Credit Scoring

Industry Finance
Potential industries Finance, Banking, Insurance
Client SME-Lending Company

Summary

The quality of the underwriting process is critical for the successful business of lending. Complete automation of the process with a machine learning model that takes a decision significantly improved the loan portfolio quality and increased the speed of decision-making. We developed a model to predict events of borrower default. The model demonstrated superior to human performance and improved the health of the loan portfolio. It also enabled a loan borrower’s positive experience due to quick decisions on their loan application.

Challenge

The customer requested to automate the underwriting process and challenged us to provide machine learning models superior to an underwriter’s expert performance. The model was developed and tested on a limited set of loan applications. External sources of information about creditworthiness had to be tested and proposed.

Solution by AI Superior

We developed an AI component to fully automate the underwriter process. More than 800 features were extracted from 14 different data sources. The component is based on a state-of-the-art machine learning approach to predict the probability of default for a potential borrower.

Outcome and Implications

The solution boosts the portfolio performance due to the drop in the number of accepted customers with defaulted loans while maintaining the overall acceptance rate of new customers. As well, developed solutions significantly increased the speed of decision-making, reducing the process from a few hours or a day to a mere fraction of a minute.