Supply chain integration and reverse logistics performance: the
moderating effect of supply complexity.
Name: Jan Brolsma (s2230348) Theme: Buyer-‐supplier cooperation Date: 12-‐07-‐‘12 Abstract:
It is widely believed that supply chain integration has a positive impact on performance in both the forward-‐ and reverse chain. As for the forward chain, there are several recent contributions that show that the level of integration is not always appropriate, as its influence on performance depends on the circumstances. Following from this line of thought, this paper explores whether similar statements can be made regarding the reverse chain. We will study 12 business units and the relationship with their key buyers and suppliers. For each unit, we will investigate the level of supply chain integration, the business conditions as a measurement of supply complexity, and their respective reverse logistics performance. The results show that higher supply complexity requires higher integrative efforts with buyers and suppliers. In contrast, supply chain integration is less appropriate under low supply complexity conditions. However, the level of integration and its influence on RL performance is restricted if the relationship with the referring unit’s buyers and suppliers is short-‐ lived, and integration and its influence on RL performance is strengthened if the referring unit is characterized by high volumes and relatively few customers.
1. Introduction
Literature suggests that managing reverse logistics (RL) can be an effective way for companies to reduce their environmental impact (Chung and Wee, 2008), and to increase their competitive position (Dowlatshahi, 2010). In order to establish a competitive RL system, supply chain integration (SCI) is considered to be important (Olorunniwo and Li, 2010), if not inevitable (Chan et al., 2010). However, recent studies with a focus on the forward supply chain reconsidered the general thought of integration as being a supply chain utopia (Childerhouse and Towill, 2011). Gimenez et al. (2012) showed that the effectiveness of SCI largely depends on the context of the supply chain. Since many aspects of the issues raised in forward supply chains are also applicable to RL (Olorunniwo and Li, 2010), it is likely that the alleged positive effect of SCI on RL might be moderated by context variables as well. This current paper seeks to investigate the effect of context on the relationship between SCI and RL performance.
Prajogo and Olhager (2012) mention that it is widely accepted, both in practice and literature, that SCI contributes to an increase in supply chain performance. Following from that, it is stated in several contributions that this effect also applies to RL. In fact, SCI is considered to be an important method to deal with the high indirect costs and uncertainty that specifically characterize the RL process (Jayaraman et al., 2008). Therefore, the ability to collaborate with various players in the reverse chain is thought to be -‐at least-‐ of equal importance as in the forward chain (Olorunniwo and Li, 2010).
Ho et al. (2002) noticed that the absence of context in most supply chain management research is a major shortcoming. In addition, Sousa and Voss (2008) emphasized the importance of context and a contingency approach in operations management. In line with this reasoning, there are several papers that investigated the effect of context on the positive relationship between SCI and performance. In general, it appears that context has a moderating effect on this relationship. However, these contributions are focused on the forward supply chain. Given that RL logistics needs a totally different approach than forward logistics (Chan et al., 2010), it seems that the applicability of these findings to RL is not straightforward. This paper incorporates context to the relationship between SCI and RL performance. We do so, by building on Gimenez et al. (2012), who investigated the effect of supply complexity on SCI and performance in the forward chain. Here, supply complexity corresponds to a certain level of uncertainty, and can be derived from business characteristics such as product variety, production volumes, life cycles, and predictability of demand.
The objective of this study is to understand the effectiveness of SCI on RL performance in different contexts. Specifically, the aim of this paper is to show that SCI is only effective on RL performance, where the context of buyer-‐supplier relationships is characterized by high supply complexity. This research partly fills the gap as stated by (Prajogo and Olhager, 2012), who argue that the reverse flows of information and materials within the concept of SCI should further be explored. We build on research that has been conducted where the focus has been on forward logistics (i.e. Van Donk and Van der Vaart (2004); Gimenez et al. (2012), and aim to find out whether these findings can be generalized to RL performance. Because the variables in our propositions consist of many dimensions, we chose to utilize the multi-‐case study approach to gather in-‐depth knowledge to test our propositions. The research will be conducted at medium-‐sized companies active in the Dutch metal fabrication industry. Within this specific industry, retailers, services, or third party reverse logistics providers, are not included. Following from previous research of Gimenez et al. (2012), a number of propositions are formulated and evaluated, by measuring the influence of supply complexity on different dimensions of SCI and RL performance.
discussed in section four. Lastly, the conclusion, implications, and future research directions are presented in section five.
