The score acts before funds leave
Visa launched an enhanced version of A2A Protect that adds a unified fraud score powered by Featurespace technology. The product evaluates suspect account-to-account transfers before funds leave an account and is delivered through a single API. [1] [2]
The system can incorporate opt-in network-level risk signals alongside an institution’s own behavioural information. Visa says the service can provide model value without requiring a bank to first accumulate a large historical dataset, then adapt through continuous learning in that environment. [2]
The reported results need careful attribution
Visa cites a Pay.UK case study in which its data detected 54% of fraud and authorised-push-payment scam value beyond existing bank-system detection, alongside a 40% reduction in false positives. [2]
These are vendor-reported case-study outcomes, not independently audited industry benchmarks or guaranteed results. Performance will depend on data quality, payment mix, thresholds, governance and how each institution handles review and customer friction. [1] [2]
Shared intelligence changes the control model
Instant account-to-account payments compress the time available for intervention. Pre-transfer scoring can add decision time, while network signals may reveal beneficiary or behavioural patterns that one bank cannot observe alone. [1] [2]
That benefit also raises governance questions around explainability, consent, false positives, model change and human escalation. Institutions need evidence that the score improves their own control environment rather than relying only on a vendor case study. [1] [2]