The Case for Interpretability: Evaluating Clustering Approach in Customer Segmentation

Customer segmentation is important for marketing and business strategy. This study explores the effectiveness of two widely used clustering algorithms, K-Means and Agglomerative Clustering, in the context of real-world customer segmentation. Using transactional data from a retail store with over 120,000 records, the analysis focuses on whether these clustering methods can produce meaningful, interpretable customer segments based on behavioral features such as frequency, monetary value, recency, and product variety. While both algorithms succeeded in forming statistically distinct clusters, the results revealed key limitations in business applicability. Clusters often lacked interpretability, were demographically homogeneous, and required additional effort to translate into actionable insights. Furthermore, both methods assume geometric similarity and require the number of clusters to be defined in advance, which can be problematic for complex, overlapping customer behavior. These findings highlight the need for clustering approaches that are better aligned with business decision-making