Pay.UK and Featurespace set out on a mission together – to help fight fraud by bringing together the combined power of data from multiple high street banks and payment service providers for the first time.
Pay.UK and Featurespace set out on a mission together – to help fight fraud by bringing together the combined power of data from multiple high street banks and payment service providers for the first time.
Featurespace achieved 43% value detection at an industry leading 5:1 false positive ratio on the fraud which evaded the bank’s existing controls. The pioneering data sharing consortium approach offered differing uplift depending on the size and type of the consortium participant but can be approximately encapsulated as offering between 5 and 15% uplift to value detection. Based on UK Finance data*, operationalising these results for the whole network would keep an additional £279 million from ever reaching criminals’ accounts.
Post completion of the pilot, the data was processed through TallierLTM™, Featurespace’s Generative AI innovation, and was able to increase fraud detection to 56% at a 5:1 ratio. At a 10:1 ratio, the technology caught 68.8% of missed fraud, and over 70% at a 10.4:1 ratio.
It is clear using this new approach has positive long-term implications for fraud detection.
Featurespace has been building payment fraud models for a decade. Leveraging our work with other ACHs/consortiums we have a platform that applies our cutting edge real-time results.
Register your interest in how you can integrate to this platform
*UK Finance: (APP) losses reach £485.2 million with TPF reaching £163.1m in 2022.
The Featurespace solution used in the PoC only requires information about the payment details such as who paid, who received the payment, the amount, and the date of the Faster Payment message. It does not require sensitive personal information such as the account holder's name or address, or information on free text payment reference fields.
Future upgrades for the Fraud Overlay system
of the consortium will include TallierLTM™, which
is a world-first Foundation Model for payments developed by Featurespace. This technology uses Generative AI to predict the future transactions that consumers are likely to make. When TallierLTM was integrated into the Fraud Overlay model during the Proof of Concept phase, the fraud value catch increased from 43% to 56% when operating at 5:1, without using any personally identifiable data.
At a 10.4:1 ratio, the technology caught over 70% of missed fraud
