The effect of backroom size on retail product availability – operational and technological solutions

Milićević, Nikola; Grubor, Aleksandar Article The effect of backroom size on retail product availability – operational and technological solutions Amfiteatru Economic Journal Provided in Cooperation with: The Bucharest University of Economic Studies Suggested Citation: Milićević, Nikola; Grubor, Aleksandar (2015) : The effect of backroom size on retail product availability – operational and technological solutions, Amfiteatru Economic Journal, ISSN 2247-9104, The Bucharest University of Economic Studies, Bucharest, Vol. 17, Iss. 39, pp. 661-675 This Version is available at: https://hdl.handle.net/10419/168940 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Retail Technologies for the 21st Century Innovation and Competitiveness in the Retail Trade Industry AE Vol. 17 • No. 39 • May 2015 661 THE EFFECT OF BACKROOM SIZE ON RETAIL PRODUCT AVAILABILITY – OPERATIONAL AND TECHNOLOGICAL SOLUTIONS Nikola Milićević1 * and Aleksandar Grubor2 1)2) University of Novi Sad, Novi Sad, Republic of Serbia Please cite this article as: Milićević, N. and Grubor, A., 2015. The effect of backroom size on retail product availability – operational and technological solutions. Amfiteatru Economic, 17(39), pp. 661-675 Abstract Amid the conditions of increasingly fierce competition, retailers are doing their best to meet the demands of their customers as efficiently as possible. Through the ever-growing level of product availability they raise the quality of service, which is positively reflected not only on the growth in sales, but also customer satisfaction. In the opposite case, the out-ofstock problem emerges, affecting not only customers, but also retailers and their suppliers. Bearing in mind, that the causes of the given problem occur most frequently in the last metres of the supply chain, in this paper we investigated the effect of backroom size on product availability, depending on the retail format. For this purpose, we used moderated regression analysis on the sample of 80 fast moving consumer goods in retail stores located on the territory of the Republic of Serbia. The obtained results pointed to opposite movements in the smallest and the largest format. Whereas in superettes the out-of-stock level lowers with the increase in the backroom size, it tends to drop in hypermarkets. Therefore, we pointed to some in-store problems that cause product stock-outs in different store formats. In addition to indicating the potential causes of analyzed relations, this paper also presents certain operational and technological solutions related to their mitigation. Keywords: retail, product availability, backroom size, replenishment, Serbia JEL Classification: M31, M30 Introduction The process of globalization and advances in information technology have changed market conditions, improving the customer position. Bearing in mind the high share of consumption in Gross Domestic Product (GDP) in most European countries and the importance of an active consumers’ policy for good market functioning (Dinu, 2006), customers can be regarded as a real power of the economy (Braşoveanu, Braşoveanu and * Corresponding author, Nikola Milićević –– milice[email protected] AE The Effect of Backroom Size on Retail Product Availability – Operational and Technological Solutions Amfiteatru Economic 662 Maşcu, 2014). With higher expectations, they devote an increasing amount of attention to what, where and when to buy, trying to satisfy their needs as cheap as possible. Amid the growing customer demands, retailers, additionally burdened by ever-rising competition, are forced to place a special accent on their shopping experience, that is, “every point of contact at which the customer interacts with the business, product, or service” (Grewal, Levy and Kumar, 2009, p.1). By means of higher level of service, retail companies are trying to deliver superior customer experience, and thus a higher customer satisfaction which, according to Gomez, McLaughlin and Wittink (2004) plays a key role in a successful business strategy. Given that customer service is manifested through product availability (Trautrims et al. 2009), one of the main tasks of retailers is providing its adequate level. A higher level of product on-shelf availability not only increases the likelihood of the customers finding and purchasing the desired product (Ton and Raman, 2010), but also motivates them to do their shopping in well-stocked stores (Dana and Petruzzi, 2001). From the S-D (service dominant) logic perspective on-shelf availability represents the key factor in value creation process (Ehrenthal, Gruen and Hofstetter, 2014). Instead of direct delivering, selling companies (manufacturers and retailers) manage and combine their resources in order to offer value