Retailer Views on Forecasting Collaboration - CiteSeerX

1 downloads 0 Views 39KB Size Report
Sep 9, 2004 - Retailer Views on Forecasting Collaboration. Johanna Småros. Helsinki University of Technology. Alfred Angerer. University of St. Gallen. Prof.
Logistics Research Network Annual Conference, September 9 – 10, 2004, Dublin, Ireland

Retailer Views on Forecasting Collaboration Johanna Småros Helsinki University of Technology Alfred Angerer University of St. Gallen Prof. John Fernie Heriot-Watt University Dr. Beril Toktay INSEAD Dr. Giulio Zotteri Politecnico di Torino

Abstract Forecasting collaboration between retailers and suppliers is suggested to bring significant benefits in the form of increased efficiency and improved customer service. Yet, only few companies seem to be engaged in collaborative relationships. Even in the grocery sector, which is one of the most active promoters of supply chain integration efforts such as Efficient Consumer Response (ECR) and Collaborative Planning, Forecasting and Replenishment (CPFR), companies seem to have been slow to embrace collaborative forecasting. In this paper, data collected through in-depth interviews with twelve leading European grocery retailers is used to examine three hypotheses suggested to explain the slow adoption rate of collaborative forecasting in the European grocery sector. The data is found to support the proposition that retailers’ lack of forecasting capabilities is a more important obstacle to forecasting collaboration than the required investments in information technology. Evidence on the different forecasting needs of retailers and suppliers is also found. Some additionals elements of forecasting collaboration are also identified. Keywords: Supply chain management, collaborative forecasting, CPFR, grocery retailing.

Introduction Recently, as a result of the introduction of the Collaborative Planning, Forecasting and Replenishment (CPFR) initiative (Figure 1), collaborative forecasting has received much attention (Aviv, 2001; McCarthy and Golicic, 2002). Developing forecasts in a collaborative fashion and operating according to a single, shared forecast is suggested to bring significant benefits to supply chain members in the form of increased forecast quality, improved communication and, through this, improved product availability, increased sales, and reduced inventory (Aviv, 2001; Helms et al., 2000; Sherman, 1998; McCarthy and Golicic 2002). Yet, despite the suggested benefits of collaborative forecasting, the adoption rate of CPFR has been slower than expected (KJR Consulting, 2002; Lewis, 2000). Lately, the enthusiasm initially sparked by outstanding results from pilot implementations by well-known companies, such as Wal-Mart and Procter & Gamble, seems to have been replaced by increasing skepticism (Corsten, 2003; Sliwa, 2002). The trade press presents an abundance of potential obstacles to forecasting collaboration and CPFR. Fliedner (2003) has examined several reports and trade journals. He presents the following list of potential barriers: lack of trust in sharing sensitive information, lack of internal forecast collaboration, availability and cost of technology or expertise, aggregation concerns (number of forecasts and frequency of generation), and fear of collusion.

1

Logistics Research Network Annual Conference, September 9 – 10, 2004, Dublin, Ireland

Develop front-end agreement Create joint business plan Create sales forecast Identify sales forecast exceptions Collaborate on sales forecast exceptions Create order forecast Identify order forecast exceptions Collaborate on order forecast exceptions Generate order

Figure 1. The nine steps of the CPFR process model. More rigorous research examining the inconsistency between theory and practice is, however, scarce. Barratt and Oliveira (2001) use a survey approach to examine the inhibitors and enablers of CPFR. They identify lack of discipline to properly execute the preparatory phases of the CPFR process model, lack of joint planning, difficulties in managing the exceptions and review processes related to sales and order forecasting, and lack of shared targets as significant inhibitors. McCarthy and Golicic (2002) present a case study approach on forecasting collaboration and criticize the CPFR process model for being too detailed and comprehensive. They recommend that companies develop collaboration practices that require less investment in human or technological resources than the CPFR process model does. Småros (2003) also examines forecasting collaboration using a case study approach and concludes that streamlined collaboration approaches are needed since retailers often seem to lack forecasting resources and processes. The most explicit attempt to explain the slow adoption rate of collaborative forecasting approach is presented in a recent paper by Småros (2004). Based on a case study of four retailer–supplier collaboration projects in the European grocery sector, four hypotheses concerning the attractiveness and feasibility of collaborative forecasting are presented: § Hypothesis 1: The technology investments required for large-scale collaboration slow down adoption of collaborative practices, but do not present a critical obstacle to forecasting collaboration. § Hypothesis 2: Retailers’ limited forecasting resources and lack of forecasting processes present a critical obstacle to CPFR-style forecasting collaboration, but not to more streamlined collaboration practices. § Hypothesis 3: Due to different replenishment lead-times and aggregation levels, retailers and suppliers have different forecasting and collaboration needs. § Hypothesis 4: Long lead-times and lack of internal integration make it difficult for suppliers to efficiently use demand and forecast information obtained through collaborative relationships in their operations. In this paper, we use data collected from in-depth interviews with leading European grocery retailers to test the first three of these hypotheses. The paper is organized as follows: First, the research method is presented. Next, the hypotheses are examined in light of the collected data. Finally, conclusions and suggestions for further research are presented. Methodology The research approach is hypothesis testing. Usually, surveys with many respondents are used in hypothesis-testing research in order to attain sufficient generalizability. However, in this case it was felt that due to the detailed nature of the data needed as well as the lack of standardized terminology, in-depth interviews would be produce more reliable information.

