Case Study: UK High Street Bank catches 83% of bad debt missed by credit scores with Kortical

Beating Traditional Credit Score with Machine Learning

UK High Street Bank wanted to test whether Machine Learning could outperform traditional credit scoring at predicting customer credit default. They partnered with Kortical and used Kortical’s AI platform to ingest anonymised transaction and CRM data, automatically transform it, create features and build thousands of ML models from a subset of the bank’s customer history.

Kortical delivered a production-ready ML model in four weeks that caught 83% of bad debt missed by the traditional credit score while declining loans to the same number of applicants; alternatively the model could be reversed to offer loans to 77% more people at the bank’s current default rate. Kortical’s platform handled over 220 million rows, accelerated feature engineering and provided explainable insights (different drivers by age group), enabling a major reduction in credit risk or a substantial increase in lending.