What threshold should we run at?

There is no universal figure, and anyone quoting one is guessing about your book. The right threshold depends on the name distribution you screen, the identifiers you hold and your review capacity. What matters to a supervisor is that it was set deliberately, tested and documented.

Why do we get so many candidates?

Because sanctions lists carry many short and common names with few identifiers, and because looser matching is the deliberate choice that prevents misses. Improving input data usually cuts candidate volume more than tuning the algorithm does.

Does fuzzy matching cope with Cyrillic or Arabic?

Only if transliteration is handled explicitly. Comparing a Latin-script customer name against a Cyrillic or Arabic list entry needs a transliteration step first; without it the strings never come close enough to score.

SCREENING GLOSSARY · GUIDEUPDATED 2026-09-04

Screening Glossary

What is fuzzy matching in name screening?

Fuzzy matching is name comparison that tolerates difference. Instead of demanding identical strings it scores similarity, so spelling variants, reordered words, missing middle names, typing errors and transliteration differences still raise a candidate.

Exact matching fails on real data almost immediately, because names arrive from passports, invoices, registries and payment messages in inconsistent forms. Fuzzy matching combines edit distance, phonetic encoding, token reordering and name-specific dictionaries. The trade-off is inescapable: looser matching finds more true matches and produces more noise, tighter matching produces a quieter queue and more misses. That is a policy decision dressed as a technical one.

Run screening View pricing

What this workflow covers

SCOPE

Set the threshold deliberately, then let analysts confirm or dismiss on identifiers rather than on the score.
Aleksandr Ivanovich Petrov, A. I. Petrov and Alexander Petrov should all raise the same entry. Exact matching returns one of them.
Normalise GmbH, Sp. z o.o., LLC, Holdings and Group out of company names before comparison, or they dominate the similarity score.
A match score is not a probability that the person is sanctioned, and it is not a disposition.

Compliance glossary

TERMS

Edit distance: A measure of how many single-character changes are needed to turn one string into another; a common component of fuzzy name matching.
Phonetic encoding: Reducing a name to a code representing how it sounds, so differently spelled but similar-sounding names can be compared.
Normalisation: Stripping punctuation, legal-form suffixes and case differences before comparison so the scoring reflects the distinctive part of the name.

Authoritative references

SOURCES

01
A Framework for OFAC Compliance Commitments
U.S. Department of the Treasury — OFAC
02
Specially Designated Nationals and Blocked Persons List (SDN) — Human Readable Lists
U.S. Department of the Treasury — OFAC

Frequently asked questions

Q&A

What threshold should we run at?: There is no universal figure, and anyone quoting one is guessing about your book. The right threshold depends on the name distribution you screen, the identifiers you hold and your review capacity. What matters to a supervisor is that it was set deliberately, tested and documented.
Why do we get so many candidates?: Because sanctions lists carry many short and common names with few identifiers, and because looser matching is the deliberate choice that prevents misses. Improving input data usually cuts candidate volume more than tuning the algorithm does.
Does fuzzy matching cope with Cyrillic or Arabic?: Only if transliteration is handled explicitly. Comparing a Latin-script customer name against a Cyrillic or Arabic list entry needs a transliteration step first; without it the strings never come close enough to score.

Screening Glossary Transliteration Alias (AKA)False positive Sanctions screening Customer screening