are based on the personal interactions between employees of the knowledge sending and receiving plants.
Next to the personal interactions, formal channels also exist in the sense of information and communication systems. These systems can be used to share knowledge among members of the manufacturing network (Bigliardi et al., 2010). In this sense, they also build on the interaction between employees working at sending and receiving plants.
Finally, literature discusses that knowledge transfer can be enabled through similarities. These can be strategic or knowledge similarities (Darr & Kurtzberg, 2000;. Both can help to overcome the possible lack of the pre-existing stock of knowledge of the knowledge receiving plant in the sense that also, for example, similar strategic orientation can enable the knowledge receiving plant to understand the provided knowledge. Strategy can be operationalised through competitive priorities such as cost, quality, flexibility, delivery or innovation (Schoenherr & Narasimhan, 2012;Szász & Demeter, 2014).

STUDIES AND ARTICLES
To measure the result of the knowledge transfer, we consider two categories. The first consists of knowledge outputs such as an increase in product, process, or technology knowledge (Argote & Ingram, 2000;Darr & Kurtzberg, 2000;Kang et al., 2010). The second consists of operational performance measures, such as cost, quality, flexibility, delivery or innovation (Hayes & Wheelwright, 1984;Rosenzweig & Easton, 2010;Schoenherr & Narasimhan, 2012;Szász & Demeter, 2014). To sum up, Figure 1 shows the conceptual framework of the research paper at hand. While the source and the recipient need to have the described preconditions like the willingness to share knowledge and absorptive capacity to take the knowledge in, the channel can have different characteristics. Since we propose that these characteristics influence the knowledge transfer activities, we group the elements of the channels into (1) effort that a company puts into the interaction between knowledge sending and receiving plants, (2) the provided collaboration tools (information systems, enterprise systems), as well as (3) the similarities between the plants. The (4) coordination mechanisms (i.e. standardisation, centralisation and autonomy) frame the interplay between the plants within the manufacturing network.

Methodology
Being a primarily exploratory research, we have chosen the multiple case study method (Yin, 1988), which gives the possibility to both discover diverse knowledge roles of the multiple plants involved and their different underlying capabilities, but also to observe general knowledge transfer similarities across different cases.
The unit of analysis is the manufacturing plant within the multinational company. To gain an understanding of lateral knowledge transfer we examined 13 manufacturing plants located in Switzerland, Romania, Macedonia and Albania. We used middle-range theory development (Merton, 1968), by linking theory and empirical work. We derived dimensions from theory and refined them through case study research. Eisenhardt and Graebner (2007) recommend the case study approach for research interests such as ours, since the topic is not well documented and relatively unknown. The qualitative research approach (Eisenhardt, 1989;Eisenhardt & Graebner, 2007;Voss et al., 2002) provided us with deeper insights into the selected case plants and allowed us to generate new insights. The plant level was selected as the unit of analysis to gain information in the needed level of detail. Moreover, based

STUDIES AND ARTICLES
on Voss et al. (2002) who consider that case studies should contain between 3 and 30 cases, and Eisenhardt (1989) who reduce this interval to 4-10 cases, an approximately equal number of cases (three to four) was targeted in each of the four participating countries, i.e. Switzerland, Romania, Albania, and Macedonia, resulting in a total number of 13 case studies. More specifically, we conducted case study analysis at 3 manufacturing plants from Switzerland (S1, S2, S3), 3 from Romania (R1, R2, R3), 4 from Albania (A1, A2, A3, A4) and 3 from Macedonia (M1, M2, M3). These plants operate in various industries like automotive, electronics, food, constructions, rail, steel, cement, and industrial equipment offering a good variety in terms of country and industry to ensure a higher validity of crosscase analysis.
Case plants were selected based on the joint fulfilment of the following criteria (Szász et al., 2019): (a) they belong to a multinational company (MNC) with at least four Table 2. Cross-case analysis of case data related to successful projects Source: own editing Ad-hoc support between plants (Social ties)