2. Theoretical framework
We review the literature on RL in general, by emphasizing its relevance to today’s supply chain practices and by comparing it with its counterpart (forward logistics). Also, the impact of SCI on forward-‐ and reverse logistics performance is studied. Following from that, the influence of contextual variables (supply complexity) on this alleged positive relation is laid out. Finally, our main propositions are presented.
2.1 Reverse-‐ and forward logistics
By definition, forward logistics is related to the movement of materials from suppliers to end customers (Jayaraman et al., 1999); (Krikke et al., 1999); (Fleischmann et al. 2000); (Dowlatshahi, 2000); (Kiesmuller, 2003). On the other hand, RL relates to the movement of materials in the opposite direction. RL can be
defined as a process by which a manufacturing entity systematically takes back
previously shipped products or parts from the point-‐ of-‐consumption for possible recycling/reuse, remanufacturing, or disposal (Dowlatshahi, 2010); (Sarkis, 2003).
Despite that research for supply chain management seems to focus mainly on its counterpart (forward logistics) (Chan and Chan, 2008), it appears that RL is not a new concept. Especially during the past decade, RL has received much attention (Dowlatshahi, 2010); (Rubio et al., 2008). This development might indicate that there is a growing consensus that RL can be of critical importance to overall corporate success: since RL can be used as a competitive strategy (Olorunniwo and Li, 2010), to save costs (Chan, 2007), and to achieve a high customer satisfaction (Autry et al., 2001). RL activities can also help to reduce the negative impact on the environment (Chung and Wee, 2008); (Gonzalez-‐Torre et al., 2005), and is often initiated by regulations requiring companies to take responsibility for end-‐ of-‐life or end-‐of-‐use products (Seitz and Peattie, 2004). Therefore, RL can be an alternative use of resources that can be both cost effective and ecologically friendly. The importance of RL for many modern-‐day organizations is thus obvious.
uncertain (Chan et al., 2010), and difficult to collect (Jayaraman et al., 2008). This makes the RL process complex and costly to manage.
Following from the above reasoning, Chan et al. (2010) state that the approach to RL activities should be totally different than the approach to forward logistics. Even so, although the approach is different, Olorunniwo and Li (2010) argue that many aspects of the issues raised in forward supply chains (e.g. transportation costs, packaging materials, capacity planning) will also be applicable to RL. Given the different characteristics between forward-‐ and reverse logistics, this paper investigates whether context has a similar effect on the relationship between SCI and performance in the reverse chain.
2.2 Supply chain integration
SCI originates from a systems perspective (Christopher, 2005), where it is generally believed that optimization of the whole achieves better performance than a string of optimized sub-‐systems (Childerhouse and Towill, 2011). Despite that integrating the supply chain has always been viewed as one of the key supply chain management initiatives (Van Donk and Van der Vaart, 2004), Gimenez et al. (2012) noticed that there are many different interpretations, types and classifications of SCI. Van der Vaart and Van Donk (2008) distinguished over twenty constructs that have been used to measure the level of SCI in survey research. One of these distinctions is between upstream (backward) and downstream (forward) integration with buyers and suppliers (Prajogo and Olhager, 2012); (Frohlich and Westbrook, 2001). Another common distinction is between internal and external integration (e.g., Stank et al., 2001; Gimenez and Ventura, 2005).
Similar to the approach used in a recent study of Gimenez et al. (2012), we build on a detailed analysis by Van der Vaart and Van Donk (2008, pp. 47), who distinguished three categories of items to understand the integration of suppliers with their key buyers:
1. Attitudes (relational aspects, trust): supply chain attitudes refer to the
attitude of buyers and/or suppliers towards each other or towards SCM in general (Van der Vaart and Van Donk, 2008). Examples are a firm’s expectation with respect to the future of their relationship with suppliers and/or buyers, how they consider problems that arise in the course of this relationship, and whether they share the responsibility for making sure that the relationship works for both parties (e.g., Chen et al., 2004; Johnston et al., 2004).
