Automatic replenishment programs: An empirical examinationAccurate demand projection, i.e., good forecasting, has seemed almost a fantasy in recent years. Shorter product development cycles, product proliferation, and global-based competition have made it more difficult to successfully manage demand.' Too often product availability and order quantities fail to "match up" resulting in problems including excessive inventory investment, poor customer service, lost sales opportunities, and inefficient production scheduling.2 Thus, many firms have sought better ways of managing inventory replenishment.New approaches to inventory management aimed at improving system-wide efficiency have become popular recently. A number of different terms are used to designate such programs; however, we will use the umbrella term automatic replenishment programs (ARPs). Automatic replenishment is used to signify an exchange relationship in which the seller replenishes or restocks inventory based upon actual product usage and stock level information provided by the buyer. Popular automatic replenishment programs include continuous replenishment planning (CRP) and vendor managed inventory (VMI). CRP and VMI are very similar except that with VMI the vendor assumes responsibility for deciding what and when to ship? Industry-specific automatic replenishment programs have also been developed. For example, efficient consumer response (ECR) is widely used within the grocery industry and quick response (QR) programs are common in the apparel industry.Automatic replenishment programs have received considerable attention in popular-press and trade publications.4 Most of the accounts are anecdotal, providing company-specific examples. Little empirical work has appeared in academic publications. One recent academic piece reported on the use of a simulation model to examine VMI. The authors found "compelling economic and operational benefits" associated with VMI.5 Thus, the current exploratory research was undertaken in order to empirically examine automatic replenishment involvement. BACKGROUND The intent of the automatic replenishment programs is to make inventory commitment more efficient. Of course, this must be accomplished without sacrificing sales or customer relations. Product availability must be balanced against carrying unnecessarily high stock levels. The solution is to substitute information for inventory. Actual sales figures and buyer-supplied information triggers replenishment. This requires close cooperation between retailers and manufacturers. The trading partners must trust one another and be willing to share proprietary/confidential information.6 The following quote succinctly describes VMI, one of the most common types of automatic replenishment currently in use. ... suppliers assume the responsibility for managing inventories at customer locations through the use of highly automated electronic messaging systems. Detailed sales and demand data are exchanged between vendors and customers, and the information is used to plan and implement product replenishment and sales strategies.7 Thus, the basic components or building blocks of an automatic replenishment program include decision support systems, product identification technology (bar coding), and electronic data interchange (EDI). The components should be considered enablers, but not an absolute requirement for ARP-type programs8 For example, if only a low number of stock-keeping units are being managed, all three may not be required. However, many automatic replenishment systems incorporate all three items. Involvement in automatic replenishment is expected to yield a range of benefits including increased sales per store for retailers due to more efficient stocking and more frequent deliveries, higher selling space productivity because of fewer stock-outs, and a need for less "back room" storage space for safety stock. Safety stock, if not totally eliminated, can usually be dramatically reduced.' Manufacturers benefit from lower manufacturing and distribution costs through better matching of supply to demand and improved production planning." In reporting on a nine week VMI case study of 18 VF Corporation stores, Nannery identified significant reductions in inventory per store, increased inventory turns, and substantial increases with respect to in-stock percentages. However, he also noted that such programs are "by no means painless for vendors. " Retailer inventories may be reduced, but manufacturer warehouse inventories actually increased, at least at first.11 SCOPE OF STUDY AND RESEARCH HYPOTHESESS Little empirical research appears to have been published examining automatic replenishment programs. The limited information that was found focuses on case-type studies, mostly in the grocery industry.12 One exception that was identified is Fiorito, May, and Straughn's (1995) survey of quick response among retailers.13 While they did not directly measure QR implementation, they did assess the retailers' utilization of electronic data interchange (EDI), automatic replenishment systems, and barcodes. The current research provides a profile of automatic replenishment involvement by type of program (VMI, ECR, etc.). The research also examines implementation status with respect to specific program components such as EDI and barcoding, information