Fuzzy matching
Also called fuzzy logic matching, approximate name matching
Fuzzy matching is a screening technique that compares names and other data against sanctions and watchlists and flags close but inexact matches. It catches spelling variations, transliterations, typos, reordered names and missing characters, giving each potential match a similarity score that is compared with a set threshold.
How it works
Screening systems break names into parts and compare them using algorithms such as edit distance, phonetic codes and token matching. Each comparison returns a score. If the score passes the firm's threshold, an alert goes to an analyst, who checks secondary data such as date of birth, nationality or address before clearing or escalating the hit.
The threshold is a trade-off. Set it too high and a designated person spelt differently slips through. Set it too low and analysts drown in false positives. Regulators expect firms to justify their settings, test them against known name variations and review them when lists or customer populations change. Matching on Arabic, Cyrillic and Chinese names needs particular care.
The rules
Requires a sanctions filtering programme with end-to-end testing, including review of data matching and whether threshold settings map to the institution's risks.
New York-regulated banks, check cashers and money transmittersDescribes testing and auditing of screening as an essential component of a sanctions compliance programme, and cites screening software weaknesses as a root cause of violations.
US persons and firms dealing in US-nexus transactionsCommon questions
Why do firms need fuzzy matching?
Names on sanctions lists appear in many spellings and scripts, and customers' data contains typos. Exact matching would miss these variations. Fuzzy matching detects close matches so a designated person cannot avoid screening through a small spelling change.
What is a fuzzy matching threshold?
It is the minimum similarity score at which the system raises an alert. A lower threshold catches more variations but produces more false positives. Firms set and document thresholds based on their risk and test them regularly.
How do analysts resolve fuzzy match alerts?
Analysts compare secondary identifiers such as date of birth, nationality, address and registration numbers with the list entry. If these confirm a different person, the alert is cleared with a recorded rationale. If not, it is escalated.
