Raw rows loaded
525,461
Online Retail Customer Segmentation
RFM-based clustering with KMeans baseline and GMM comparison
Rows after cleaning
407,664
Customers segmented
4,312
Chosen clusters (K)
3
Model Selection
| Model | Silhouette Score | Interpretation |
|---|---|---|
| KMeans (k=3) | 0.4117 | Best baseline |
| GMM (k=3) | Run notebook cell for exact value | Secondary comparison |
Cluster Profiles (Relative RFM)
| Cluster | Segment Name | Recency | Frequency | Monetary | Size |
|---|---|---|---|---|---|
| 0 | Champions | 34.06 (low) | 2.18 (high) | 8.00 (high) | 1,361 |
| 1 | At-Risk | 249.76 (very high) | 0.85 (low) | 5.59 (low) | 955 |
| 2 | Potential Loyalists | 54.24 (mid) | 1.07 (mid-low) | 6.16 (mid) | 1,996 |
Key Plots
Customers per Segment
Average Revenue per Customer by Segment
Customer Segments (PCA projection)
Recommended Actions
- Champions: offer early access, loyalty perks, and referral incentives.
- Potential Loyalists: drive 2nd/3rd purchase with bundles and personalized cross-sell.
- At-Risk: run win-back campaigns with stronger offers and reactivation journeys.
Project Outputs
- Notebook:
notebooks/01_customer_segmentation_starter.ipynb - Export file:
data/processed/customer_segments.csv - Environment:
environment.yml
Note: Frequency and Monetary values in clustering were log-transformed for modeling stability. Use the exported file for business-facing dashboards and campaign execution.