2. Practices (specific activities): Supply chain practices are defined as tangible
activities or technologies that play a role in the collaboration of a focal firm with its suppliers and/or customers. Examples are the use of Electronic Data Interchange (EDI), integrated production planning, Vendor Managed Inventories (VMI) and delivery synchronization (e.g., De Toni and Nassimbeni, 1999; Frohlich and Westbrook, 2001; Kulp et al., 2004).
3. Patterns (modes of communication): Supply chain patterns relate to the
facility, frequent face-‐to-‐face communication and high, corporate-‐level communication on important issues (e.g., Chen et al., 2004; Duffy and Fearne, 2004; Bagchi and Skjoett-‐Larsen, 2005).
The above listed categories cover a broad range of items, which are used to measure different aspects of SCI. Given that SCI can be interpreted in many different ways, this classification should provide us with a deep and insightful understanding of SCI. Moreover, we are able to distinguish between three aspects of SCI, which can help us to explain and discuss our results.
2.3 Influence of SCI on forward-‐ and reverse chain performance
In a recent study, Gimenez et al. (2012) mentioned that the majority of studies that investigated the influence of SCI on performance found positive relationships, both in integration with customers and with suppliers (upstream and downstream). In fact, Childerhouse and Towill (2011) reviewed the literature concerning SCI, and found that it is emphasized by many authors that integration is an essential attribute of modern day supply chain management.
In the light of this perspective on SCI, there are several studies that found a positive relationship between SCI and performance in the reverse chain as well. In their case study, Jayaraman et al. (2008) showed that SCI could -‐and should-‐ be used to cope with the high indirect costs and uncertainties that specifically apply to RL. Based on a survey that was send to 600 US companies, Olorunniwo and Li (2010) also found that investment in information technology together with information sharing and collaboration with partners is critical to RL performance. Furthermore, Chan (2007) mentions that RL requires cooperation between two or more companies in order to be effective. Similarly, Chan et al. (2010) state that at least some level of collaboration across the supply chain is a prerequisite for every RL system.
In general, it appears that most studies have found a positive relationship between SCI and performance in both the forward, and the reverse chain. Following Daugherty et al. (2002), RL performance can be defined as two distinct dimensions: operating/financial performance, and satisfaction. These dimensions show much resemblance to the ones used by Gimenez et al. (2012), who also constructed two categories of variables to measure supply chain performance in the forward chain: (1) cost (e.g., transportation costs) and (2) service (e.g., delivery speed). Therefore, the cost and service dimensions can be used to represent RL performance.
2.4 Influence of context on integrative practices
of integrative practices were appropriate in case of lower business condition complexity. In a comparable study, de Treville et al. (2004) concluded that the considerable resources required for demand integration are only applicable when there is sufficient demand variability. In addition, Cox (2001) mentions that each relationship should be assessed whether they should be fully integrated or not, since it should be matched to supplier and customer dependency.
Recently, Childerhouse and Towill (2011) reconsidered the assumed positive effect of SCI on performance, by building on the work of Frohlich and Westbrook (2001). They emphasized that the current debate in literature is not about full-‐ or no integration. Rather, it seems that it is more about the level of integration, and that it depends on the context variables of the individual value stream.
Although the moderating effect of context on the relationship between SCI and performance has been proven, the above-‐mentioned studies focus on forward logistics. Given the different characteristics between forward-‐ and reverse logistics, it remains questionable whether the effect of context on the positive relationship between SCI and RL performance has a similar effect.
2.5 Supply complexity as a context variable
This paper analyzes the influence of context on the relationship between SCI and RL performance. More specific, we aim to investigate this effect by using supply complexity as a context variable. By building on the research of Gimenez et al. (2012), we consider supply complexity on the level of the individual (buyer-‐supplier) relationship, and assume that it corresponds with the level of uncertainty within the supply link.
