Fuzzy Matching in Sanctions Screening: How It Works

Why Exact Name Matching Fails

Sanctions lists contain names from every language and writing system. When transliterated into Latin characters, multiple valid spellings exist. A simple exact-match search will miss dangerous matches.

Common Name Variations

Arabic names can be transliterated in dozens of ways. Russian Cyrillic maps differently depending on the system. Chinese names appear in Pinyin, Wade-Giles, or other romanizations. Even European names vary across languages.

How Fuzzy Matching Works

Modern screening tools combine multiple algorithms: trigram similarity compares three-character sequence overlap, Levenshtein distance measures minimum edit operations, and Soundex and Metaphone compare phonetic similarity rather than spelling.

Setting the Right Threshold

Every fuzzy matching system has a confidence threshold. Setting it too high misses matches; too low generates excessive false positives. Most programs use 70 to 85 percent, adjusting based on risk appetite.

Isarud Approach

Isarud combines trigram similarity, Levenshtein distance, and phonetic matching. Results are scored 0 to 100 percent, with matches above 85 percent flagged high-confidence and 70 to 85 percent flagged for manual review.