propositions to potential customers. Only after all parties, including customers, have integrated their resources and created preconditions for successful service exchange, does value co-creation start (Vargo, 2011). Retailers play an integrating role in this process, enabling the exchange of services by making manufacturers’ products available to the customers (Ehrenthal, Gruen and Hofstetter, 2014). However, when out-of-stock (OOS) situation occurs, manufacturer and retailer value proposition to the customer is either altered (if customer substitutes or delays the purchase) or not realized (if customer cancels the purchase). Out-of-stock situations are one of the most frequent problems faced by customers, both in brick-and-mortar and in online shopping conditions. The situation when they cannot find the product of the desired brand, shape and size at the designated or expected place questions the achievement of their primary goal regarding its purchase and use. Furthermore, in addition to wasted time and energy, it also creates additional costs, which can be transactional, opportunity-related or substitutive by nature, depending on the customers’ response (Campo, Gijsbrechts and Nisol, 2000). Bearing in mind that shopping goals, as an important element of customer behaviour, influence how customers perceive the retail shopping environment and its individual elements, shopping behaviour, and satisfaction with the shopping experience (Puccinelli et al. 2009), their failure due to out-of-stock problem negatively affects retailers as well. The increase in the OOS rate in retail stores results in decreased customer satisfaction (Angerer, 2005), which may have a negative effect not only on store loyalty (Goldfarb, 2006), but also indirectly on the retailer’s business performance (Musalem et al. 2010). According to Andersen Consulting (1996), the out-of-stock problem costs the average grocery retailer 0.3 to 0.5% of the customer base. In addition to indirect, stock-out can also directly affect retailers business. If the customers cancel their purchase, change the store or substitute the out-of-stock product with the cheaper brand or item, the retailers will be faced with loss of sale (Ehrenthal and Stolzle, 2013). Gruen and Corsten (2007) assessed these losses on 4% of their annual sales. While Retail Technologies for the 21st Century Innovation and Competitiveness in the Retail Trade Industry AE Vol. 17 • No. 39 • May 2015 663 sale losses due to OOS were estimated at 175 million Euros in Netherland, in Germany they were estimated at 1 billion Euros per year in the food retail channel (Verhoef and Sloot, 2006). The fact that even the most successful companies are not immune to stock-outs is testified to by the world’s greatest retailer, Wall-Mart, which lost almost $3 billion in 2013 due to the given problem (Rosenblum, 2014). Like retailers, manufacturers are also affected by out-of-stocks. According to Gruen and Corsten “lost sales due to OOS items on average cost them $23 million for every $1 billion in sales” (2007, p.1). In addition to sale losses, the decrease of brand loyalty (Goldfarb, 2006) and the exchange of inaccurate distribution and inventory information (Ehrenthal Gruen and Hofstetter, 2014) are also some of the problems caused by stock-outs. Due to the significance of product on-shelf availability and all the effects that stock-out may cause in the supply chain, this paper analysis the OOS problem in the context of backroom size. It is structured as follows. In the literature review section, in addition to main root causes, we devoted special attention to in-store operations such as replenishment and ordering processes. After the methodology section, where we described conceptual model, data and method used for evaluating the impact of backroom size on OOS rate in various retail formats, we presented research results with their discussion. Before the conclusion, for better understanding and solving out-of-stock problems concerned in this research, in implication section we have described several operational and technological solutions that can be used in retail sector. 