تخصیص خرده فروشی قفسه: تجزیه و تحلیل مقایسه ای از رویکردهای هیوریستیک و متا هیوریستیک

This research presents a retail shelf-space decision model that incorporates a nonlinear profit function, vertical and horizontal location effects, and product cross-elasticity. We propose a linear programming formulation of the nonlinear profit function that can solve the shelf-space problem optimally. We describe potential advances in heuristic and meta-heuristic algorithms and compare the approaches through simulations and a field experiment. We discuss the impact of the number of item facings, vertical location, and horizontal location (e.g., we find the vertical location effect is approximately double the size of the horizontal location effect on profit performance).

Retail shelf-space management is one of the most difficult aspects of retailing. A significant reason is that while retail shelf-space is fixed, the numbers of new potential products (Dreze, Hoch & Purk 1994), customer wants (Corstjens & Doyle 1983), and competitors (Grewal et al., 1999 and Hansen, 2009) are constantly growing and evolving. At the same time, customers are consolidating shopping trips toward multi-purpose shopping (e.g., Popkowski et al. 2004). Thus, the success of any retailer depends on its ability to match its changing environment by continually deciding between how much of which products to shelve where and when. Indeed, the shelf location of products can significantly affect the products, and thus merchandise category, performance (Dreze, Hoch & Purk 1994). Thus, retailers benefit by expanding their focus from product-level performance to the total shelf-space configuration. The practice of analyzing shelf-space costs and product performance has been standard for several decades in retail practice and literature (e.g., Wickern 1966); many retailers have now adopted software programs such as Spaceman or Prospace for creating planograms. These programs can display historical product sales, profits, or inventory turnover information. In past years, the actual decisions of items to shelf location were usually made through human judgment because of the near infinitesimal combinations of shelf-space arrangements. As a result, shelf-space software programs are often only used as a visual template, and not to perform analysis. However, advancements in computing resources have permitted the development of more complex shelf-space models that are more consistent with consumer decision making (e.g., Borin and Farris, 1995 and Urban, 1998).3 Retail corporate buyers, category managers (who are employed by manufacturers), and retailer consultants can use these shelf-space models to improve their decision making, resulting in better financial performance. In this research, we first integrate the following three important elements into the basic shelf-space decision model: (1) a nonlinear profit function, (2) location effects, and (3) product cross-elasticity. In contrast to much of the research that has used nonlinear programming to model the nonlinear profit function, we propose a novel linear programming formulation of the nonlinear profit function that can solve the shelf-space problem optimally. We then extend the retail shelf-space literature by comparing potential advances in heuristic and meta-heuristic algorithms of the shelf-space model. We compare the different approaches/extensions through simulations and a field experiment. Unlike prior experiments that look at one store using a before-after scenario (i.e., lacking a control group), we investigate differences between three different configurations (i.e., including a control group) using a before-after comparison. Thus, we can control for both natural growth in consumer populations (in the control group) as well as change due to the novelty of making any changes at all (by having two different change groups of stores). The results indicate that the meta-heuristic approach outperforms the heuristic approach (and is closest to the linear programming formulation) in the simulations and the heuristic and control group in the natural field experiment. In summary, we find that the number of facings, vertical location, and horizontal location each have a significant, near-equivalent impact. We conclude with a discussion of limitations and implications.

Retail shelf-space optimization may be complex, but it is not a conundrum. The model put forth in this paper (1) assists practitioners in resolving the decision on how much of which products to shelf when and where, and (2) moves the literature towards a resolution on the debate concerning whether a heuristic or meta-heuristic approach provides more robust and faster estimation. Tests of the model show that the horizontal position, vertical position, and number of facings each affect the performance of the items within the shelf-space. Nonetheless, we caution that this model, or any other model, needs to be balanced with an appropriate perspective on store atmospherics and “retailtainment.” We echo the caution of Wickern (1966, p. 41), that “the success of retailing consists not only of selling merchandise, but also of the nature and completeness of the assortment.” For a growing number of customers, shopping is a hedonic event (e.g., Arnold & Reynolds 2003), affecting store patronage intentions (Grewal et al. 2003). Retailers often place product SKUS together according to size or brand, creating a category image. Shelf-space profit maximization could work against such imagery. Research is missing that balances shelf-space optimization and atmospherics. Until it is addressed, practitioners would be wise to also keep a “human touch” in the planogram design process.