Fisher (1997) was one of the first authors to consider context in supply chain management. His results indicate that a high level of uncertainty is related to market-‐responsive supply chains and innovative products. In contrast, it was also found that a low level of uncertainty is related to efficient supply chains and functional products. In line with these findings, Childerhouse and Towill (2002), and Lee (2002), state that uncertainty is one of the main drivers for SCI. Following from this line of thought, both the study of Van Donk and Van der Vaart (2004), and Gimenez et al. (2012), used the level of uncertainty experienced within the link between a buyer and supplier to gain knowledge about the influence of context on supply chain practices. By building on work of Aitken et al. (2003), and Childerhouse et al. (2002), Van Donk and Van der Vaart (2004), and Gimenez et al. (2012) used several indicators to measure business conditions. Aitken et al. (2003), and Childerhouse et al. (2002), refer to these indicators as the DWV3 approach: Duration of the life cycle, time Window for delivery (both relate to the required lead and response time for delivering products), Volume, Variety, and Variability.
(instead of business conditions). They chose to do so, by arguing that the used indicators actually characterize supply and not the single business entity.
The above-‐mentioned study of Van Donk and Van der Vaart (2004), found a positive relationship between the level of complexity and the level of integration in the supply chain. By building on this research, Gimenez et al. (2012) investigated the effectiveness of SCI in different contexts. More specifically, they showed that SCI is only effective in buyer-‐supplier relationships characterized by high supply complexity. In this paper we try to find out whether we can generalize these findings to RL. In the next section, we will develop two propositions in line with the results of Gimenez et al. (2012).
2.6 Propositions
In section 2.1, we have observed that reverse logistics requires a different approach than forward logistics. Despite the different approaches, the impact of SCI seems to have a positive effect on both forward-‐ and reverse logistics performance. Literature indicates that the positive effect of SCI on performance in the reverse chain is possibly even stronger than in the forward chain. Also, we have seen that this effect in the forward chain is moderated by context variables. This current paper will analyze the role of context on the positive relationship between SCI and RL performance, specifically by exploring whether the acknowledged effect of supply complexity on the relationship between SCI and performance can be validated/confirmed with regard to the reverse chain. This should give insight into when SCI is appropriate (under which circumstances) and when not, regarding its positive influence on RL performance.
Given the presented theoretical background, we are able to formulate two core propositions. These propositions are based on the following three considerations: (1), RL performance can be measured in terms of cost and service; (2), the level of SCI can be represented by practices, patterns, and attitudes; and (3), the context of a supply chain can be represented by the level of supply complexity.
Proposition 1: If supply complexity is high, SCI contributes to the improvement of one or more aspects of RL service and cost performance.
Proposition 2: If supply complexity is low, SCI does not contribute to the improvement of one or more aspects of RL service and cost performance.
Figure 1: Conceptual model
3. Methodology
The focus of this research is to understand the effect of context on the alleged positive relationship between SCI and RL performance. Although we have seen in section 2.3 that there are several survey-‐based contributions that investigate the role of integration in the reverse chain, we chose to conduct a multi-‐case study. The main reason for choosing this research approach is because the variables in our propositions consist of many dimensions. As such, we expect that a lot of rich data will be lost when using survey research. Therefore, the multi-‐case study approach should help us into gathering in-‐depth knowledge to test our propositions: we expect that qualitative data can help us to understand and discuss our findings. Surely, case study research can be an appropriate method to understand a relatively new concept, for answering questions related to why a phenomenon exists, and for conducting more exploratory research aimed at understanding an unknown phenomenon and learning about possible not foreseen variables influencing the phenomenon (e.g. Meredith, 1998). We aim to examine the positive effect of SCI on RL performance in different contexts. In order to do so, we have chosen to conduct a multiple case study by means of semi-‐structured interviews (site visit). Including multiple cases should increase the external validity (Voss et al., 2002). Also, this method should increase the chance of detecting different contextual factors (supply complexity) and different integrative aspects with respect to the different buyers and suppliers (Van Donk and Van der Vaart, 2004).
