Supporting Retail Sale by Using a Combination of RFM model and Clustering Algorithm

In the increasingly competitive retail sector, understanding customer behavior to formulate personalized marketing and sales strategies is imperative. This study proposes an automated customer segmentation approach based on the analysis of multidimensional transaction data from global shopping centers, specifically utilizing a comprehensive retail dataset from Istanbul spanning from 2021 to 2023. The core methodology integrates the RFM (Recency, Frequency, Monetary) analytical framework with an unsupervised machine learning clustering algorithm. The RFM model is applied to quantify individual buyer engagement by measuring the time elapsed since the last purchase (Recency), the total number of transactions (Frequency), and the total expenditure (Monetary). Subsequently, a clustering algorithm is employed to analyze these variables, thereby uncovering underlying behavioral patterns and grouping customers with similar characteristics into meaningful structures. The results demonstrate the successful partitioning of the customer base into three strategic segments: Regular Customers (39.65%), Infrequent Customers (41.08%), and Potential Customers (19.26%). The quality and distinctness of these clusters are validated through robust evaluation metrics, achieving a Silhouette Score of 0.339, a Calinski-Harabasz Score of 68829.85, and a Davies-Bouldin Score of 1.11, indicating well-defined and adequately separated groupings. This practical implementation provides businesses with a powerful tool to comprehensively understand consumer profiles, thereby optimizing targeted outreach campaigns and enhancing overall customer retention strategies.