STUDIES AND ARTICLES
manufacturing plants, (b) their MNC is a leading company in its field, with the headquarters (HQ) in a developed country, and operations in at least three countries, (c) the plant to be interviewed is not an isolated player (Vereecke et al., 2006), i.e. it is actively engaged in knowledge sending and/or receiving to/from other units from within the MNC. Companies with no clear knowledge roles for the analysed plant or companies that were not transparent enough regarding their knowledge transfer activities and underlying were excluded from this study.
Within the qualitative research approach, we used semi-structured interview as the main method of data collection, with an interview protocol uniformly applied in each country. Each manufacturing plant was interviewed at least once, but in some cases follow-up interviews were conducted to clarify specific knowledge transfer aspects. Interviews were targeted at the highest managerial level who oversee the knowledge transfer projects, but also actively participate in facilitating knowledge transfer activities. Researchers have participated in multiple interviews in mixed teams from different countries to enable a uniform understanding of the data collected. Field data were collected from December 2015 until March 2017.
In order to analyse our research questions, we first asked the interviewees general questions about the plants (environmental conditions, strategies and involvement in knowledge transfer activities, embeddedness in network). Second, they were asked to evaluate their embeddedness in the knowledge transfer activities within their networks, more exactly to evaluate in general how much (1) information and (2) innovation they send and receive and (3) how much training they offer to employees from other plants and how much training they received from other plant staff in comparison to other plants in the network. These dimensions were adapted from the work of Vereecke et al. (2006). The first dimension covers the amount of information transferred, which needs to be distinguished from pure data based exchange and refers to more explicit data concerning day-to-day activities related to products, processes, technology, management or services (i.e. meaningful information related to manufacturing). In addition, the second dimension aims at capturing innovation, which is related to a more tacit type of knowledge. Transferring innovation from one plant to another means that there are no routines established and most often, its implementation is based on a combination of knowledge and information. Third, trainings were also assessed as a frequent form of complex knowledge transfer within the manufacturing network of MNCs.
Next, in order to derive differences between beneficial and less beneficial knowledge transfer, we asked interviewees to identify a successful and a less successful knowledge transfer project and to explain what content explicitly has been transferred between plants. Altogether we examined 25 examples of knowledge transfer projects, as one of the 13 plants refused to identify an unsuccessful project. By having multiple knowledge transfer projects, we aimed at getting a better understanding of the specific content transferred within these processes. We were also interested in the factors that made the interviewees consider a knowledge transfer project beneficial or not which were mainly related to the process (e.g. time and resources needed to transfer knowledge) or the outcome (e.g. new process technology introduction) of knowledge transfer processes.
We then discussed the different prerequisites for knowledge transfer based on factors derived from literature. In addition, to gain in-depth insights, we encouraged the interviewees to discuss additional relevant factors. It has to be noted that each knowledge transfer project was investigated from a unilateral perspective, but we aimed to include both sender and receiver plants in our sample to gain a better understanding of both roles within manufacturing networks.
All interviews lasted between two and three hours, were taped and afterwards transcribed. The contents of all interviews were summarised into a manuscript containing the details of each knowledge transfer discussed. Afterwards, the research team conducted a cross-case analysis to compare and to gain a proper understanding of knowledge transfer processes at each manufacturing plant. Besides the interview data, company documentations, archive data, manuals, we used industry publications and personal observations in order to formulate conclusions (Szász et al., 2019).

Data analysis and findings
Following the analysis of interview data, complemented by archival data and personal observation, we categorized the 13 plants in four groups in terms of their involvement in knowledge sending and knowledge receiving activities (Table 1): -Group 1 (Net senders) consists of those plants that send a lot of knowledge to other plants, but receive only minimal knowledge from others. -Group 2 (Balanced actors) and Group 3 (Active receivers) are intermediate groups. The plants in group 2 send and receive knowledge approximately to an equal extent. Plants in group 3 are also involved in both sending and receiving knowledge, but they receive somewhat more and send somewhat less than plants belonging to Group 2. -Group 4 (Net receivers) consists of those plants that are mainly knowledge receivers, and engage only rarely in knowledge sending activities, and if they do so, they send only low amount of knowledge.

