Location differs between traditional and new retail: A comparison analysis of Starbucks and Luckin Coffee in China using machine learning

The distinction between new retail and traditional retail has been particularly emphasized and shaped by the influence of e-commerce giants and rapid technological advancements in China. However, related literature paid little attention to the nonlinear effects of location determinants, and the comparison between traditional and new retail. By taking Starbucks and Luckin Coffee in China as examples, this study investigated the nonlinear effects of influencing factors on traditional and new retail locations using the XGBoost model and interpretation approaches (relative importance and partial dependence plots). Results show that traditional and new retail brands with overlapping products do not overlap in terms of location choice. For example, regarding demographics, new retail is more sensitive to population density, population mobility, and the proportion of young people than traditional retail. Regarding the social economy, per capita GDP and rent have an obvious threshold effect on the promotion of traditional retail density, while they have a marginal effect on new retail. Regarding competition, traditional retail is keen to face peer challenges head-on, while new retail will choose to avoid high-competition areas. These findings are effective in understanding the differences between emerging new retail and traditional retail from the perspective of location selection.