Malaysian banks can now assess small business loans using transaction history instead of credit scores

Researchers built a machine learning system that scores loan risk by analyzing bank statements instead of credit bureau data, then tested it on 611 Malaysian small businesses. The approach works: models trained on transaction patterns were 25% better at predicting default risk than models using application data alone, suggesting banks have a new way to approve loans for businesses that have no credit history.

Small and medium businesses make up 96% of Malaysian firms but are locked out of formal lending because they lack credit scores. This research shows a concrete, measurable alternative: transaction data already exists in banks' systems and is predictive. The practical implication is immediate—any bank with access to customer statements can deploy this system tomorrow, without waiting for policy change or new infrastructure. The constraint was never the data; it was knowing the data worked.