2

Logistics Research Network Annual Conference, September 9 – 10, 2004, Dublin, Ireland

Sample The sample of companies examined consists of twelve European grocery retailers. When contacting companies, the aim was to include leading retailers operating in different markets. The companies included in the study represent six regions: the Nordic countries, the UK, Northern Continental Europe, Western Continental Europe, Central Continental Europe, and Southern Continental Europe. All of the companies are major players within the grocery sector. In fact, only two of the twelve companies are outside of the top three (measured in market share) in their respective target markets. The companies form a rather heterogeneous group with turnovers ranging from approximately 1 billion euros to over 20 billion euros. Most of the companies have several retail chains or store formats, ranging from small neighborhood stores to very large hypermarkets. In addition, the companies differ in private label penetration. Furthermore, the companies differ in their views on forecasting collaboration: three of the companies are currently involved in large-scale forecasting collaboration, one has established permanent collaboration processes with a few suppliers, five of the companies are or have been involved in collaboration pilots, and four of the companies did not mention being involved in any collaborative forecasting initiatives at all. Data collection Data was collected through in-depth structured interviews with between one and three persons from each of the companies. Some of the respondents were interviewed several times. The interviewees were mainly directors and managers in charge of logistics and supply chain management, development, category management, or information technology. The interview questionnaire was designed to solicit and capture information regarding key performance indicators, the companies’ logistics processes, supplier collaboration, and logisticsrelated challenges and opportunities. The questionnaire was first tested on one of the respondents. After this, it was slightly revised. Based on the interviews, company specific case reports were compiled. These reports were checked internally by the members of the research group as well as externally by the interviewees. The data collection effort lasted from August 2003 to June 2004. Results H1: The technology investments required for large-scale collaboration slow down adoption of collaborative practices, but do not present a critical obstacle to forecasting collaboration We start the examination of Hypothesis 1 by looking at the three large-scale implementations of forecasting collaboration encountered during the research. Two of these implementations focus on promotions. In one of the cases, the process is based on the retailer’s central organization developing an initial system-generated forecast and then checking it with stores and suppliers. The other implementation focuses on retailer-supplier collaboration and includes more information technology (IT) support, such as alerts (accessible to both retailer and supplier) when demand is developing differently than predicted. The third large-scale collaboration is a supplier-driven version of the CPFR process model in which the retailer generates forecasts based on its history data and the suppliers view these forecasts in order to pick up trends and identify exceptions compared with their own forecasts. The identified exceptions are then either handled automatically or manually by the retailer. Two observations can be made based on this information. Firstly, large-scale forecasting collaboration is possible without special technology. Secondly, companies have been able to provide the IT solutions required for supporting more complex collaboration processes. When discussing collaborative forecasting with the retailers, technology issues rarely came up. The following observations concerning the role of IT were made: § “Sure, IT helps, but it’s the time investment that is heavy.” § “We constantly make and plan IT investments to improve our own processes. Although many of these investments are important for future collaboration, they are primarily made in order to improve our own efficiency.” § “We need to get our internal systems in order first.” § “We are currently evaluating different IT tools to make collaboration more efficient.”

3

Logistics Research Network Annual Conference, September 9 – 10, 2004, Dublin, Ireland

