Fincom

Phonetic Fingerprint, the patented core, turns a name into a mathematical representation of its sound across 44 languages, a deterministic step rather than inference. Supervised machine learning in the entity resolution engine weighs date of birth, address and identifiers on top, and removing it would leave the patented engine standing.

Alerts go to compliance officers in the firm's case management system for adjudication, which fits a setting where blocking a payment or filing a report has statutory consequence. Thresholds, what proceeds when no alert fires and how a bank tunes sensitivity are not published.

Published measures include alert rates cut from an industry 30 to 50 percent to about 5 percent and screening under 200 milliseconds, and every hit carries a score from phonetic distance and weighted attributes. The claim of no missed hits has no recall figure or method, and no validation package maps to SR 11 7.

AgriBank, FCB reports false positive hits falling from 95 percent to about 8 percent, then to between 3 and 5 percent over the following months. That figure carries no method or baseline definition.

A deterministic phonetic core does not learn from customer traffic, so one bank's volume does not obviously sharpen a competitor's results. Whose adjudications label the supervised entity resolution layer, and whether tuning is per tenant, is not said.

The privacy policy dated February 2, 2022 covers the website only, names no subprocessors and sets out no individual rights. Customer names, dates of birth and identification numbers screened against watchlists, and the alert records that follow, get no described handling.

Awards such as RegTech100 and FinCrimeTech50 placements fill the homepage, but they are not security credentials. Neither a SOC 2 report nor an ISO certification or trust portal is named.

Screening is organized around OFAC sanctions obligations, with Verification of Payee and readiness for the 2026 Nacha mandate offered as products. Naming the rules served is a clear supplier position, though no register entry or filing channel stands behind it.

Transliteration from Arabic, Cyrillic or Chinese yields many valid spellings, so false positives fall hardest on people with non Latin names, and phonetic matching eases that burden. Such matching is product design, not governance: false positive rates by name origin and a model sign off process are unpublished.

Officer review before an outcome is final lets a wrong match be cleared and the payment released. Someone whose payment is held over a name resemblance is not told why and is given no route to suppress a recurring false match, and nothing stands behind the output.

In house builds of Phonetic Fingerprint and the entity resolution engine are described in technical detail, which keeps the model chain short. Watchlist data providers and hosting are not named, though list currency governs whether a screen is valid at all.

Nasdaq Verafin announced in June 2025 that it is building the screening into its anti financial crime platform, and the real time compliance system connects to Finastra's real time payments platform through the Boost API. Microsoft, Cognizant and Delta Capita also appear on the partners page.

Data is maintained in the EU and the US according to a privacy policy written for the website, and no hosting provider, region choice or private option is stated for the screening platform. European and American bank customers often set strict rules on where screening data is processed.

Savings are what gets published, more than 80 percent lower screening operating cost and alert handling cut from eight minutes to 30 seconds, while price and billing dimension stay behind a contact form. Those figures size the business case, not the bill.

Separate AML suites for financial institutions, trade finance and ecommerce each carry their own material, with Cadence Bank, The National Bank of Indianapolis and Jefferson Bank among named banks. Payment coverage spans wires, ACH and SWIFT messaging, with perpetual KYC and Verification of Payee among the products.