1. Literature Review The first publications related to out-of-stock situations in retailing appeared in 1960s and 70s (Walter and Grabner, 1975). Although attention in these was mostly devoted to customers’ reactions in OOS situations, they also partly raised some issues related to the root causes of this problem. According to Gruen, Corsten and Bharadwaj (2002), 72% of out-of-stocks occur as a consequence of retail in-store practices (retail store ordering and replenishing causes), while the remaining 28% are related to supply chain processes (replenishment and planning). With the rate of 47%, problems in ordering and forecasting activities (such as inaccurate inventory, bookkeeping and forecasts) represent major OOS causes. On the other hand, insufficient or busy store staff, infrequent, late or no shelf filling, congested backrooms, bad planogram execution, receiving errors and shrinkage are typical replenishing problems that cause 25% of shelf stock-outs. In addition to these, in their global report, Gruen, Corsten and Bharadwaj (2002) cited several other out-of-stock causes, such as inadequate shelf capacity, inverse effect of inventory, advertising and price changes, new product phase in and out, and manufacturer minimum order sizes. That the problem of stock-out occurs in the “last 50 metres” of the supply chain has been confirmed by results of other studies as well. According to Roland Berger Consultants (2003), over 85% of all out-of-stocks are caused by retailer in-store operations. By using a common approach in seven different European retail markets, they developed a standardized root cause catalogue that comprises 13 major and 49 sub-root causes. Survey results have shown that all four top root causes are related to retail store practices: store ordering (35%), delisting by store staff (30%), shelf replenishment (12%) and inventory inaccuracy (11%). AE The Effect of Backroom Size on Retail Product Availability – Operational and Technological Solutions Amfiteatru Economic 664 McKinnon, Mendes and Nabateh (2007) conducted interviews with supermarket managers in order to identify the reasons for out-of-stocks of three FMCG product categories (Fast Moving Consumer Goods). The results of their study indicated that 65% of all stock-outs were caused at the store. Similar to these results, Aastrup and Kotzab (2009), who analysed OOS situations at 42 retail stores, as Ehrenthal and Stolzle (2013), whose research included five European retailers, found 98% and 51,5%-94% of stock-outs to have been caused by in-store operations, respectively. Bearing in mind that top root out-of-stock causes originate from problems in store-related operations, some store characteristics may be considered as important factors of on-shelf availability. The results of several studies (Roland Berger Consultants, 2003; Fernie and Grant, 2008; Aastrup and Kotzab, 2009) have shown that OOS rates differ between retail stores depending on their size or formats which they belong to. In addition to size, on a sample of 84 products in 10 retail stores of a major European retailer, Angerer (2005) analyzed a few more store-related characteristics as work intensity, SKU density, store manager experience and backroom size. According to his research (2005), stores with too many or too few employees per square meter of salesroom, high SKU density and inexperienced store managers have higher out-of-stock rates. In relation to backroom, Angerer (2005) pointed to the existence of a positive correlation between backroom size and out-of-stocks. He explained this relationship through a counterproductive effect of “having too much stock” on shelf availability. Furthermore, besides Gruen and Corsten (2007), who analysed the results of Angerer’s research (2005), a number of other authors (Ton and Raman, 2010; Eroglu, Williams and Waller, 2011) in their own researches elucidated the negative impact of higher inventory levels on shelf availability through poor backroom-to-shelf replenishment process. The negative impact of higher inventory level, Waller et al. (2010) attributed to a „backroom logistics effect”. 2. Methodology Research Following studies concerning OOS causes, in our research we analyzed the effect of backroom size on shelf availability (expressed with average out-of-stock on a store level). However, as shelf availability varies between different stores sizes, in addition to Angerer’s research (2005), besides backroom size and out-of-stock level, our analysis included store format as moderated variable (moderator) as well. Figure no. 1: Conceptual model By analysing this model we investigated relations between mentioned variables in different store formats (figure no. 1). Thereby, the emphasis was on the smallest and the largest one. Backroom size Out-of-stock Retail formats: - Supperettes, - Supermarkets, - Hypermarkets Retail Technologies for the 21st Century Innovation and Competitiveness in the Retail Trade Industry AE Vol. 17 • No. 39 • May 2015 665 2.1. Sample Size and Variables Our sample consisted of 30 retail stores of a retailer that ranges among three top retailers on the Western Balkans. All stores are located on the territory of the Republic of Serbia. In terms of size, they were divided into three groups (Lovreta, Petković and Končar, 2009): 8 superettes (up to 400 square meter salesroom), 12 supermarkets (400 – 2000 square meter) and 10 hypermarkets (over 2000 square meter). In collaboration with the retailers’ supply chain director, we chose 80 FMCG products classified into 10 categories from each