The client needed to segment customers to enhance risk-based monitoring and apply targeted thresholds based on behavioral patterns and risk profiles.

Developed a hybrid customer segmentation model.

1. Engagement with Stakeholders: Developed an initial segmentation model frame, identified business-driven components to start segmentation model development.

2. Segmentation Process: Used a combination of top-down (expert/domain-driven) and bottom-up (data-driven) approaches, applied machine learning algorithms (K-means clustering, Hierarchical clustering) to demographic and transactional data.

3. Model Deployment: Deployed the model using a cloud-based ML pipeline (AWS/Azure), set up governance and control framework for periodic runs.

4. Visualization and Testing: Used graphical and statistical tests (density distribution, histogram, box-plot) to interpret clusters and validate segmentation effectiveness.

Improved risk-based monitoring and increased productive alerts.

- The segmentation allowed for effective risk-based monitoring with targeted thresholds, increasing the percentage of productive alerts.

- The hybrid approach enabled the bank to monitor customers more effectively, resulting in better risk management and compliance.