3.1 Selecting cases
In order to collect the necessary case data, we chose to include medium-‐sized companies who are active in the Dutch metal industry. We have done so, because companies in this industry are working with materials (metal) that can be re-‐ used/recycled, and where the raw materials typically fluctuate in price (metal prices). This should ensure at least some level of RL activities. A sample of all the Dutch companies who are (economically) active in the metal industry, were identified with the help of freely accessible information from the Dutch Chamber of Commerce (KvK, 2012). In order to be sure of a certain level of professional management, we only included companies that were listed to have more than 50 FTE’s. This resulted in a list of 192 companies. Not all the companies on this list were relevant due to the
following. Firstly, some of the companies sold their products to end-‐customers. Since we were interested in the RL performance between buyers and suppliers in a B2B market, these were excluded. Secondly, the companies who were working on a long-‐ term project basis (for example complete buildings or oil rigs) were excluded as well. This research is focused on manufacturing companies, who produce a variety of components for different industries, which may include project-‐based production companies. In order to make these selections possible, we collected the information available from the Internet pages of each company. The resulting list contained 35 companies. In addition, 8 companies were added, by relying on the researcher’s personal network. While five of these eight companies are not on the KvK-‐list (because they were not listed as ‘metal industry’ companies, rather as for example
‘fabrication’ or ‘machining’ industry), they are active in the metal industry, and
operate in a B2B environment with more than 50 FTE. Therefore they are comparable with the companies selected from the KvK-‐list.
A total of 43 (35+8) companies were contacted by phone (companies in the researcher’s region had highest priority), and 16 of them expressed their interest in the research. Six of the contacted companies directly agreed to make an appointment for the interview. An e-‐mail with more information about the research was send to interested companies who did not directly agree to make an appointment, which was followed by another phone call a few days later. In total, a number of 12 companies agreed to participate in the research, which resulted in a response rate of 27%. It is hard to say if this sample is representative, but it seems that our sample reflects the huge diversity that is listed in our initial KvK-‐list of 192 companies: as can be concluded from the range of size (30-‐400 FTE’s), physical processes, main markets, and number of customers (see Table 1).
Table 1: General characteristics of the sample units
3.2 Interview protocol and data collection
As was mentioned earlier, we relied on semi-‐structured interviews to gather the data. To guide these interviews, a list of open questions has been composed in a protocol (see Appendix I), which consisted of three parts:
these business conditions. Some common aspects are quantitavely reported for each individual unit. In doing so, we can compare the characteristics of the units, since most conditions can be interpreted in a similar fashion. Additional information is mentioned wherever necessary.
• The second part referred to more in-‐depth questions regarding the links with the key buyers and suppliers to measure the level of SCI for each unit. Similar to the approach used by Gimenez et al. (2012), this second part was divided into three separate areas: integration of practices, patterns and attitudes. Although the work of Giminez et al. (2012) is based on survey research, the variables that were used to measure SCI are quite abundant. Therefore, these variables were used to structure and interpret the (rich) gathered data. In order to do so, each variable/aspect was graded with a value of 0, 1 or 2. If the referring integrative aspect is typically applicable at both the company’s key supplier(s) and key buyer(s), it is graded with the value 2. If the referring aspect is typically applicable at either the company’s key supplier(s) or key buyer(s), it is graded with the value 1. If the referring aspect does not typically apply to both the key buyer(s) and key supplier(s), it is graded with the value 0. The total score of each unit regarding SCI is the sum of these values. The upper 4 scores are graded as ‘high’, the middle 4 scores as ‘medium’ and the lower 4 scores as ‘low’.
