The Bank for International Settlements Innovation Hub published findings from Project Hertha, a joint experiment with the Bank of England, on whether transaction analytics in real-time retail payment systems can help identify coordinated financial crime while using a minimal set of data points to preserve user privacy. The project concludes that payment system analytics could serve as a supplementary tool for banks and payment service providers (PSPs) to improve detection of suspicious activity. Testing on a state-of-the-art synthetic dataset covering 1.8 million accounts and 308 million transactions, banks and PSPs that incorporated network-level indicators derived from payment system data identified 12% more illicit accounts than they would have found using their own monitoring alone, with a 26% improvement for previously unseen typologies. Payment system-only monitoring was slightly less effective than bank/PSP monitoring overall (39% versus 44% of illicit accounts identified), but was most valuable for complex schemes spanning many accounts across multiple institutions. The work assumed no change to existing institutional responsibilities and no sharing of private customer data with the payment system operator, and it underlined the importance of labelled training data, a robust feedback loop from bank investigations, and explainable AI to support investigations and reporting. The report notes that deploying comparable solutions in practice would raise practical, legal and regulatory issues that were out of scope, and it highlights potential follow-on work on transaction tracing, collaborative investigations, and extending the approach to cross-border and large-value payment systems and to cryptoasset networks.
Bank for International Settlements - Innovation Hub 2025-06-05
Bank for International Settlements Innovation Hub and Bank of England find payment system analytics can help banks identify 12% more illicit accounts
The BIS Innovation Hub and Bank of England’s Project Hertha found that real-time retail payment system analytics can supplement banks’ and payment service providers’ monitoring to better detect coordinated financial crime while using minimal data to preserve privacy. Testing on a large synthetic dataset, adding network-level indicators from payment system data enabled identification of 12% more illicit accounts overall and 26% more for previously unseen typologies, particularly complex schemes across multiple institutions, though payment system-only monitoring remained slightly less effective than institutional monitoring.