Site Selection Analysis and Prediction of New Retail Stores from an Urban Commercial Space Perspective: A Case Study of Luckin Coffee and Starbucks in Shanghai

In the context of digital transformation, examining the differences in commercial site selection and the factors influencing these decisions holds significant practical value for understanding market adaptation strategies across varying business models and predicting future industry trends. This study divides the research area into 100 m × 100 m grids and employs a random forest model and related interpretability methods to conduct an empirical analysis of the site selection and influencing factors of Luckin Coffee and Starbucks stores in Shanghai. By integrating the prediction results with existing planning documents, this study achieves a coupling between urban spatial structure and location strategies. The findings indicate the following: (1) The random forest model demonstrates high accuracy in predicting new retail store locations, with an accuracy rate of 90.0% for Luckin Coffee and 92.2% for Starbucks. (2) The influence of traditional factors on the expansion of new retail coffee stores is declining, while Luckin Coffee’s layout demonstrates a stronger reliance on urban functional zones. (3) Relative suitability is derived by calculating the difference between the predicted probability values and the normalized kernel density values. In the central activity areas of the city, the relationship between site selection probability and suitability exhibits an inverse correlation, with Starbucks generally showing higher relative suitability overall. (4) Suitable areas for both brands’ site selections are spatially contiguous and integrated within the urban fabric, which suggests significant growth potential for both brands in the main urban areas. This study not only focuses on commercial optimization but also offers theoretical and methodological insights by exploring how different retail models interact with urban spatial structures, thereby contributing to the fields of retail geography and spatial governance.