The fact that the retailers, even the ones that are involved in or have piloted retailer-supplier forecasting collaboration, had so little to say about IT (except for IT solutions related to their own internal processes) when discussing forecasting collaboration is rather surprising. This observation, in combination with the two previous ones, supports the hypothesis that IT does not provide the main, or even a very critical obstacle to forecasting collaboration. H2: Retailers’ limited forecasting resources and lack of forecasting processes present a critical obstacle to CPFR-style forecasting collaboration, but not to more streamlined collaboration practices In order to examine Hypothesis 2, we again return to the three large-scale implementations identified. In only one of the cases, the collaboration process employed is similar to the CPFR process model. This collaboration is, however, rather supplier-driven with the retailer providing system-generated forecasts for suppliers to examine and compare with their own forecasts. When looking at the retailer’s forecasting capabilities, it can be noticed that it has dedicated forecasting resources and rather sophisticated tools for forecasting. Th e other two large-scale implementations focus on promotions. Both of the retailers involved in this kind of collaboration use rather sophisticated forecasting tools. In addition, one of the retailers has dedicated forecasting resources. The other retailers in the sample are different from these three companies. The companies employ rather basic forecasting tools (except for one company that is currently implementing a more powerful system). In addition, none of them have dedicated forecasting resources. Three observations can be made based on this information. Firstly, the only retailer involved in CPFRlike collaboration has dedicated forecasting resources. Secondly, collaboration tends to be focused on promotions and more streamlined, i.e. containing less steps and tasks, than the CPFR process model. Thirdly, all of the companies involved in large-scale collaboration have fairly sophisticated forecasting tools. When examining the companies’ willingness to engage in other kinds of collaboration, it can be noticed that seven of the companies interviewed are involved in replenishment collaboration with their suppliers. In addition, four companies regularly exchange sales data with suppliers. This indicates that companies that are not involved in collaborative forecasting are still able to collaborate in other ways, especially by giving suppliers access to demand data. Finally, we examine the retailers’ comments on forecasting collaboration in general and the CPFR process model in particular (the following list contains topics that were mentioned by at least two of the retailers): § Many retailers mentioned that the CPFR process model in its suggested form requires too much work and resources, and, therefore, is unfeasible, § Some retailers pointed out that the process model should be seen as a conceptual framework rather than an actual process to be implemented. When seen as a framework, CPFR is considered valuable as its highlights the right things and works as a change initiative, § Many companies are interested in collaborating with their suppliers, in one way or another, but some retailers are skeptical about the value of collaboration, § Several companies stated that the retailers need to develop their own forecasting and replenishment processes before real collaboration can take place. Based on the interviews, it can be concluded that most retailers do, indeed, lack dedicated forecasting resources and that their forecasting tools often are rather simple. The retailers involved in large-scale collaboration are all exceptional in that they have more sophisticated forecasting tools than the other retailers in the sample. In some cases they also have dedicated forecasting personnel. These observations support the notion that intensive CPFR-style collaboration requires forecasting capabilities that most retailers, even leading ones, currently do not have. An interesting observation that is slightly different from the original hypothesis is the emphasis on retailer forecasting tools and data (for generating forecasts) rather than on explicit forecasting resources (for reviewing and adjusting the forecasts). This is probably the result of the implementations found in the companies generally being rather streamlined and more focused on

4

Logistics Research Network Annual Conference, September 9 – 10, 2004, Dublin, Ireland

information sharing (sharing of system-generated forecasts) than on actually developing forecasts in co-operation with suppliers. The results also support the second part of the hypothesis by demonstrating that also companies that lack advanced forecasting processes, tools, and resources, can engage in more streamlined collaboration, focusing on information sharing. H3: Due to different replenishment lead-times and aggregation levels, retailers and suppliers have different forecasting and collaboration needs To examine Hypothesis 3 we look at some key performance indicators: § Order-to-delivery lead-times between manufacturers and retail distribution centers: The reported average order-t o-delivery cycles range from about 14 hours to 48 hours with a median value of 20 hours for the companies in the sample, § Order-to-delivery lead-times between retail distribution centers and stores: The median value of the reported average lead-times is 32 hours, with reported lead-times ranging from about 14 hours to over 32 hours, § Service level from manufacturers to retail distribution centers: The reported average service levels range from about 95% to 98%, with a median value of 98% for the companies in the sample, § Service level from retail distribution centers to stores: The median value of the reported average service levels is 98%, with reported service levels ranging from 97% to 99%. These figures verify that lead-times are, indeed, short and service levels high, both for companies who do collaborate and for companies who do not collaborate. Due to the short lead-times and high service levels, the retailers do not have the same forecasting needs as suppliers, especially on an aggregate level, just as suggested by Hypothesis 3. Rather than investing in attaining a very high forecast accuracy, the retailers can, in many cases, rely on the responsiveness of their supply chains. One retailer even commented that it only considers forecast errors that are larger than +/- 50%. If the error is smaller than that, the retailer is able to cope with it. Interestingly, this retailer is one of the companies involved in forecasting collaboration with suppliers. When digging deeper into the data, a potential reason that may explain why some companies invest so much in forecasting and forecasting collaboration despite the already high service levels and short lead-times emerges. There is some indication that the companies that have the most developed automatic store ordering systems (i.e. systems that cope or are being developed to cope with not only stable demand but also events, such as promotions and important seasons) and are following a centralized approach to managing store replenishment (i.e. replenishment methods and parameters are set by the retailer’s central organization) have invested or are investing in forecasting tools. These companies also seem to be most interested in forecasting collaboration. Conclusions The data collected from interviews with leading European grocery retailers was found to support all three hypotheses. In addition, some new aspects on retailer-supplier collaboration were identified. Firstly, IT does not, indeed, appear to be a critical obstacle to forecasting collaboration. Based on the retailers’ comments and the fact that one of the three large-scale implementations of forecasting collaboration has been set up with little IT support, it seems that the biggest technological challenges do not lie in the collaboration interface itself, i.e. in exchanging data or communicating with suppliers, but rather within the retailers’ internal processes. Secondly, the data also supports the observation that the resource-intensive nature of CPFR-like collaboration is a major obstacle hindering retailers from engaging in this sort of collaboration. In fact, only one collaboration arrangement resembling the CPFR process was encountered in the sample of companies examined, and this collaboration had been implemented by a retailer with dedicated forecasting resources – a rarity amongst the companies interviewed. Furthermore, supporting the other part of the hypothesis, more streamlined collaboration, such as exchange of sales data, is something almost all retailers are capable of.

