Name matching and fuzzy logic
Listed names never arrive spelled the same way twice. Fuzzy matching is how the control catches variation — at the price of innocent lookalikes.
L0 Explain simply
An everyday analogy: names are unreliable labels. The same person can be written Mohammed, Muhammad, or Mohamed depending on who transcribed the name from Arabic; family names come first in some cultures and last in others; a typist can drop a letter anywhere. A gatekeeper who insists on letter-perfect spelling will miss the very person the poster describes, because posters and visitor records are written by different hands. So the gatekeeper is trained to treat close spellings and similar sounds as worth a second look. The cost is honest and unavoidable: more innocent people with similar names get stopped for a moment. Screening accepts that trade deliberately — a check that only catches perfect spellings is barely a check at all.
L1 Core concepts
Exact string comparison fails for reasons built into the data, not because anyone did anything wrong. Transliteration — rewriting names from Arabic, Cyrillic, Chinese, and other scripts into Latin letters — has no single correct answer, so one person legitimately appears under many spellings. Naming conventions differ: patronymics, name order, compound family names. Payment data adds abbreviation and truncation. Fuzzy matching answers this: algorithm-based techniques that match a name whose spelling, pattern, or sound is close to a list entry without being identical. List issuers help from their side by publishing aliases and original-script names, which is exactly why screening engines match against every alias on an entry rather than only the primary name — a hit on an alias is how a known alternate name of a target gets caught.
L2 Practitioner view
The practitioner's dial is the match threshold. Each comparison produces a similarity score; alerts fire above a configured line. Tighten it and alert volume falls but the risk of missing a genuinely misspelled target rises; loosen it and the queue fills with namesakes. There is no universally right setting — the calibration is a documented risk decision, usually differentiated by list, by data type, and by field: a match on a passport-number field warrants different handling than a similar-sounding word in free text. Weak aliases are commonly configured to enrich investigations rather than alert alone. The two linked scenarios show both directions of the same machinery: an innocent name collision that must be cleared with evidence, and a true match caught only because the engine screened an alias.
L3 Technical details
Under the hood, matching is a pipeline, and every stage is a control decision. Normalisation strips case, diacritics, punctuation, titles, and legal-form suffixes so that superficial differences stop mattering. Tokenisation splits names into parts so ordering differences can be compared. Then similarity techniques do the scoring: edit-distance measures count character changes between strings, phonetic algorithms compare how names sound, and n-gram methods compare overlapping fragments. Engines combine several, weighted by field. Each technique widens what the filter can catch and admits more lookalikes, which is why institutions define the variation classes they intend to catch — transliteration variants, transpositions, missing name parts — and test against constructed examples of each. A capability nobody has verified is an assumption, not a control.
Sources & standards1
- Market practice
Wolfsberg Group Sanctions Screening Guidance ↗ — The Wolfsberg Group · Fuzzy matching definition; screening technology considerations
Wolfsberg guidance is industry market practice, not law; institutions vary in how they apply it.
L4 Standards & sources
What the authoritative material actually establishes: the Wolfsberg Group guidance defines fuzzy matching as algorithm-based techniques for matching names whose spelling, pattern, or sound is close but not identical, and frames threshold calibration as a risk-based decision the institution must be able to explain — no regulator or standards body prescribes a specific algorithm or numeric threshold. On the data side, the UN consolidated list publishes names split across multiple name fields with original-script versions and quality-graded aliases, and OFAC's SDN List likewise carries alias records — the raw material fuzzy matching works on. The honest summary: effectiveness expectations are principles-based, the list data is official, and the matching mathematics in between is the institution's documented, testable responsibility.
Sources & standards3
- Market practice
Wolfsberg Group Sanctions Screening Guidance ↗ — The Wolfsberg Group · Definition of fuzzy matching; risk-based calibration of screening technology
Wolfsberg guidance is industry market practice, not law; institutions vary in how they apply it.
- Official requirement
United Nations Security Council Consolidated List ↗ — United Nations Security Council · Name fields, original-script names, and a.k.a. quality grading in list data
Published in XML, HTML, and PDF formats and updated as committees act; each name is subject to the measures of its specific sanctions committee, not one uniform regime.
- Official requirement
Specially Designated Nationals and Blocked Persons List (SDN List) ↗ — US Department of the Treasury, Office of Foreign Assets Control · Alias (a.k.a.) records on SDN entries
Updated continuously; machine-readable formats are distributed via OFAC's Sanctions List Service. Every list entry appearing in this site's examples is fictional.
Sources for this topic3
- Market practice
Wolfsberg Group Sanctions Screening Guidance ↗ — The Wolfsberg Group · Screening technology and fuzzy matching
Wolfsberg guidance is industry market practice, not law; institutions vary in how they apply it.
- Official requirement
United Nations Security Council Consolidated List ↗ — United Nations Security Council · Original-script names and alias grading
Published in XML, HTML, and PDF formats and updated as committees act; each name is subject to the measures of its specific sanctions committee, not one uniform regime.
- Simplified educational illustration
Payments Signal editorial teaching models — Payments Signal
What this simplifies: Matching pipelines are simplified to a linear sequence and algorithm families are named without vendor specifics; real engines combine techniques in proprietary ways. Name examples illustrate transliteration variation only, and all matched parties in the linked scenarios are fictional. This topic explains how controls catch name variation; it deliberately offers no guidance on producing it.
Used wherever diagrams, scenarios, figures, or example values are didactic constructions rather than sourced facts; every such use carries a simplifications disclosure. All people, companies, banks, and list entries in examples are fictional.
Deepest material on this page: L4 — Standards & sources. Where a topic stops short of implementation depth, that is a deliberate coverage decision, not an oversight — see coverage.