systems capabilities of the respondent firms, and program success, to date, in terms of achieving specific automatic replenishment-related goals. In addition to providing an overview of ARP involvement, the research also examines specific research hypotheses relating to ARPs. Figure I provides a conceptual framework for investigating the relationships between factors associated with automatic replenishment programs. In this study, a multi-dimensional perspective is taken regarding the nature of ARP implementation, information systems capabilities, and ARP performance constructs. The operationalization of these constructs is discussed later in the paper. Automatic replenishment programs are information intensive. The underlying philosophy is that effective use of information can replace extensive commitment to inventory holdings." In today's business environment, inventory re-order decisions are generally made frequently and for small amounts, and poor decisions are likely to result in out-of-stocks. The replenishment systems don't typically allow for substantial buffers of safety stock. Accurate and timely information is critical to avoid out-of-stock problems. Commitment to automatic replenishment, as evidenced by implementation of various program components such as Electronic Data Interchange, point-of-sale scanning, and joint forecasting, generates valuable market-related data. For example, with automatic replenishment, firms focus on working with trading partners to develop cross-firm demand projections, generally do more planning, and can precisely track actual sales. Customer demands are subject to change, but the automatic replenishment systems capture those variations or "hiccups. " 15 Thus, firms with greater ARP involvement have more support data to guide decision-making. Even more importantly, it is likely that automatic replenishment involvement acts as a driver for expanding information systems capabilities. As firms get more involved in automatic replenishment, i.e., add more implementation components or expand the program scope, they will expand and up-date information systems capabilities to provide the necessary support. Therefore, we hypothesize: H1: Information systems capabilities are positively influenced by automatic replenishment program implementation. As was stated previously, the intent of ARPs is to make inventory commitment more efficient through precise planning and replenishment practices. Typical ARP goals include reduced costs, faster inventory turns, and increased/improved customer service. Review of the literature, particularly Fiorito, May, and Straughn's 1995 article, indicates automatic replenishment programs are complex and incorporate a wide range of building block-type components. Certain components, i.e., barcoding and scanning capability, are more essential to automatic replenishment success. Other items, such as advance ship notices, are not essential to ARP involvement, but offer attractive customer-valued add-ons. Automatic replenishment program development can be very resource-intensive. However, as the range of ARP components utilized or implemented is increased, program effectiveness should be improved. For example, cross-docking operations are not essential to offering automatic replenishment, but the use of cross-docking can improve speed and efficiency of inventory replenishment. Further, because inventory management is generally so dependent upon information, the automatic replenishment program effectiveness/success is also linked to information systems capabilities.16 Timely, readily usable information is needed to support quick decisions associated with automatic replenishment." Automatic replenishment mandates that firms be responsive to market conditions. Inventory levels must be precisely tracked at the SKU level and the information subsequently used to trigger re-shipments. Therefore: H2: Automatic replenishment program effectiveness is positively influenced by: a. automatic replenishment program implementation and b. information systems capabilities. Finally, automatic replenishment program effectiveness should influence overall ARP performance. Firms are being pushed to greater productivity, particularly in terms of faster order cycles." Customers want faster deliveries and more tailored shipments in terms of small size orders of exact product mixes." ARPs offer the potential of developing a satisfied customer base; customers can rely upon products to be in stock as needed. As Cottrill (1997) notes, the programs ensure that the right products reach the shelves at the right time and at lower cost and "thus boost sales and profits."20 Thus: H3: ARP performance is positively influenced by automatic replenishment program effectiveness. RESEARCH METHODOLOGY Based upon review of the literature and interviews with five logistics executives (three manufacturers and two retailers), an automatic replenishment-related survey was developed. The instrument was subsequently pre-tested with six people (three logistics executives, two consultants, and one academic). Using their input, the survey was modified and expanded. Two versions (only differing slightly) were developed-one for manufacturers and one for retailers. A list of manufacturer and retailer members was secured from the Council of Logistics Management and telephone calls