store: 6 personal hygiene care products, 6 household care products, 6 soft drinks, 8 products made from sugar (including sugar), 4 edible oils and fats, 12 cereal-based products and flour, 11 spices and aromas, 6 coffee brands, 15 sweets, and 6 salty snacks. In relation to this, attention was dedicated to best-selling as well as products of special importance for customers (difficult to substitute) and all of them were available (listed) in selected stores during the observation period. In this research we used data obtained from stores POS terminals for 2013. We obtained daily sales and inventory data for all 80 products in each store. They were used for calculating out-of-stock rate (which is most frequently used as product availability indicator), first on product and then on store level. Using POS estimation method (Hausruckinger, 2005; Gruen and Corsten, 2007) out-of-stock rate (OOS index) for item i in store s produces the ratio of lost (LS) and expected sales (ES) in units, over a given period of time, where the lost sale is the difference between the average and real sale: OOSis = LSis * 100 / ESis (1) However, as Hausruckinger’s approach for estimating expected sales corridor floor can be problematic for FMCG products with high sales volatility, for its calculation we also relied on features proposed by Papakiriakopoulos and Doukidis (2011). After calculating product OOS rates we calculated the mean OOS level for each store (OOSs). In addition to out-of-stocks, our analysis included backroom size as a store variable. Following Angerer (2005) we presented it as the ratio of backroom size to sales room. 2.2. Moderated Regression Analysis As relation between backroom size (BS) and out-of-stock on a store level (OOSs) may depend on retail format, it can be investigated with the use of moderated regression analysis. Therefore, our model, besides dependent variable (out-of-stock), includes one continuous predictor (backroom size), one categorical moderator (retail format) and their interaction. According to Frazier, Tix and Barron (2004), both the predictor and the moderator should be analyzed before structuring the equation. While categorical variable needs to be coded, continuous predictor needs to be centred or standardized. Bearing in mind that retail format as categorical variable has G = 3 levels (superette, supermarket and hypermarket), according to West, Aiken and Krull (1996) two code variables (G – 1) must be built into our regression model (C1 and C2). We used dummy variable coding system to represent them. AE The Effect of Backroom Size on Retail Product Availability – Operational and Technological Solutions Amfiteatru Economic 666 Table no. 1: Dummy coding system Base superette supermarket hypermarket dummy codes C1 C2 C1 C2 C1 C2 superette 0 0 1 0 1 0 supermarket 1 0 0 0 0 1 hypermarket 0 1 0 1 0 0 Among three versions of dummy coding system (presented in table no. 1), we have chosen the first one with superette as a comparison group (in which both code variables have 0 values). In the two remaining groups a value of 1 is alternately given to code variables (C1 in supermarket group and C2 in hypermarket group) for contrasting with comparison group. Although not a necessary requirement for moderator regression analysis (Whisman and McClelland, 2005), on the recommendation of many authors (Aiken and West, 1991; West, Aiken and Krull, 1996; Cohen et al. 2003) we centred continuous predictor (backroom size), i.e. converted it to deviation score form (West, Aiken and Krull, 1996). In this regard we replaced the predictor BS with BS’ (Whisman and McClelland, 2005): BS’ = BS – mean (BS) (2) Not only that centring reduces multicollinearity problems (Whisman and McClelland, 2005), but according to West, Aiken and Krull it also “yields the regression model that is most analogous to the familiar ANOVA model” (1996, p. 14). In addition to this operation, similar effects could be obtained from standardizing continuous predictors, i.e. converting them to “z scores” (Frazier, Tix and Barron, 2004). In many studies (West, Aiken and Krull, 1996; Frazier, Tix and Barron, 2004; Cohen et al. 2003), interaction term was presented as the product of predictor and moderator variables using the newly centred/standardized continuous variables or coded categorical variables. As our analysis included two coded variables (C1 and C2) we created two interaction terms, one for each coded variable (BS’C1 and BS’C2). Opposite to continuous predictor, these product terms, as dependent variable and coded variables as well, do not need to be centred or standardized (Fraizer, Tix and Barron, 2004). After all variables were prepared, we structured the regression model, presented with the following equation: OOSs = b0 + b1BS’ + b2C1 + b3C2 + b4(BS’C1) + b5(BS’C2) (3) Our full regression model is consisted of one centred predictor, two coded variables and two product terms. According to Whisman and McClelland the simple rule in forming the moderated regression model is “that the components of any products must always be included when testing the moderator effect” (2005, p. 113). For testing auto-correlation and multicollinearity, we used Durbin-Watson test and variance inflation factors. 