OFAC Fuzzy Matching Glossary Part 1
Core terms for approximate name matching that catches sanctions list entries despite spelling differences.
TL;DR
TL;DR: Fuzzy matching is approximate string matching that catches a name even when it is spelled, transliterated, or punctuated differently from a listed entry. It trades more false positives for fewer dangerous false negatives.
What fuzzy matching is
Fuzzy matching compares two strings and scores their similarity rather than demanding an exact match. This matters for sanctions screening because a listed name can appear in many forms: different transliterations from non-Latin scripts, reordered name parts, typos, and missing punctuation. An exact-match engine would miss those variants. A fuzzy engine catches them but also flags lookalikes that are not actually listed, known as false positives, which a screening workflow must then handle.
Why it matters for agent payments
For an agent that pays counterparties automatically, a missed match is the catastrophic error, because OFAC enforces strict liability and penalties start at $356,000 per violation. A fuzzy matching engine that errs toward false positives is the safer failure mode: a legitimate payment can be reviewed and released, but a missed blocked party cannot be undone. The sanctionsai.dev API combines fuzzy matching against 19,218 SDN names with exact wallet screening over 947 listed addresses, returning clean or flagged in under 100 ms.
Terms covered in Part 1
- Fuzzy match: a similarity-based, not exact, comparison.
- Transliteration: converting a name between writing systems.
- False positive: a non-listed name flagged as a possible match.