Data-Driven Customer Segmentation Using the RFM Model and K-Means Clustering: A Longitudinal Analytical Study of Online Retail Behavior Using Python and Orange3

In the contemporary digital economy, the ability to effectively analyze and segment consumer behavior has become a

critical determinant of organizational competitiveness and strategic decision-making. This study provides a

comprehensive quantitative analysis of customer behavior using the Recency–Frequency–Monetary (RFM) model

combined with K-Means clustering techniques, applied to longitudinal transactional data from an international online

retail platform during the period 2010–2011. Adopting a data-driven analytical framework, the research integrates

statistical modeling and machine learning techniques within a Python-based computational environment, supported by

the Orange3 data mining platform. The study constructs behavioral profiles based on three core dimensions—recency of

purchase, purchase frequency, and monetary value—allowing for the identification of distinct customer segments. The

clustering process is validated using the Silhouette coefficient and further supported by inferential statistical testing through

one-way analysis of variance (ANOVA), ensuring the robustness and internal consistency of the segmentation model.

The findings reveal the existence of three statistically significant customer segments: high-value (champion) customers,

inactive or lost customers, and regular or potentially valuable customers. Each segment exhibits distinct behavioral and

economic characteristics, enabling the formulation of targeted marketing strategies aimed at maximizing customer lifetime

value and optimizing resource allocation. The study contributes to the existing literature by demonstrating the

effectiveness of integrating traditional marketing analytics models with contemporary machine learning techniques, while

also providing practical implications for e-commerce firms seeking to enhance customer relationship management

through data-driven segmentation approaches.