How Fuzzy Matching Works in Sanctions Screening

Why Exact Matching Is Not Enough

If you only search for exact name matches against the SDN List, you will miss a lot. Names can be transliterated from other languages in multiple ways. People use nicknames, abbreviations, and alternate spellings. Typos happen. A sanctioned individual named "Aleksandr" might also appear as "Alexander," "Aleksander," or "Alexandr." Exact matching would miss all of those variations.

This is why sanctions screening tools use fuzzy matching, a set of techniques that find names that are similar but not identical to an entry on a sanctions list.

Trigram Similarity

Trigram matching is one of the most effective techniques for sanctions screening. It works by breaking names into overlapping groups of three consecutive characters (trigrams). For example, "Smith" becomes: "smi", "mit", "ith". Two names are considered similar when they share a high percentage of trigrams.

Trigram similarity is good at catching spelling variations, transpositions, and missing or extra characters. It handles the kinds of differences you commonly see in sanctions data, where names have been transliterated from Arabic, Chinese, Russian, or other scripts.

Phonetic Matching

Phonetic algorithms identify names that sound alike, even if they are spelled very differently. The most common phonetic algorithms include:

• Soundex: An older algorithm that groups names by their consonant sounds. It is simple but limited.
• Metaphone and Double Metaphone: More sophisticated algorithms that handle a wider range of pronunciation rules, including non-English names.

Phonetic matching is especially useful for names transliterated from languages that use non-Latin scripts. "Qaddafi" and "Gadhafi" look very different in writing, but phonetic matching recognizes that they sound similar.

Edit Distance (Levenshtein Distance)

Edit distance measures how many single-character changes (insertions, deletions, or substitutions) are needed to transform one string into another. "Smith" and "Smyth" have an edit distance of 1. "Johnson" and "Jonson" also have an edit distance of 1.

Edit distance is straightforward and catches simple typos and minor spelling differences. It is less effective for names with major structural differences, which is why it works best in combination with other techniques.

Balancing Sensitivity and Specificity

The challenge with fuzzy matching is finding the right balance. If your matching threshold is too loose, you will get flooded with false positives. Every "John" will match every "Jon," and your compliance team will spend all day reviewing noise. If your threshold is too tight, you will miss legitimate matches, and that creates real regulatory risk.

Most screening tools allow you to adjust the matching sensitivity. The right setting depends on your risk tolerance, your volume of screening, and your capacity to review results. OFACScreen lets you set your own threshold, and we default to a level that balances thoroughness with practicality.

Best Practices for Fuzzy Matching

• Use multiple techniques together. No single algorithm catches everything. Combining trigram, phonetic, and edit distance matching gives you better coverage.
• Test your settings. Run known sanctioned names through your screening tool and make sure they are flagged. Also test common names to understand your false positive rate.
• Review and adjust. If you are getting too many false positives, tighten your threshold slightly. If you are not getting any hits at all, that is a red flag that your threshold may be too high.
• Document your methodology. Regulators want to understand how your screening works. Be prepared to explain your matching approach and why you chose your sensitivity settings.

OFACScreen uses a combination of trigram similarity and phonetic matching to deliver accurate results with a manageable false positive rate. Our matching algorithms are tuned for sanctions data specifically, so they handle transliterated names and international characters well.