Fraudhalo came out of a specific problem: watching generic fraud rules cause as much merchant revenue damage as the fraud itself.
Nadia El-Sayed was running fraud operations at a mid-size Atlanta payment facilitator when a card-testing wave arrived in November 2023. The attack wiped out 22 percent of the month's chargeback budget in under two weeks.
The standard response was to tighten the rules engine. Nadia did that — and discovered the problem: every rule she added to stop one fraud pattern increased false-positive declines on legitimate small merchants by two to five percent. A landscaping company's busy Saturday looked indistinguishable from a carding run. A local restaurant's holiday spike flagged as velocity abuse.
The cure was causing as much revenue damage as the fraud itself.
The insight behind Fraudhalo: generic fraud rules are calibrated for high-volume card-not-present environments — enterprise card networks, large direct issuers. Applied to SMB processor portfolios, they systematically misread the irregular purchase sizes, seasonal spikes, and local geography that define legitimate small-merchant transactions.
The fix wasn't better generic rules. It was merchant-specific baselines.
A merchant-specific baseline calibration layer, tested at two regional payment facilitators in Georgia, reduced false-positive decline rates by roughly half within 45 days — while chargeback rates stayed flat.
Full real-time scoring API purpose-built for SMB payment processors, pairing baseline calibration with card-testing burst detection. Narrow focus, no ambitions to replace the processor's gateway.
Give SMB payment processors fraud detection that knows the difference between a suspicious transaction and a small-business Saturday.
Fraudhalo, Atlanta, GA — Founded 2023No two merchants have the same transaction baseline. Risk decisions that ignore merchant-specific context produce wrongful declines at scale. Every Fraudhalo model starts with per-merchant calibration, not a global threshold.
Sub-40ms is not a marketing claim — it's the constraint the card authorization window imposes. If the scoring decision doesn't arrive before the issuer response, it's not real-time fraud detection. We build to that constraint, not around it.
Every transaction Fraudhalo declines generates a structured risk explanation record. Risk operations teams shouldn't have to reconstruct why a decision was made when a merchant calls to dispute a decline or when a chargeback response is due.
Fraudhalo is built for SMB payment processors running 5,000–50,000 daily transactions. We're not trying to serve enterprise card networks or replace gateway infrastructure. Narrow scope means every feature traces back to an actual pain point in that specific operating context.
Nadia El-Sayed and Kofi Mensah founded Fraudhalo in late 2023. They were joined by Renata Kovacs and Omar Saleh — all with direct experience in SMB acquiring fraud.
Meet the team