A predictive analytics framework for cannibalization-aware retail expansion: A case study of a restaurant chain

Location selection plays a central role in retail chain expansion, influencing customer reach, operational efficiency, and overall network performance. In multi-unit systems, expansion decisions must balance market coverage and accessibility against the risk of internal demand redistribution, or cannibalization. Traditional site selection approaches often rely on geographic distance and trade-area overlap metrics; however, such spatial heuristics frequently fail to capture how customers adjust their behavior following the introduction of new branches. This study examines whether customer-level behavioral data can improve the assessment of intra-brand cannibalization risk before it occurs. Drawing on real-world transaction records from a restaurant chain, we develop a machine learning–based predictive framework that integrates behavioral indicators such as order recency, frequency, and monetary value with spatial features including distance and trade-area overlap. By modeling individual customer switching behavior rather than aggregate outcomes, the framework enables scenario analysis across alternative location choices. Empirical findings indicate that behavioral engagement features, particularly order frequency, play a key complementary role alongside spatial proximity measures in explaining customer switching patterns. While geographic factors remain relevant, they appear insufficient as standalone predictors of cannibalization risk. Rather than a formal spatial-optimization procedure, the proposed approach provides a detailed, behavior-based decision-support tool for restaurant chain expansion planning, demonstrated through a case study of a restaurant chain.