were made to a random sampling. The intent was to screen potential participants by ascertaining whether their firms were involved in any type of automatic inventory replenishment. Of the 762 contacts made or attempted, 247 (32.4%) were ineligible because their firms were not utilizing automatic replenishment or had not fully implemented/ operationalized the systems. Of the remaining 515 contacts, 24 refused to participate in the survey, 209 did not respond after repeated phone messages, and 282 gave permission for mailing the survey. Surveys were mailed to the 282 individuals and reminder cards were sent two weeks later. A total of 104 surveys were returned. Six had excessive missing values resulting in 98 usable surveys. Of the 98 usable surveys, 75 are manufacturing firms and 23 retailing firms. The respondent firms' annual sales volume averaged $2.3 billion. The respondent base represents a wide range of industries. The most highly represented industries include: food and beverage (27. 1 %), electronics (12.6%), apparel (9.7%), chemicals (9.7%), miscellaneous manufacturing (6.3%), pharmaceuticals (4.2%), and medical equipment (4.2%). As recommended by Armstrong and Overton (1977), the potential for non-response bias was assessed by comparing early versus late respondents .21 The last quartile of responses, assumed to be most similar to non-respondents, was compared to the first three quartiles of responses. Comparisons of group means on individual survey questions revealed no significant differences for the variables analyzed. Thus, non-response bias is not considered to be an issue with respect to the study results. RESULTS AND DISCUSSIONThe research findings provide a profile of automatic replenishment program involvement and the results or benefits that the respondent firms have realized to date. As shown in Table 1, the most common type of automatic replenishment program is vendor managed inventory (VMI). Nearly one-half of the respondents firms are utilizing VMI. A high level (close to 40% of respondents) of involvement is also noted in continuous replenishment programs (CRP). Lower levels of involvement are noted in the other types of automatic replenishment-supplier-managed inventory, quick response, jointly managed inventory, efficient consumer response, and distributormanaged inventory. However, these exploratory findings may be a function of the sample and are not necessarily generalizable. Slightly over one-third of the respondents indicated their firms are involved in more than one type of automatic replenishment. Automatic Replenishment Program Implementation Fiorito, May, and Straughn (1995) identified 15 quick response-related components or elements that are commonly associated with QR.22 The authors used those items along with input provided during the interview phase of the research to develop a modified list comprised of 17 automatic replenishment-related elements. The survey respondents were asked to indicate their firms' implementation status for each of the 17 items. A 7-point scale was utilized with I = not implemented, 4 = somewhat implemented, and 7 = fully implemented. Results are presented in Table 2. The top five components, in terms of degree of implementation, represent the basic building block/hardware-type components (EDI, barcoding on shipping container labels, and barcoding at the retail level) as well as usage of cross-functional teams and adoption of the automatic replenishment system for basic goods. The respondents have emphasized hardware items, developing a team approach to coordinate various functional areas to support the automatic replenishment, and coverage of basic goods. Developing necessary capabilities and determining what the programs should cover are the logical starting points for automatic replenishment involvement. The mean scores for the first five items range from 5.04 to 5.47 (with 7 indicating full implementation). It is not surprising to find that the respondent firms are not reporting full implementation. Automatic replenishment-type programs have only become popular and in wider use within the last few years; full implementation and subsequent realization of a wider array of benefits will take time. Among the respondent firms, the average length of time automatic replenishment has been utilized is 4.63 years (standard deviation 2.51). The next highest implementation levels are noted for pre-season planning with trading partners (4.84) and automatic forecasting for staple goods (4.83). The underlying philosophy of automatic replenishment programs is sharing of information in order to reduce resource commitment and improve efficiency/service. The respondent firms have placed relatively high levels of importance on working with trading partners to develop seasonal plans in advance and to automate those areas that are generally most manageable or easiest to project, e.g., staple goods. Less emphasis has been placed on implementing the next five items - Advance Ship Notices (ASNs), electronic payments,joint planning of replenishment/promotion, direct store delivery of products, and cross-docking operations (means range from 4.14 to 4.49). Joint or collaborative planning has just begun to receive more attention. Supply chain managers are realizing that trading partner collaboration extending beyond one company's boundaries