Promoting impulse buying by allocating retail shelf space to grouped product categories

Abstract

This paper addresses a problem where a retailer seeks to optimize store-wide shelf-space allocation in order to maximize the visibility of products to consumers and consequently stimulate impulse buying. We consider a setting where the retailer, because of product affinities or the retailer’s historical practice, has pre-clustered product categories into groups each of which must be assigned to a shelf. On the basis of its location in the store layout, each shelf is partitioned into contiguous shelf segments having different anticipated customer traffic densities. The retailer seeks to assign each group of product categories to a shelf, to determine the relative location of product categories within their assigned shelf, and to specify their allocated total shelf space within given lower/upper bounds. We propose a 0–1 integer programme that takes into account expected customer traffic densities within the store, groups of product categories, their relative profitability, and the desirability to keep certain product groups in the same aisle, with the objective of maximizing the impulse buying profit. The proposed model is grounded in a preprocessing scheme that explores feasible assignments of subsets of product groups to available aisles by iteratively solving an -hard subproblem and is numerically observed to greatly outperform an alternative mixed-integer programming formulation. We demonstrate the usefulness of and the enhanced tractability achieved by the proposed approach using a case study motivated by a grocery store in New England and a variety of simulated problem instances.

Appendices

Appendix A

Total visibility of a product, v _p

The current form of v _p ,

assumes that product p can be bought multiple times along each edge where it is allocated. This can be seen as a reasonable approximating assumption since a product is typically allocated to one or two edges at most.

Exact form under an assumption on the direction of walking along a shelf

Assume that a customer walks along Shelf b according to the order of edges, 1, 2, 3, …, . Then, a customer either buys Product p from Edge 1, with probability (k₁s_p1)/c₁. Otherwise, if the customer does not buy p from Edge 1, the customer could buy p from Edge 2, (1−(k₁s_p1)/c₁)(k₂s_p2)/c₂. In addition, if the customer does not buy p from Edges 1 and 2, the customer could buy p from Edge 3, (1−(k₁s_p1)/c₁)(1−(k₂s_p2)/c₂)(k₃s_p3)/c₃, and so on. The probability that the customer buys p from edge i is:

This gives a total visibility of

where sums over empty sets are set equal to 0.

Comparing the exact and the approximate form

Note that (35) and (36) can be rewritten as:

This gives another justification for (35) as a first-order approximation of (36). For example, if a product p is allocated to Edges 1 and 2 of a shelf, then:

Appendix B

The generation of traffic densities, unit revenues, and impulse purchase rates

Figure B1 demonstrates a heat map for the grocery store in New England, examined in Section 3. As illustrated in Figure B1, shelf segments that are close to the end-of-aisles/entrances tend to have higher customer traffic density, whereas shelf segments located in the middle of aisles usually have low customer traffic. On the basis of that, to estimate the numerical values of customer traffic density k _e , shelf segments were divided into three categories, having low, medium, high customer traffic density as shown in Figure B1. Then the numerical values were randomly generated using the following intervals for illustrative purposes:

• k _e ∈[0.05, 0.25) for shelf segments with low traffic density.

• k _e ∈[0.35, 0.65) for shelf segments with medium traffic density.

• k _e ∈[0.75, 1) for shelf segments with high traffic density.

Figure B1

Shelf segment attractiveness/visibility levels.

Moreover, unit revenues and impulse purchase rates of product categories are listed in Table B1. Unit profit margins of product categories are confidential and, therefore, not provided. Instead, we displayed unit revenues for interested readers.

Table B1 Unit revenues and impulse purchase rates of product groups for a grocery store in New England, the United States