Visa folds Featurespace scoring into its A2A fraud service for banks

Visa has rebuilt A2A Protect around a single fraud score drawn from Featurespace models, aimed at stopping account-to-account scams before a transfer is authorised. It is the first Featurespace-derived capability Visa has put into the market since buying the company. What changed A2A Protect now returns one unified risk score per account-to-account transfer instead of separate signals, delivered to banks over a single API. Visa says the model uses transfer learning so a joining institution gets network-level risk intelligence immediately rather than waiting months for a model to train on its own transaction history, and without needing other banks to join a data consortium first. Banks that opt in to network-level intelligence sharing additionally receive signals on emerging scam hotspots and coordinated fraud activity seen elsewhere on the network. Each alert carries a plain-language reason for the flag. Visa cites internal results of over 50% more fraud caught, over 40% fewer unnecessary alerts, and a 75% lift in detection during the first six months of deployment. The service scores bank-to-bank transfers; it is not a change to Visa card scheme rules or interchange.