has the potential for significant system-wide gains. More emphasis is being placed upon developing closer ties between trading partners supported with advanced technology and information. Inventory management is easier because of the close alignment. The other four items represent "add-ons" or additional services that can complement automatic replenishment programs, but are not absolutely required. Utilization of Advance Ship Notices provide buyers with exact information as to what they will receive and when. If shortages are involved, the buyers have additional time to make other arrangements as necessary. Electronic payments are another "extra" that helps to build closer relationships. Electronic payments are faster and generally more accurate than traditional payment methods. Both direct-store delivery and cross-docking operations can result in faster deliveries and eliminate or reduce the need for buyers to warehouse or temporarily store product. Joint forecasting, electronic point-of-scale scanners, and automatic forecasting for fashion/seasonal goods are less common (range 4.03 to 4.08). While the respondents appear to be somewhat involved in joint planning for replenishment and promotion, they are slightly less involved in joint forecasting with trading partners. This may be because of the complexities involved or because of a reluctance to share such proprietary information. The lower implementation of electronic point-ofsale scanning may be due to the costs involved and/or may be industry-specific, i.e., some types of businesses/products are less likely to use POS scanning. The lower level of automatic forecasting for fashion/seasonal goods is not unexpected. Fashion items and seasonal products are difficult (and risky) to project in advance. Inaccurate advance forecasts can be costly in terms of excessive product because of poor assortments, lost sales due to out-of-stocks, etc. Whereas staple type items, with predictable sales patterns, are more appropriate for automatic replenishment and longer-range projections, fashion items are not as good of a fit with the systems. The lowest levels of implementation are noted for activity-based costing (3.55) and vendor-marked merchandise or pre-ticketing (3.21). Once again, these are components not essential to basic automatic replenishment, but representing valuable support areas. Activity-based costing can be utilized to secure system improvements and identify problem areas. Vendor-marked merchandise is highly valued by buyers and considered an important value-added service dimension, but is also often a service request that sellers find time/labor intensive and expensive to provide. Information Systems Capabilities Respondents were asked to indicate their firms' information systems capabilities with respect to nine items. The nine items were selected based upon information provided during interviews with practitioners. A 7-point scale with I = not capable, 4 = somewhat capable, and 7 = highly capable was used. Results are presented in Table 3. The respondents indicated their firms are doing best in terms of being able to provide daily download of information (mean = 5.70). They are doing relatively well at getting daily access to information needed and are only slightly less capable in terms of accuracy of information (5.33) and timeliness of information (5.16). Overall information systems capabilities are also fairly strong in terms of being able to format information to facilitate usage (5.07), availability of information (5.01), internal connectivity/compatibility (4.88), and formatting information as needed on an exception basis (4.65). The emphasis is on providing customized information support - easy to use, non-standard reporting as needed that can also facilitate internal coordination or linkages across functional areas. Information systems capabilities are only slightly lower in terms of real-time information capabilities (4.59) and external connectivity/compatibility (4.58). Greater complexity is involved in providing instantaneous access to information as well as in developing cross-boundary information systems to allow coordination with trading partners. Automatic Replenishment Program Success In order to assess how well automatic replenishment programs are performing, the respondents were provided with a list of automatic replenishment-related goals (see Table 4). The list was initially developed based upon review of the literature and subsequently refined, utilizing input received during initial interviews and the pre-test phase of the research. A total of 11 items were included and assessed on a 7-point scale with I = not at all effective, 4 = somewhat effective, and 7 = extremely effective. Automatic replenishment programs have been very effective in achieving some of the most basic program-related goals. The respondents indicated their programs had been highly successful in terms of improving/increasing customer service levels (5.47), fewer stock-outs (5.33), improved reliability of deliveries (5.15), and faster inventory turns (4.94). The respondents reported only slightly less program success related to achieving reductions in key business areas. They are doing moderately well in terms of reduced over-stocks (4.79), inventory (4.77), returns and refusals (4.62), handling (4.56), costs (4.50), and product