In grading the different aspects of SCI, we can easily compare the different units. Additional (qualitative) information is mentioned wherever necessary. • Lastly, the third part of the interview protocol is concerned with RL
performance. This paper views RL performance in a multi-‐dimensional manner. In doing so, we are able to apply a broad construct of RL performance, which should provide a better and more detailed understanding of the influence of SCI on RL performance. Therefore, following Daugherty et al. (2002), RL performance can be defined as two distinct dimensions: (1), operating/financial performance; and (2), satisfaction. Except for a few aspects that were specifically designed to measure variables in the reverse chain (e.g. environmental regulatory compliance; recovery of assets), the survey research of Daugherty et al. (2002) shows much resemblance to Gimenez et al. (2012), who also constructed two categories of variables to measure supply chain performance in the forward chain: (1) cost (e.g., transportation costs) and (2)
service (e.g., delivery speed). Following from that, this paper will measure RL
performance based on the existing measures of Gimenez et al. (2012), combined with aspects that specifically apply to RL, which were adopted from Daugherty et al. (2002).
The questions in this section were rather open (appendix A), leaving much room for the interviewees to explain how they think that SCI practices/patterns/attitudes contribute to the performance of their RL operations. In doing so, we found some variables that typically applied to multiple units. As stated, we made a distinction between RL performance in terms of cost and service. The typical, and more specific, variables were grouped in these two groups of measures accordingly.
Both qualitative and quantitative data was collected. This approach enabled us to collect rich data on the one hand, and to compare all cases with regard to a number of quantitative aspects on the other hand. Also, all the questions were asked with respect to both the buyer-‐ and the supplier side of the company’s relationships. In doing so, the level of integration of the whole supply chain was covered. In order to maintain consistency in the resulting data, the first interview was taken by all three researchers. After that, each company was individually interviewed by one of the researchers. The interviews took between one and two hours. The interviews were recorded, and the summaries of the interviews were sent back to the companies for verification. Also, the organizations were occasionally asked for further information.
4 Results
This section reports the main results of our study: the integrative practices, patterns, and attitudes regarding the links with both key buyers and suppliers (table 2), the business conditions of the supplying units (table 3), and the performance of RL practices (table 4). Finally (4.4), the results are combined and weighed against our propositions. In the next section (5), the results are discussed, whilst reflecting on literature.
4.1 Integrative practices, patterns, and attitudes
Table 2 presents the integrative practices, patterns, and attitudes regarding the links with both key buyers and suppliers. The results show that there is a large variation between the units concerning the level of SCI. Recorded examples of integrative practices with regard to the actual physical flows are frequent deliveries (fixed delivery days), packaging customization, and product identification systems. The practices with regard to the information exchange and cooperation differ in terms of joint-‐production planning and logistics activities; a few units involve both their key buyers and suppliers in making their production planning (forecast), whilst some other units do not involve their key suppliers/buyers at all.
Despite that the exchange of information and collaboration with both key buyers and suppliers are seen as important aspects at nearly all units (except for units F and J), the investment in information technology seems to be lacking behind at most units. For example, several units indicated that they do not use the MPS/MRP algorithms provided by their respective ERP systems. Also, nearly all units (except for units D and I) do not have an integrated IT system with any of their key suppliers/buyers.
With respect to the communication between the unit’s key buyers and suppliers, it can be noticed that most units describe it as informal. The contact moments are usually not planned, and mostly flow through the respective sales (in case of the unit’s suppliers) and purchasing (in case of the unit’s buyers) departments. Reported means of information are e-‐mail and telephone. Also, mutual company visits (e.g. every quarter) are common methods to exchange information.
Table 2: Integrative practices, patterns and attitudes
4.2 Business conditions
J that it is impossible for them to come up with any forecast at all, since their demand is extremely uncertain.
Units B, C, D, F and K, show a low level of supply complexity, which is derived from the relatively high sales volumes, low demand uncertainty, large batch sizes and simple operational routings (e.g. number of machines). Higher levels of supply complexity (units A, E, G, H, I, J and L) are derived from the relatively low sales volumes, high demand uncertainty, and complex operational routings.
