Are Spatial Neighbors Complements or Substitutes? Discovering a Visits-based Local Market Structure

Firms interact with their competitors and complementors within a multi-firm network environment. Knowledge of such market structure is critical for improving firm's marketing and operation strategy designs. We formulate a new problem to discover the local market structure - complementary and substitution patterns based on visits - for a particular store or service outlet. We develop a scalable and general framework, which relies only on easily accessible and privacy-preserving aggregated mobility data. We then develop a semi-parametric spatial econometrics model and use an instrumental variable approach to estimate the visit spillovers. Using data for three brands, we show that our approach can identify neighboring outlets that act as complements or substitutes in relation to the visits. Interestingly, we find that the publicly-recognized competitors do not necessarily show any spillover effects at the local level. In addition, we show that in the case of a complementary relationship, visits to a neighboring store led to an average increase of 3.9%~37.9% in visits to the focal brand's store in our dataset. Meanwhile, substitution effects from other stores resulted in a reduction in visits by up to 33.5%. These findings underscore the importance of location-based analysis to identify complementary and substitutionary relationships for visitations. Building on the discovered local market structure, we develop two types of centrality measures to score the importance both of neighboring brands and of local stores belonging to a focal brand. We demonstrate that our centrality measures informed by the local market structure can improve cross-promotion efforts and web design performance of local marketplaces using policy simulations