Fraud Protection & Compliance
As volumes of account-to-account consumer payment transactions grow — such as person-to-person (P2P) transfers made via Zelle, which is forecast to take over Venmo in 2018 — so does the opportunity for fraudsters to exploit demand deposit accounts (DDA). With a view to stopping real-time payments fraud, FICO recently released the Retail Banking Consumer fraud model.
The new model was built on over a decade of real-time payments consortium data and leverages FICO’s deep analytic expertise in building custom account-to-account retail banking models. Beyond P2P transactions, the Retail Banking Consumer model applies fraud analytics to DDA consumer payments processed on traditional channels such as internet and mobile banking.
Consortium Data Speeds Model Adoption and Efficacy
While custom fraud models are based only on a single set of bank-specific data, the Retail Banking Consumer 1.0 model incorporates the rich global data lake of retail banking activity within the FICO® Falcon® Intelligence Network. We applied behavioral analytics on this ample and diverse set of worldwide payment activity and accumulated fraud scenarios.
Institutions new to the FICO® Falcon® Platform can rapidly implement the model with very little upfront historical data requirements when they agree to join the consortium. They don’t need to provide extensive months of retail banking data, or commit to a complex model design and delivery process. Once they are part of the Falcon Intelligence Network, they can use the consortium model, and over time that bank’s data will be incorporated into model refreshes, allowing the consortium model to become more tuned to that institution’s specific transaction and fraud patterns. Regularly scheduled retrains occur at no cost to the client, allowing the model to remain robust while incorporating the latest innovations from FICO’s analytics team.
Advanced Real-Time Payments Fraud Analytics Under the Hood
While using the new models are easy, the model itself is extremely sophisticated. It includes:
- Multi-level profiling: Profiling of both ends of the transaction, compactly representing the rich history of behavior of the payer at the account and customer level, as well as the behavior of the beneficiary account.
- Global Intelligent Profiling: A fixed-size profiling only of the most relevant beneficiary accounts outside of the client’s banking system.
- Behavior Sorted Lists: Tracking of recurring behavioral distinct payment features such as preferred beneficiary countries or beneficiary accounts for a given debiting account or customer, and establishing their favorites to better identify unusual payment behavior.
- Multi-Layer Self Calibrating analytics: FICO’s self-calibrating machine learning technology allows the model to dynamically adjust to changes in predictive variable trends in the real-time. This keeps the model robust and relevant in an ever-changing payment space.
- Peer grouping: Grouping transactions/accounts with similar behavior and risk features, to help identify truly abnormal behavior in the context of behavioral soft clustering. This feature compares outliers against the distribution of activity within a given computed peer group.
These features allow FICO to use its retail banking consortium data and advanced analytic technology to help financial institutions defend against real-time payments fraud and other DDA-based payment types.
To keep up with my latest thoughts on fighting fraud with advanced analytics, artificial intelligence and ML, follow me on Twitter @ScottZoldi.