Not all franchises are the same: an unsupervised machine learning approach to strategic segmentation

Purpose The franchising sector has emerged as a strategic investment avenue, offering scalability, standardization and risk mitigation. However, franchise chains differ significantly in their ability to create value. This study aims to investigate how different franchise profiles generate value, considering financial, operational and institutional attributes. Design/methodology/approach A quantitative methodology was used using unsupervised machine learning techniques. Initially, the K-Means algorithm served as an exploratory segmentation tool. Subsequently, principal component analysis (PCA) was applied for dimensionality reduction, followed by hierarchical density-based spatial clustering of applications with noise (HDBSCAN) as the primary clustering technique. This approach enabled the identification of dense cluster structures and the detection of atypical franchise chains. Variables such as franchise fee, payback period, franchisee support and franchise chain satisfaction were analyzed. Findings HDBSCAN outperformed K-Means by revealing clusters that more accurately represented the underlying data structure. Six distinct strategic profiles were identified, including an outlier cluster (Cluster −1), comprising highly differentiated franchise chains. The clusters exhibited patterns linked to institutional reputation, organizational maturity and sector engagement. The results indicate that value creation extends beyond initial investment and relies on a combination of structured support, operational standardization and symbolic positioning. Research limitations/implications Although the applied clustering techniques effectively captured structural nuances, the study was limited to a specific set of financial and operational variables. Future research could benefit from incorporating behavioral, digital and regional dimensions to enrich the understanding of value creation in franchise chains. Originality/value This study contributes to franchising literature by integrating advanced unsupervised learning techniques to segment franchise chains and uncover strategic profiles. The findings provide valuable guidance for franchisors and investors, emphasizing the centrality of institutional reputation and support mechanisms in sustaining franchise success. The use of HDBSCAN represents a methodological contribution by enhancing the accuracy of cluster detection in complex data environments.