Customer Segmentation Using RFM Analysis and Clustering for Promotion Decision Support

Abstract— This study presents an analysis of a customer-level dataset for RFM-based customer segmentation and promotion decision support. The supplied CSV contains 4,269 customer records and 13 columns, including Recency, Frequency, MonetaryValue, Tenure, cluster identifiers, segment names, churn-risk categories, and IQR outlier indicators. The file matches the labeled output documented in the customer-segmentation-rfmt GitHub repository, which identifies UCI Online Retail as the transaction source. This paper analyzes the supplied labels and profiles; it does not independently reproduce the transaction-to-cluster pipeline. The dataset contains three segments: Champions (948 customers; 22.2%), Promising (1,634; 38.3%), and At Risk (1,687; 39.5%). No missing values were detected in the CSV. Promising has the highest mean Frequency (188.29) and Monetary Value (4,412.44), while At Risk has the highest mean Recency (174.20 days). The paper reports a silhouette diagnostic of 0.0673 for its tested three-feature representation; the repository’s score of 0.351 for k = 3 is a repository result, not an independently reproduced result. Findings describe the supplied data and do not establish optimal clustering or promotion effectiveness. Full validation requires the original transactions, documented reference date, preprocessing steps, and clustering evaluation. The decision-support workflow presents segment profiles and candidate promotions for business review, not proven campaign outcomes.