STUDIES AND ARTICLES
only partly support the existing results from literature discussing plant age and the participation in knowledge transfer activities as correlating factors. During the interviews, the interviewees were asked to identify successful and less successful knowledge transfer projects in order to derive differences between beneficial and less beneficial knowledge transfer. Table 2 shows the cross-case results of case data related to successful projects. All units investigated, no matter if they are knowledge sending or knowledge receiving plants, engage in formal (job rotation, centrally coordinated visits of other plants, trainings, standardised documentation) and informal (ad-hoc support between plants) efforts to facilitate knowledge exchange. Furthermore, all plants have various types of collaboration tools in place (such as information sharing tools, web, intranet, video conferencing, email, file sharing, and even more complex systems which can facilitate knowledge transfer activities). When it comes to similarities, all knowledge exchanging plants have similar process knowledge and most of them also have similar product, process and/or technology knowledge. The level of standardisation and centralisation is overwhelmingly high, with the majority of strategic and operational decisions making activities at headquarters. Overall, if the knowledge transfer projects were successful, the plants reported either a knowledge output in a process or a technology related knowledge increase.
In order to differentiate between beneficial and less beneficial knowledge transfer, we asked the interviewees to identify successful and less successful knowledge transfer projects and to explain what content explicitly has been transferred between plants. From this resulted 25 examples of knowledge transfer projects, one of the plants refusing to comment on unsuccessful projects. Table 3 and Table 4 highlight some examples of prerequisites leading Table 3.

Examples of successful knowledge transfer projects
Source: own editing

Company Role
Project description Supporting factors S1 Knowledge sender Duplication of production line at sister plant. Knowledge sending plant provided technological and process knowledge and supported employees at sister plant in implementing and ramping-up production.

Effort
-Monthly video conferences coordinated by central -Every six month visit of employees from knowledge sending plant at knowledge receiving plant -High effort in establishing personal ties between employees of knowledge sending and receiving plant Collaboration tools -All necessary documentation stored in information sharing software Similarities -Establishment of similar production line as at sending plant -No prior related knowledge in place Coordination mechanism -High level of standardisation in documentation Success -Technological knowledge output R3 Knowledge receiver Implementing of TPM system. Other plant provided knowledge of how to implement TPM and how to follow the rules. TPM was implemented and running after 1,5 years at interviewed plant. To compare: average implementation time within company is 3 years.

Effort
-Centrally coordinated visits of other plants that had TPM already implemented -One employee from knowledge sending plant served as a consultant for receiver plant -Consultant made frequent visits at knowledge receiving plant Collaboration tools -All necessary documentation stored in information sharing software Similarities -Similar processes at knowledge sending and receiving plant -No prior related knowledge in place Coordination mechanisms -High level of standardisation -High level of centralisation; Autonomy at HQ Success -Process knowledge output M3 Knowledge receiver New process knowledge transferred from sister plant to M3. In it, the sister plant enabled M3 in how to implement and use a process extensions to the existing production process.

Effort
-Ad-hoc visits to other plants to see how they conducted process step of interest for selection purpose -One employee from knowledge sending plant served as a consultant for receiver plant -Job rotation with knowledge sending plant was key -Network of peers contacted for ad-hoc questions (social ties)  (Table 3) and some, where the knowledge transfer was unsuccessful (Table 4). The comparison between Table 3, showing examples of successful projects, and Table 4, showing examples of unsuccessful projects, displays that the successful projects were characterised by a high level of effort in personal contact (monthly video conferences, frequent employee visits from knowledge sending plants at knowledge receiving plants, personal ties between the knowledge sending and knowledges receiving plants, job rotations between sending and receiving plants, consultancy activities, frequent visits of key knowledge sending personnel at the knowledge receiving plants, ad-hoc visits to other plants). As the manager of S3 commented: "Knowledge transfer is a people's business" or as the plant manager of M3 summed up: "Systems are not enough. Network is the key factor …. The network of people." Even though in all successful projects, information about the emphasised projects were in the company databases, they were only useful if the data quality was appropriate. And even then, the information alone was not useful if there were no employees from the knowledge sending plant explaining the content of the documents and data repositories to their peers at the knowledge receiving plants.
Prior-related knowledge showed to be of no necessity for a successful knowledge transfer. In all three examples provided in Table 3, no prior-related knowledge was in place. In comparison to this, two out of three examples of the unsuccessful projects had prior-related knowledge based on the emphasised project in place.
With regard to the coordination mechanisms, the companies showed similar characteristics, no matter if the project was successful or not. All companies have a medium to high level of standardisation and a medium to high level of centralisation with strategic decision autonomy always at headquarters and operational autonomy partly at headquarters and partly at the plant level.

