Know your competitor! Analyzing and predicting the location of competing stores: The case study of Valora at Swiss railway stations

Abstract

Location choice in retailing is a key subject of retail location theory, but is also of great practical relevance. Retail companies must assess the demand and competition situation and try to anticipate the behavior of their competitors. This study examines location choice patterns of two convenience food formats from Valora, Avec and k kiosk, at Swiss train stations. The study combines an analytical and a predictive modeling approach. Possible location factors for the two store types are derived from the literature. Publicly available data from the SBB (Schweizerische Bundesbahnen) serve as the basis of the analysis. Binary logit models are built for the formats examined in order to identify the determinants of location choice. Machine learning algorithms are used to check and optimize the predictive ability of the models. It turns out that people boarding, alighting and changing trains at train stations (which represent the main demand for convenience stores at railway stations) are an important determinant of location choice. The more frequent a train station is, the more likely it is that Avec or k kiosk will be present there. Furthermore, format-specific clustering and avoidance patterns emerge. Both Valora formats show an avoidance of each other. While Avec tends to avoid competing convenience supermarkets, this is not the case with k kiosk. With the help of machine learning, the predictive ability of the models can be greatly improved. A prediction model with high specificity and sensitivity is built for k kiosk and applied on a real case.