Table 3: Business conditions
4.3 RL performance
Finally, table 4 presents the performance of RL practices at each unit. Also here, a large variety of RL performance cost and service dimensions have been found. Some units have reported only minimal effects of SCI on RL performance (unit J), whilst others reported quite extensive effects (unit E). It stood out that most units focused on reducing the volume of materials that flow backwards. Certainly, the units try to keep the returns of faulty products or leftover materials to an absolute minimum. However, there were also some units who reported a shared packaging/container system. Here, the reverse flow includes materials such as packaging/container items, where the RL is organized between two participating (buyer/supplier) organizations.
stands out is that many of the unit’s managers indicated that ‘the bigger the
customer/buyer (in percentage of turnover), the higher the RL performance in that particular link is’. This also seems to hold true with respect to the size of the
referring organization: as many of the investigated units have to deal with a relatively large number of small customers and/or buyers, the unit’s managers stated that bigger organizations have better procedures regarding RL in order than smaller organizations (without exceptions).
Table 4: RL performance
4.4 Combined results
A generalization of the combined results is visualized in figure 1. The horizontal axes show the level of SCI, the horizontal axis shows the level of supply complexity, and the size/darkness of the bubble indicates the effect on RL performance (the bigger/darker, the higher the effect). The arrows are placed to help explain the results (see below).
Figure 1: Combined results
Arrow A:
The units that are related to a low level of SCI (e.g. Units F and J) report similar effects to RL performance, namely: lower levels of returned items and more accurate flow of information (less errors). This whilst the units that are related to a high level of SCI (e.g. B, G, E, I), commonly report RL performance effects such as: faster response times, higher quality and more trust. Given these findings, as expected, it can be said that units with higher levels of SCI relate to more dimensions of RL performance and cost. In contrast, a relatively low level of SCI corresponds with a low level of RL performance. The interpretation of these results will be further discussed in section 5.1.
Proposition 1 and 2:
Now we have seen that the results confirm that SCI has a positive influence on RL performance, we are interested whether context (supply complexity) has a moderating role on this relationship. Here, we can see that there are some considerable differences between the units:
• Units E, G, H and I show a relatively high level of supply complexity, as well as a high level of SCI and RL performance. This seems to fit in nicely with our proposition (1), that if supply complexity is high, SCI is an appropriate method to improve RL performance.
• Units D, F and L show a relatively low level of supply complexity, as well as a low level of SCI and RL performance. This seems to fit in nicely with our proposition (2), that if supply complexity is low, SCI is not an appropriate method to improve RL performance.
• Units A and J show a relatively high level of supply complexity, whilst the level of SCI and its influence on RL performance is relatively low. These unexpected results might be explained by looking at the unit’s respective business characteristics (see section 5.2).
• Units B, C and K show a relatively low level of supply complexity, whilst the level of SCI and RL performance is relatively high. Also here, the results might be explained by looking at similarities among the unit’s business characteristics (see section 5.2).
To conclude, the results are rather mixed: whilst some findings confirm our first proposition (1) that when supply complexity is high, a high number of SCI dimensions are associated with RL performance, there are also results that do not unambiguously confirm this effect. The same can be said about our second proposition (2) that when supply complexity is low, there are only a few SCI dimensions that can be associated with RL performance. Also here, there are some findings that confirm this proposition (2), whilst others do not. In section 5.2, we focus on interpreting and discussing these findings, specifically by reflecting at the more qualitative results, as well as literature.
5. Interpretation and discussion
The objective of this paper is to investigate the effect of context on the alleged positive relationship between SCI and RL performance. The results were presented in the previous chapter (tables 1-‐4 and figure 1). This present chapter interprets and discusses the results, whilst reflecting on the theoretical framework (section 2), and additional literature.
5.1 SCI and RL performance
What stood out in the interviews was that units specifically indicated that SCI is equally, if not more, important to performance in the reverse chain than in the forward chain. Especially the impact of integrative practices on the information flow in RL is regarded as crucial for success, since the level of documentation (procedural information) is higher than in the forward chain. Nearly all units (except for unit J) have procedures in place when materials/items are flowing backwards, with the goal of handling (in case of problems) the issue as fast as possible, and to prevent it from happening again. For example by gathering information such as: Is it our mistake?
