The client is an Indian private sector bank headquartered in Mumbai and offers services across six verticals: corporate and institutional banking, commercial banking, branch and business banking, retail assets, development banking and financial inclusion, treasury and financial market operations. As of December 2020, it has a network of 372 branches with 5,843 employees and 394 ATMs across 28 states and union territories.
Created two models to identify customers who would flow to the next bucket (who fails to pay even in the next month) or remain stable / normalise (pay the pending amount or clear previous dues).
For customers who flow to the next bucket:
Segmented the defaulting customers into 5 different bands (from very low risk to very high risk) based on the risk score from the model
The output of the model is used by the collections and strategy team to uses a feedback technique called “test and learn” where the client assigns different strategies for different cases/buckets and run a test. This helps the client to recover its account receivables more efficiently.