Measuring the impact of shelf position on a retailer's sales

Abstract

Prior academic work in retailing suggests that a substantial fraction of purchasing decisions are made inside the store, where customers must choose from a large variety of products. In-store marketing tactics such as shelf space management then become particularly relevant as the number of products that could be displayed increases, while shelf space remains limited. Although extant research has shown a significant relationship between product location and sales, most studies are mainly focused on space elasticity and less is currently known about the effects of shelf position on sales. In addition, the potential reverse causality in the shelf position/sales relationship and the high costs of conducting controlled experiments have made it even more difficult to obtain managerial implications. The purpose of this study is to contribute to this stream of research by examining the effects of shelf position on sales using a novel dataset about product position collected using a combination of robotics, internet of things and machine learning technologies. The data considers product position data for two stores of a major supermarket chain in Latin America and sales data provided by one of the retailer s vendors covering five product categories. Using these data, we first perform a descriptive analysis to explore possible (non-causal) associations between variables. This analysis suggests that products located at the edges of an aisle sell between 37.0-69.6% more than those located near the center, while products located closer to the checkout counters sell between 24.3-48% more than those located farther away. The results also suggest that products located at the mid-height of the shelves sell 2.7 to 5.6% higher sales than those placed on the bottom shelf. We then propose a quasi-experimental approach to draw causal inferences about the impact of shelf position on sales. This requires us to rely on a semiautomatic method to detect changes in shelf position and hence identify suitable changes for a quasi-experimental approach, which we implement with a difference-in-differences analysis. In contrast with the results from the descriptive analysis, the quasi-experimental analysis finds no significant impact of the horizontal position of a product on its sales. However and consistent with the descriptive analysis, this analysis also suggests that by getting closer to the vertical center of a shelf, a product could increase its sales by 15.2-23.4%, while increasing the height of a product can decrease its sales by 26.6%.