The Influence of Customer Movement between Sales Areas on Sales Amount: A Dynamic Bayesian Model of the In-store Customer Movement and Sales Relationship

摘要

Recent years have seen active research that utilizes information combining geographic data and sensor data, called geospatial information, in urban planning, medical care and marketing. In this study, we focus on RFID technology that records position information (i.e., spatial information) of shopping carts in a supermarket, and estimate the latent space-time structure of the store as observation data of customers’ visits. Then, we propose a dynamic Bayesian model for sales analysis, which extends the conventional state-space model to include the spatiotemporal structure. From the results of the model analysis, it is obvious that supermarkets have clear periodic structures in units of time periods and weekly structures, and they are dynamically related to the adjacency of each sales area. By utilizing the visualization of the space-time structure of the sales area, it is possible to easily inform the store manager about the influence of customers’ visits on sales outcomes.