5

Logistics Research Network Annual Conference, September 9 – 10, 2004, Dublin, Ireland

An interesting observation in this context is the role of retailer forecasting systems in supporting collaboration. It seems that the access to tools generating high-quality forecasts may be an even more critical component of forecasting collaboration than retailer forecasting resources, especially when the focus of the collaboration is on exchange of forecast information rather than on joint forecasting. Finally, data on key performance indicators collected from the companies supports the hypothesis that retailers, owing to short lead-times and high service levels, have different forecasting and collaboration needs than suppliers. Retailers can in many cases rely on the responsiveness of their supply chains, reducing the need for accurate forecasts. However, also in this context an interesting observation was made. There is some indication that the implementation of centrally controlled automatic store ordering may provide retailers with an incentive to invest in forecasting and forecasting collaboration, even when lead-times are short and service levels high. Discussion and further research The study identified some new viewpoints on the tested hypotheses. These need to be further examined. Especially in the grocery sector, the role of forecasting tools as enablers of collaboration merits further research. Another interesting research question is why some companies invest more in forecasting tools and resources than other companies, despite their seemingly similar business environments - do aspects such as private label penetration or automatic store ordering explain these differences? Research focusing on different industries is also needed. The collaboration practices employed in different sectors should be examined and compared to those found in the grocery sector. In addition, more general concepts, such as how decentralized versus centralized control in companies affect supply chain management and inter-company collaboration, provide interesting research subjects. Based on the study, it seems that centralized control provides better support and also increased need for collaboration efforts compared with decentralized control. In addition, the power aspect is interesting. It seems that the power of European grocery retailers’ in their supply chains makes it possible for them to attain good service even without collaborating. References • • • • • • • • • • • •

Aviv, Y. (2001), “The effect of collaborative forecasting on supply chain performance”, Management Science, Vol. 47, No.10, pp. 1326-1343. Barratt, M. & Oliveira, A. (2001), “Exploring the Experiences of Collaborative Planning Initiatives”, International Journal of Physical Distribution & Logistics Management, Vol. 31, No. 4, pp. 266289. Corsten, D. (2003), “CPFR – The silver bullet?”, Presentation delivered at the CIES Supply Chain Conference 2003 in Nice. Fliedner, G. (2003), “CPFR: an emerging supply chain tool”, Industrial Management & Data Systems, Vol. 103, No. 1, pp. 14-21. Helms, M.M., Ettkin, L.P. & Chapman, S. (2000), “Supply chain forecasting – Collaborative forecasting supports supply chain management”, Business Process Management Journal, Vol. 6, No. 5, pp. 392-407. KJR Consulting (2002), CPFR baseline study – manufacturer profile, Grocery Manufacturers of America. Lewis, L. (2000), “CPFR solutions”, Progressive Grocer, Vol. 79, No. 4, pp. 28-32. McCarthy, T. & Golicic, S. (2002), “Implementing collaborative forecasting to improve supply chain performance”, International Journal of Physical Distribution & Logistics Management, Vol. 32, No. 6, pp. 431-454. Sherman, R. (1998), “Collaborative planning, forecasting & replenishment (CPFR): Realizing the promise of efficient consumer response through collaborative technology”, Journal of Marketing Theory and Practice, Vol. 6, No. 4, pp. 6-9. Sliwa, C. (2002), “CPFR clamor persists, but adoption remains slow”, Computerworld, Vol. 36, No. 27, p. 10. Småros, J. (2003), “Collaborative forecasting: A selection of practical approaches”, International Journal of Logistics: Research and Applications, Vol. 6, No. 4, pp. 245-258. Småros, J. (2004), “Forecasting collaboration in the European grocery sector: Observations and hypotheses”, Working paper, Helsinki University of Technology.

6