Realta launches adaptive fraud routing to improve payment risk

By Divya Shah

Realta has launched Adaptive Routing, a new fraud decisioning architecture that directs payment transactions to the risk models best suited to evaluate them, helping financial institutions improve fraud detection while reducing false positives.

The platform is already delivering results for a U.S.-based payments processor, identifying three times more fraud while reducing false alerts by 89%.

Traditional fraud systems typically route different transaction types through the same decision-making process, regardless of characteristics such as transaction value, payment channel or merchant type. Realta said this approach often leads to increasingly complex rule sets that become difficult to maintain and can contribute to false declines and customer friction.

The new Adaptive Routing architecture evaluates multiple transaction attributes, including payment amount, merchant category, currency and channel, within a single assessment. Payments are then routed to domain-specific risk models designed to analyse particular fraud scenarios.

A component of the platform is the Eclipse Agent, which filters false positives in real time before alerts reach fraud analysts. Realta said risk decisions are returned in less than 100 milliseconds while maintaining audit-ready records for compliance and governance purposes.

The platform also aims to improve transparency in fraud investigations by providing analysts with visibility into the weighted risk categories and specific rules that contributed to each decision, rather than relying solely on a single risk score.

In addition, Adaptive Routing integrates with Realta’s unified decisioning engine, enabling organisations to share intelligence across fraud prevention, underwriting and anti-money laundering (AML) workflows throughout the customer lifecycle.

The company said the approach helps organisations increase transaction approvals and operational efficiency without significantly increasing fraud losses or review workloads.