damage (4.46). The programs are effective at achieving stated objectives of better matching supply and demand-and in a cost-effective, quality conscious manner. The lowest level of program success is associated with reducing the need to discount product (3.96). Automatic replenishment attempts to exactly match supply and demand. Even with careful monitoring,joint planning, and other initiatives aimed at predicting demand, mismatches will occur due to market conditions, changing consumer preferences, etc. Frequently the decision is made to markdown products rather than carrying them in inventory or returning product to vendors. Research Hypotheses The dependent variables in this study are information systems capabilities, ARP effectiveness, and overall ARP performance. Therefore, the unit of analysis is the established automatic replenishment program of the firm. To test the operational model presented in Figure 1, measurement scales were either adapted from previous work or developed as necessary. While performance scales are easily adapted from the literature, scale development was particularly critical within the dimensions of ARP implementation, information systems capabilities, and ARP effectiveness, since representative constructs had yet to be established. In order to assist in this development, Exploratory Factor Analysis (EFA) with principle component extraction and varimax rotation was applied to each category of these explanatory variables. Understanding that a "blind" EFA can result in factors that lack substantive meanings and are inappropriate for construct development, a thorough analysis of the substantive meanings of the factors and corresponding items was performed to purify the factors. Following Kerlinger's (1986) benchmark, all factors with an eigenvalue greater than one were included in the study.23 Results of the EFA are shown in Table 5. All scales exceed Nunnally's (1978) suggested reliability minimum of .60. Complete scales and reliability estimates are provided in Table 6.24 H1 investigates the issue of whether information systems capabilities are positively influenced by automatic replenishment program implementation. As firms commit to and expand ARPs by adding additional elements, they are likely to expand information systems capabilities as well. As detailed previously, automatic replenishment program component implementation was measured utilizing 17 items (see Table 2) and information systems capabilities were gauged in nine areas (see Table 3). The factor analysis of the 17 implementation scale items produced three dimensions. The implementation items loaded along the following dimensions-forecasting/planning (planning issues associated with set-up and managing automatic replenishment including items such as automatic forecasting, pre-season planning with trading partners, and joint forecasting); barcoding/scanning (incorporating barcoding at retail level, scanning, etc., i.e., the mechanics necessary to physically process and track inventory); and delivery (the follow-through required to efficiently move product, i.e., barcoding on shipping container labels, cross-docking, direct store delivery, etc). Thus, the three implementation dimensions represent a temporal or phased perspective of ARPs: plan for inventory deployment; take care of mechanics necessary to move the product smoothly; and then complete the cycle by actually making sure delivery is made. The three implementation dimensions were used for analyses relating to HI. Also, the nine information systems capabilities items were collapsed into two dimensions: information timeliness and information compatibility. Information timeliness incorporates availability and accuracy in a timely manner; information compatibility focuses on formatting and ease of connectivity issues. Based upon review of the literature, it was believed that information systems capabilities will be positively influenced by the implementation of automatic replenishment program components (tested with respect to the implementation of barcoding/scanning, forecasting/planning, and delivery components). While the results indicate that both forecasting/planning and delivery implementation significantly influence information timeliness (beta =.297, p . 10). Barcoding has traditionally been thought to decrease the time in which it takes information to flow between buyer and seller; however, die complexities involved and other system compatibility issues may explain this finding. Interestingly, it is the automatic replenishment-related components involving joint decisions and close interaction (forecasting/planning and delivery) that influence information timeliness, highlighting the critical nature of synergy between firms. Further examination of the issue centered on whether information compatibility is positively influenced by the implementation of the three automatic replenishment program components. Again, the relationship was supported for forecasting/planning (beta = .418, p H2 addresses the issue of whether automatic replenishment program effectiveness is positively influenced by ARP implementation level and by information systems capabilities. Eleven items (see Table 4) were used to assess respondents' ARP effectiveness. For analysis purposes, the eleven items were collapsed into two dimensions: cost-related (reductions