Churn & Cross-sell Prediction in Leading Bank
Problem
- Lack of integrated customer data
- High churn ratios and lack of insight about the reasons of customer disengagement and inactivity.
- Need for predictive action plan for high churn risk customers.
- Need to deliver targeted offers for current customers from product portfolio.
Action
- Our scope was to design and implement an analytics datamart and apply churn prediction and cross-sell projects
- Designed and implemented analytics datamart which integrates customer, product and sales transaction data
- Built churn prediction models for credit cards
- Built crossell model for consumer credit
- Implemented an ML-Ops pipeline for managing machine learning models
Tool Stack
- SQL Server, Python, Airflow
70
%Churn Model Coverage Ratio
15
%Potential decrease in high churn risk
20
%Potential increase in crossell offers