Data flow · RFM segmentation

RFM segmentation of 1,983 bank clients

RFM segmentation of 1,983 bank clients A synthetic 10,000-row transaction log is cleaned by dropping its flagged anomaly rows down to 9,492 transactions and 1,983 clients, scored into R/F/M quartiles and six combination-based segments, then tested for monetary concentration — Gini 0.324, with the top 20% of clients holding 38.2% of revenue — before export. 01INGEST 02CLEAN 03RFM 04SCORE 05SEGMENT 06VERIFY ANALYSISPIPELINE ARTIFACTS& EVIDENCE 508 DROPPED SEG + SPEND SHARE OF CLIENTS vs REVENUE CLIENTS REVENUE RATIO Champions 126 1.91 Loyal 376 1.43 At Risk 349 1.34 Potential Loyalists 79 1.16 Need Attention 410 0.62 Hibernating 643 0.61 Gini 0.324 — top 20% hold 38.2% of revenue 1,983 clients · rfm_output.csv SRC Transaction Log 10,000 txns · seed 42 DB CLN Parse & Drop 9,492 txns survive DB DB RFM RFM Table 1,983 clients · r·f·m DB TB SCR Quartile Score R/F/M 1–4 · sum 3–12 TB TB SEG Segment Map combo rules → 6 labels TB TB EXC Anomalies Excluded 508 rows · is_anomaly=1 DB VAL Concentration Check Gini 0.324 · top20 38.2% TB FL OUT Deliverables CSV · Excel · charts Streamlit · dbt boards FL NODE SOURCE STAGE FOCAL · GINI EXCLUDED · DROPPED DELIVERABLE SHARE PANEL DATA DB DATASET TB TABLE FL EXPORT LEFT CHIP = INPUT · RIGHT CHIP = OUTPUT FLOW DATA HANDOFF FOCAL · SEG + SPEND ANOMALY DROP