Discussion and conclusion
The subsequent paragraphs briefly summarise our findings in light of the research questions of this paper.
First, related to RQ1.a, our study revealed that both knowledge sending and knowledge receiving plants

STUDIES AND ARTICLES
engage in formal and informal efforts to facilitate knowledge exchange. The knowledge transfer is more successful in case of similar process knowledge or similar product/technology knowledge base. Moreover, all companies have a medium to high level of standardisation and a medium to high level of centralisation with strategic decision autonomy always at headquarters and operational autonomy partly at headquarters and partly at the plant level. We align with , who consider that these aspects (strategic similarities and product/product family/technology similarities) are prerequisites for a knowledge transfer to be beneficial. Second, related to RQ1.b, despite the often-discussed fact that the information systems are necessary for a successful knowledge transfer, our data mirror a different perspective. Only if a personal interaction between the knowledge sending and receiving plant is established, the knowledge transfer has been considered as beneficial. In line with this, several authors have criticised the general assumption that information systems can support the knowledge transfer within organisations. Authors claim that if there is no overlap between the knowledge sender and receiver in their underlying knowledge base, knowledge transfer based on information systems alone does not work (Alavi & Leidner, 2001). As stated above, our data do not reveal the importance of prior knowledge stock. Instead, our conclusion is more in line with the statement of Roberts (2000), saying that information systems will never be able to entirely replace face-toface interactions (RQ1.a) between knowledge sender and receiver.
Third, in relation to RQ1.c, we conclude based on our data that the pre-existing stock of knowledge, claimed to be necessary to have the ability to absorb provided knowledge, does not hold true in all of our cases. With this, our data do not support the necessity of causal ambiguity for a beneficial knowledge transfer (Gupta & Govindarajan, 2000;Szulanski, 2000). Instead, the personal contact seems to be able to overcome non-existing prior knowledge stocks, which is an important implication for newly established plants with less knowledge or plants located in countries with lower levels of pre-existing knowledge base.
Fourth, related to RQ2, we conclude that at least a medium level of standardisation needs to be in place to enable the possibility of knowledge transfer, but the coordination mechanisms alone seem not to be responsible for a successful knowledge transfer.
To sum up, managers seeking beneficial knowledge transfers within manufacturing networks should take the importance of personal interactions during the knowledge exchange process into account. It is worth to invest in personal exchange activities, such as mutual plant visits, joint projects, informal meetings, joint training programs and team-buildings, to bind social ties even prior to knowledge exchange activities, as these social ties support a higher frequency of interactions and a higher willingness to participate in knowledge exchange activities, and ultimately can secure the success of transferring knowledge between peer plants belonging to the internal network of the same MNC.
Overall, our study contributes to operations and knowledge management literature by exploring the role of some plant-level and network-level factors in successful knowledge transfers, but it is also limited in its generalizability given that case study research was conducted. Nevertheless, the variance between the cases in terms of country, industry and age offers a good basis to formulate propositions for further research attempts. Further research on a larger scale should verify our results and quantify the importance of personal factors compared to structural factors in successful knowledge transfers. Future research should also involve additional factors, such as disseminative capability, absorptive capability dimensions, to investigate whether there is an interaction between these factors and the ones involved in our study that can further improve successful knowledge transfers. Another highly relevant future research issue is related to the COVID-19 pandemic as it influences the way employees work at companies and poses an important limit to personal interactions which was deemed as the most influential prerequisite of successful knowledge transfer. How changing work habits will influence knowledge transfer projects remains, thus, an issue for further investigation.