3. Results and Discussion For testing the interaction effects, hierarchical multiple regression was conducted in two steps. According to Frazier, Tix and Barron (2004) in the first step we entered centred continuous variable and coded variables (as predictor and moderator variables), followed by interaction terms in the second step. Then, we compared the reduced regression model (without interaction terms) with full regression model given in equation 3. Retail Technologies for the 21st Century Innovation and Competitiveness in the Retail Trade Industry AE Vol. 17 • No. 39 • May 2015 667 Table no. 2: Hierarchical multiple regression Mod. R R2 Adj. R2 Std. E. Change Statistics DurbinWatson R2 F df 1 df 2 Sig. F 1* 0.856 0.732 0.701 0.010 0.732 23.701 3 26 0.000 2 ** 0.904 0.817 0.779 0.009 0.085 5.577 2 24 0.010 1.743 * OOSs = b0+ b1BS’ + b2C1 + b3C2 **OOSs = b0+ b1BS’ + b2C1 + b3C2 + b4(BS’C1) + b5(BS’C2) Comparing to the reduced model, the R2 change related to the interaction terms was 0.085 (table no. 3). It means that the interaction between backroom size and retail formats explained an additional 8.5% of the variance in out-of-stocks. The results of F test (F(2,24) = 5.577, p<0.05) also confirmed that adding interaction terms to the model resulted in accounting for statistically significant more variance in shelf stock-outs. Following DurbinWatson statistic (1.743), there is no auto-correlation in a regression analysis. In addition, values of the variance inflation factor (VIF), which are lower than 10, indicate that there are no potential problems of multicollinearity for variables. Table no. 3: Regression coefficients (full model) Model Variables Unstandardized Stand. t Sig. VIF B Std. E. Beta 1 Intercept 0.066 0.004 16.944 0.000 BS’ -0.004 0.012 -0.032 -0.304 0.763 1.104 C1 -0.021 0.005 -0.530 -4.244 0.000 1.514 C2 -0.044 0.005 -1.055 -8.113 0.000 1.642 2 Intercept 0.068 0.003 19.973 0.000 BS’ -0.033 0.014 -0.295 -2.423 0.023 1.943 C1 -0.024 0.004 -0.591 -5.405 0.000 1.571 C2 -0.042 0.005 -1.009 -8.589 0.000 1.813 BS’C1 0.061 0.024 0.271 2.537 0.018 1.501 BS’C2 0.081 0.029 0.307 2.768 0.011 1.612 As presented in Table no. 3, values of regression coefficients changed after adding interaction terms. While the coefficient of backroom size variable had negative values in both models, it was statistically significant only in the second one (full regression model), with p = 0.023. So, without interaction terms, we would have concluded that backroom size did not have significant relation with shelf-out-of-stock. In addition, in full regression model, coded variables and interaction terms also had significant t-tests. Opposite to regression coefficients of coded variables, which were negative, regression coefficients for interaction terms had positive values. The facts that the increment in the squared multiple correlation (R2) is significantly greater than zero and that the regression coefficients of interaction terms significantly differ from zero, support the thesis that relationship between backroom size and out-of-stock level differs among different store formats. In order to test these relations within each store format, following Whisman and McClelland (2005) we rearranged the equation 3 into: OOSs = (b0 + b2C1 + b3C2) + (b1+ b4C1 + b5C2)BS’ (4) AE The Effect of Backroom Size on Retail Product Availability – Operational and Technological Solutions Amfiteatru Economic 668 The obtained equation represents the relationship between OOSs and BS’, where the term in the first set of brackets (b0 + b2C1 + b3C2) represents the intercept and the term in the second set of brackets (b1+ b4C1 + b5C2) represents the regression slope. Bearing in mind that our categorical variable consists of three groups, after substituting the values of the dummy codes, we simplified the equation 4 for each group:  OOSs = b0 + b1BS’, for superettes (C1 = 0; C2 = 0), (5)  OOSs = (b0 + b2) + (b1+ b4)BS’, for supermarkets (C1 = 1; C2 = 0), (6)  OOSs = (b0 + b3) + (b1+ b5)BS’, for hypermarkets (C1 = 0; C2 = 1), (7) The equations 5, 6 and 7 are simple regression equations that show the regression of out-ofstock (dependent variable) on the backroom size (continuous predictor) for three different retail formats (signified with code variables). For superettes, the regression coefficient b1 gives the regression of OOSs on BS’. For supermarkets the regression of OOSs on BS’ is (b1+ b4) and for hypermarkets the regression of OOSs on BS’ is