Automated credit management with predictive analytics

Banking

Business

Large
~ 250 employees

Type of solution

Automation and risk model for credit assessment

- 30%

losses due to default on consumer loans

Challenge

Credit process with manual validations, long response times and a rating based on an expert scorecard of fixed rules, with limited ability to discriminate between good and bad customers and adjust risk appetite.

Solution

An end-to-end solution was developed that replaced the expert scorecard with a predictive model (XGBoost), trained with proprietary data, macroeconomic context and credit history from the risk center.

Technologies used

Python · XGBoost · AWS API Gateway · MLflow

Impact

Better discrimination between good and bad customers, reducing losses due to default. Greater efficiency in origination: more customers are accepted within the risk appetite, and fewer valid opportunities are missed.

Related services

The capabilities that made this project possible.

Advanced Analytics

Data Engineering