Retail store location screening: A machine learning-based approach

Abstract

With numerous location choices across dispersed markets and a lack of detailed store-level information, the initial screening process for selecting store locations is challenging. We propose a machine learning-based model that uses public city-, competitor-, and point-of-interest (POI)-level data, including target group indices (TGIs), and apply machine learning to recommend sites based on predicted store performance. We demonstrate the effectiveness of our approach with real data from a jewelry retailing chain. Three machine learning approaches were developed and tested using data from 743 same-brand jewelry stores, and we find that a customized sequential ensemble model performs the best and outperforms the best available industry benchmarks. Our approach offers a new scalable and cost-efficient screening process for retailers to identify potentially top-performing locations.