Machine Learning-Driven Customer Segmentation: A Behavior-Based Approach for F&B Providers

This study explores behavior-based customer segmentation by integrating Recency, Frequency, and Monetary value (RFM) analysis with the K-Means++ clustering algorithm. Using one year of invoice-level transactional data from a Romanian Food and Beverage (F&B) provider serving restaurants and coffee shops, the research aims to deliver actionable insights to enhance marketing and sales strategies. After standardizing the dataset to address scale differences, the Elbow Method was applied to determine the optimal number of clusters, resulting in five distinct customer groups: Champions, Loyal Customers, Promising, Hibernating Customers, and Lost Customers. Notably, the Champion segment, consisting of a single customer, accounts for 15% of total sales, highlighting both profitability and dependence risks. Loyal and Promising customers were identified as the most strategically valuable segments for targeted retention and growth initiatives. The clustering results were validated through visualization techniques and internal metrics, confirming the effectiveness of the segmentation. By relying exclusively on transactional data, this approach ensures GDPR compliance and offers a scalable framework for continuous monitoring and dynamic strategy adaptation. The findings provide immediate financial implications for the company, illustrating the potential of machine learning-driven behavior-based segmentation in B2B markets with frequent, recurring transactions.