How did it happen? Etc. Besides the reverse flows that follows from problems (e.g.
quality problems, wrong forecast, missing tolerances), other common examples include joint product development and creating standardized specifications. As was mentioned in several interviews, the bigger the organization (referring unit’s buyers/suppliers), the better this (procedural) information flow is organized. In line with this reasoning, common effects of SCI on RL performance are faster response times and lower levels of returned items. Indeed, as was mentioned by one of the unit’s managers: ‘All information that is flowing backwards has significant value into
taking a step forwards.’
to obtain information about stock levels, sales forecasts and technical drawings. At the units where such systems were in place, the acquired data is validated through contact moments between the referring buyers and suppliers. One important remark however, is that most units who have the possibility to log on to their buyers information systems, do not know how to use or interpret the data.
Although there is not an integrated information system, we have seen that there are many other integrative aspects that do take place. As such, our results confirm that SCI has a positive effect on RL performance (see figure 1, arrow A). Also, it seems that all units have some level of SCI and RL activities. This is in line with the findings of Chan et al. (2010), who argue that every RL system requires at least some level of SCI in order to function. Considering the observations, SCI seems to be rather a prerequisite than an option in order to achieve success in the reverse chain.
5.2 Effect of supply complexity on SCI and RL performance
While the results of most units confirm our propositions, there are a few results that are not directly evident. Therefore, this section focuses on these doubtful findings that were mentioned in 4.4, by interpreting and discussing their meaning.
When looking at units with a relatively high supply complexity, we expected to find a highly positive relationship between SCI and RL performance. Despite that the results of units E, G, H and I fit in nicely with our proposition (1), the results of A and J do not, because they show a weak relationship between SCI and RL performance. Although it is not directly clear from the general business characteristics, which were shown tables 1 and 3, a good explanation might be that the two units both operate in a project-‐like fashion. Namely, almost all sales orders (especially at unit J) are one of a kind, and they both have a relatively large amount of customers. It rarely happens that two sales orders require the exact same products in exactly the same configuration. Therefore, the prompt costs, material requirements, and capacity are highly uncertain, if not impossible, to predict. Given these specific characteristics of units A and J, it comes as no surprise that their respective levels of SCI are low, since the relationship with their buyers (and thus with most of their suppliers) is usually short-‐lived: once the order (project) is finished, the relationship is finished as well. In line with this reasoning, the effect of SCI on RL performance is also low, simply because there is only little integration between buyers and suppliers. Reported effects of integrative efforts on RL performance are lower chance of mistakes (higher quality), and disputes. In the rare case something is flowing backwards, for example faulty products, it is usually resolved through a financial compensation.
till medium, which indicates a relatively stable demand. Therefore, the high levels of SCI and RL performance might be due to the unit’s customer demographics, because the number of customers is relatively small, and a large percentage of the turnover is generated at the largest buyer. Following from that, Germain et al. (2008), found that formalization (which often goes together with standardization) is effective if the variability is low. Given this reasoning, the combination of large volumes with few customers seems to naturally drive the relationship with the unit’s buyers and suppliers to a process oriented approach, where standardization and formalization is related to aspects of RL service and cost performance.
Considering the above reasoning, the interpretation and discussion of the results suggest that both of the propositions that were made in section 2.6 are plausible, except for two groups of units. In the first group, we have found several units where the level of SCI was restricted through the short-‐term nature of the referring units relationships with their buyers and suppliers. The impact of SCI on RL performance is therefore restricted as well. In the second group, we have found several units where a combination of large volumes with few customers naturally drives SCI to a process oriented approach, thus having a positive effect on RL performance.
In conclusion, the findings indicate that the impact of SCI on cost and service aspects of RL performance is higher in case of highly complex supply conditions (proposition 1), and lower in case of low complex supply conditions (proposition 2). However, proposition 1 is not true for units where the relationship with its buyers and suppliers is short-‐lived. Here, high supply complexity is related to a low level of SCI and its impact on RL performance. Conversely, proposition 2 is not true for units which are characterized by high volumes and relatively few customers. Here, low supply complexity is related to a high level of SCI and its impact on RL performance.
6. Conclusions