RFM segmentation identified 4 distinct customer clusters from 936 customers. The dominant segment contains 497 customers (53% of base), while the smallest has 7 (1%). Overall cluster cohesion (silhouette = 0.529) indicates strong separation.

Analysis included 936 unique customers from 2010-12-01 to 2011-12-09 (373 days span). Transactions aggregated to customer level for RFM computation. K-Means clustering with k=4 performed on standardized Recency, Frequency, and Monetary metrics.

Input: 2000 transaction rows from 967 unique customers. Removed 514 rows with missing customer ID and 45 rows with returns (negative quantity). Final analysis: 936 customers with 1446 total valid transactions. No missing values in date or price fields.

Recency ranges from 0 to 373 days (median 107). Frequency spans 1 to 29 transactions (median 1). Monetary values range 0.19 to 1790.00 (median 16.37). Wide spreads in all three dimensions indicate heterogeneous customer behaviors ideal for segmentation.

Segment distribution is highly unbalanced, ranging from 7 customers (Cluster 2) to 497 customers (Cluster 3). Lost comprises 53.1% of the customer base.

Cluster 3 is the most cohesive segment (silhouette = 0.599) with mean RFM of R=67 days, F=1.2 txns, M=$18.74. Cluster 2 is the least cohesive (silhouette = 0.095) with R=23, F=16.0, M=$737.13. Stronger silhouette scores indicate tighter, more meaningful clusters.

Clear segment separation visible: Cluster 2 leads in purchase frequency (mean 16.0 txns) while Cluster 2 dominates in monetary value (mean $737.13). Positive F-M correlation across all clusters indicates higher purchase frequency correlates with higher total spend. Sub-clusters may emerge within larger segments.

Cluster 2 represents the most engaged segment with median recency of 23 days, while Cluster 1 is the most at-risk with median 269 days. Dormant segments (high recency) are prime targets for win-back campaigns. Low-recency segments warrant loyalty initiatives.

Average silhouette score across all clusters is 0.529 (strong separation). Scores range from 0.095 to 0.599. Values > 0.5 indicate well-separated clusters; < 0.25 indicate potential overlap or weak definition. Review low-scoring clusters for possible re-segmentation.

Segments mapped to business personas: Lost (n=353), Champions (n=7), and others. Each segment receives tailored recommendations based on RFM profile and lifetime value. See table for specific engagement strategies per segment.