in costs, inventory holdings, overstocks, etc.) and service-related (faster inventory turns, improved customer service, improved reliability of deliveries, etc.). This perspective is consistent with the literature in that the initial contributions of automatic replenishment programs have been from cost containment and improved service from suppliers. Therefore, ARP effectiveness, as measured by cost-related and servicerelated elements, should be positively driven by the three automatic replenishment program components (barcoding/scanning, forecasting/planning, and delivery) and by information systems capabilities (information timeliness and information compatibility). Regarding ARP cost-related effectiveness, two factors were found to significantly influence the construct. Forecasting/planning implementation (beta = .274, p In investigating the service-related effectiveness of ARP systems, two factors were found to drive the construct. Again, information timeliness was shown to be a strong driver of ARP effectiveness (beta = .521, p H3 examines the potential of a relationship between overall ARP performance and ARP effectiveness. ARP performance was measured using a 3-item scale including overall performance of the ARP relationship, profitability of the ARP relationship, and the impact of the ARP relationship on market performance. For each, a 7-point scale was used with endpoints of highly unsuccessful/highly successful, highly unprofitable/highly profitable, and little impact/great impact respectively. Both ARP cost-related and service-related dimensions were found to be significant antecedents of ARP performance (beta =.371, p CONCLUSION Considering that automatic replenishment programs are still relatively new, fairly widespread involvement was found among the sampling base. Although approximately one-third of the firms contacted have not implemented such programs, many of them advised that their companies are either considering automatic replenishment or are currently in the process of putting the systems in place. Among the respondent firms, automatic replenishment is still in the early stage of utilization (on average, slightly less than 5 years). For the most part, the basic hardware-type items including EDI and barcoding are in place at many firms. Initial emphasis has focused on staple items; automatic replenishment and joint forecasting is first applied to the most manageable products rather than seasonal or highly trendy/fashion items. An operational model (Figure 1) was developed detailing three potential ARP-oriented relationships. Results were somewhat mixed regarding the first two areas of interest (Hl: Information system capabilities are positively influenced by automatic replenishment program implementation and H2: Automatic replenishment program effectiveness is positively influenced by: a) automatic replenishment program implementation and b) information systems capabilities). Information system timeliness as well as information compatibility were found to be positively influenced by the forecasting/planning and delivery components of ARP implementation, but not by the barcoding/scanning component of ARP implementation. These findings may be explained by limitations of the sample or by the relatively short period of time ARP systems have been in place. Further research is warranted to confirm and further explicate the current findings. However, the findings with respect with H3 were more clear-cut. Support was found for a relationship between firm performance and automatic replenishment program effectiveness. Cost and service-related effectiveness associated with automatic replenishment program implementation positively influences ARP performance. Automatic replenishment program involvement is highly resource and time intensive. The programs are not easy to set-up and operate; however, based upon the findings of the current research, there is a pay-off. Overall performance is positively influenced by ARP effectiveness. The exploratory research and interviews indicate strong practitioner interest in ARPs. It is likely that trading partners will be demanding more automatic replenishment deliveries. This has important customer service and relationship implications-practitioners will need to be prepared to offer automatic replenishment. As previously discussed, the results suggest a phased development or ordering of implementation components which should be of value in making early implementation decisions particularly regarding resource allocation and information technology applications. The implications for researchers are probably even more marked. Because of the high practitioner interest, the potential to be gained through ARP implementation, and the fact that only minimal work has been conducted in the area, ARP-related research offers significant opportunities for expanding knowledge. Greater insights are needed as to the most appropriate ARP candidates, i.e., either by product type or business type. Also, research is needed to determine minimal commitment levels needed, i.e., determination of typical threshold levels or minimum involvement required to make involvement in automatic replenishment programs worthwhile. Additionally, research should focus on identification of the primary determinants of customer satisfaction with respect to ARP service.