given as (b1+ b5). For the regression of OOSs on BS’ for three retail format groups, b1, (b1+ b4) and (b1+ b5) represent simple slopes, that according to West, Aiken and Krull (1996) are completely comparable to the ANOVA simple effects. On the other hand, b0, (b0 + b2) and (b0 + b3) are intercepts for superette, supermarket and hypermarket groups respectively. After using values of regression unstandardised coefficients presented in table no. 4, we calculated intercepts and simple slopes for all three groups:  OOSs = 0.068 – 0.033BS’, for superettes, (8)  OOSs = 0.044 + 0.028BS’, for supermarkets, (9)  OOSs = 0.026 + 0.048BS’, for hypermarkets. (10) While there was a negative simple slope (-0.033) for superettes, for other two groups they were positive. Thereby, positive correlation is much stronger in hypermarkets (0.048) than in supermarkets (0.028). In addition, we tested the statistical significances of these slopes. According to Cohen et al. (2003), instead of centred continuous predictor (BS’), we added three variables (BS’1 for superettes, BS’2 for supermarkets and BS’3 for hypermarkets) in the full model (presented in equation 3, representing the backroom size effect for each of the three retail format groups. In them, each group's BS’ values were coded on a variable for which all other groups were coded 0. Regression coefficients for these variables reflect the slopes of out-of-stock on backroom size for each retail format group. In table no. 4, we presented the reproduced full regression model. Table no. 4: Regression coefficients (reproduced model) Variables Unstandardized Stand. t Sig. VIF B Std. E. Beta Intercept 0.068 0.003 19.973 0.000 C1 -0.024 0.004 -0.591 -5.405 0.000 1.571 C2 -0.042 0.005 -1.009 -8.589 0.000 1.813 BS’1 -0.033 0.014 -0.216 -2.423 0.023 1.046 BS’2 0.028 0.020 0.124 1.408 0.172 1.014 BS’3 0.048 0.026 0.181 1.851 0.077 1.260 R2 = 0.817, p < 0.01, Durbin-Watson 1.743 Retail Technologies for the 21st Century Innovation and Competitiveness in the Retail Trade Industry AE Vol. 17 • No. 39 • May 2015 675 Puccinelli, N.M., Goodstein, R.C., Grewal, D., Price, R., Raghubir, P. and Stewart, D., 2009. Customer Experience Management in Retailing: Understanding the Buying Process. Journal of Retailing, 85(1), pp.15-30. Raman, A., DeHoratius, N. and Ton, Z., 2001. Execution: The Missing Link in Retail Operations. California Management Review, 43(3), pp.136-152. Roland Berger Strategy Consultants, 2003. ECR – Optimal Shelf Availability Increasing Shopper Satisfaction at the moment of truth. [pdf] ECR Europe. Available at: <www.gs1.ch/docs/default-source/prozesse-dokus/ecr-europe-osa---optimal-shelfavailability.pdf?sfvrsn=2> [Accessed 14 July 2014]. Rosenblum, P., 2014. Walmart's Out Of Stock Problem: Only Half the Story? [online] Available at: <http://www.forbes.com/sites/paularosenblum/2014/04/15/walmarts-outof-stock-problem-only-half-the-story/> [Accessed 18 August 2014]. Swisslog, 2013. Picking Technologies: Improve the Speed and Accuracy of Order Fulfillment. [online] Available at: <http://www.swisslog.com/en/Solutions/WDS/SupportingPicking-Technology> [Accessed 12 September 2014]. Szmerekovsky, G.J., Tilson, V. and Zhang, J., 2011. Analytical Model of Adoption of Item Level RFID in a Two-echelon Supply Chain with Shelf-space and Price-dependent Demand. Decision Support Systems, 51(4), pp.833-841. Ton, Z. and Raman, A., 2010. The Effect of Product Variety and Inventory Levels on Retail Store Sales: a Longitudinal Study. Production and Operations Management, 19(5), pp.546-560. Trautrims, A., Grant, D.B, Fernie, J. and Harrison, T., 2009. Optimizing On-shelf Availability for Customer Service and Profit. Journal of Business Logistics, 30(2), pp.231-247. Vargo, S.L., 2011. Market Systems, Stakeholders and Value Propositions: Toward a Service-dominant Logic-based Theory on the Market. European Journal of Marketing, 45(1/2), pp.217-222. Verhoef, P.C. and Sloot, L.M., 2006. Out-of-Stock: Reactions, Antecedents, Management Solutions, and a Future Perspective. In: M. Krafft and M.K. Mantrala, eds. 2006. Retailing in the 21st Century. Heidelberg: Springer. pp.239-253. Waller, M.A., Williams, B.D., Tangari, A.H. and Burton, S., 2010. Marketing at the Retail Shelf: Exploring Moderating Effects of Logistics on SKU Market Share. Journal of the Academy of Marketing Science, 38(1), pp.105-117. Walter, C.K. and Grabner, J., 1975. Stockout Models: Empirical Tests in a Retail Situation. Journal of Marketing, 39(3), pp.56-68. West, S.G., Aiken, L.S. and Krull, J.L., 1996. Experimental personality designs: Analyzing categorical by continuous variable interactions. Journal of Personality. 64(1), pp.1-48. Whisman, M.A. and McClelland, G.H., 2005. Designing, Testing, and Interpreting Interactions and Moderator Effects in Family Research. Journal of Family Psychology, 19(1), pp.111-120.