Evaluating Promotions in Shopping Environments: Decomposing Sales Response into Attraction, Conversion, and Spending Effects

Retailers' marketing objectives can be classified into three broad categories: attraction effects that focus on consumers' store-entry decisions, conversion effects that relate to consumers' decisions about whether or not to make a purchase at a store they are visiting, and spending effects that represent both dollar value and composition of their transactions. This paper proposes a framework that incorporates all three of these effect categories and examines their influence on store performance. Specifically, store sales are broken down into four components: front traffic, store-entry ratio, closing ratio, and average spending. Using inexpensive and readily available infrared and video imaging technology, it is possible to measure these four components in a wide variety of retail environments, allowing retailers to obtain a richer understanding of the effectiveness of promotional activities on store sales. A set of twelve hypotheses based on the economics of information and promotion literatures is proposed. These hypotheses relate the presence of various promotions price, clearance, and new product, promotion scope, and the type of out-of-store communication vehicle used by retailers to each of the four store sales components. The proposed approach is then applied in two different empirical settings, both to test formally the hypotheses and to demonstrate more generally the richness of the information the approach can provide. The first application involves a Canadian apparel store that sells ladies' casual wear. The second application is based on a U.S. sporting-goods retail chain that sells a variety of sporting goods, including sportswear, sports shoes, and sports equipment. A joint model of four simultaneous equations using front traffic, store traffic, number of store transactions, and store sales as the endogenous variables is then formulated for the applications. Promotional factors are used as explanatory variables, along with a number of additional control variables including length of operation, day of week, holidays, seasonality, and weather. Seemingly unrelated regression is used to estimate the model efficiently. A comparison model that includes only store sales as the endogenous variable is estimated for comparison with the joint model. Results from these applications indicate that the proposed framework provides more detailed information about promotional effectiveness than more traditional models of store performance. The effects of specific promotional decisions on store performance are described. Specifically, price promotions have little impact on front traffic, but positively affect store entry and likelihood that a consumer will make a purchase. The effect of price promotion on consumers' spending in a store is also significant, but varies in sign with the type of promotion employed. Second, while greater promotional scope enhances store entry, promotions with narrow scope seem to have negative impact on store traffic. The effects of promotion scope on store performance also seem to be moderated by the scope of merchandise carried by the retailer. Increased promotional scope appears to have a greater effect on store traffic and consumers' spending for a multicategory retailer than for a more focused seller. Third, clearance promotions have a weaker effect on store entry when compared to other multiple-category promotions, while new-product promotions have a positive impact on conversion. Finally, newspaper advertisements, when compared to targeted coupons, have a stronger effect on store attraction but a weaker effect on spending. In addition to understanding the key drivers of store sales, retailers are also interested in determining whether or not their promotions affect store profitability. An assessment of the profit impact cannot be based on the change in overall store sales because promotions may affect various items or product categories inside a store differentially, and gross margins may not be the same for all items or categories. Although gross margin and item-or category-specific sales were not available for the two applications studied, the paper describes how such information can be integrated with the output of the proposed joint model to arrive at a richer understanding of how promotions affect overall store profitability. Finally, managerial and academic implications of this work are described, and potential extensions of the joint model are suggested.