Fuzzy Matching
Definition
A string-comparison technique that identifies near-identical records by measuring edit distance or phonetic similarity rather than requiring character-exact matches. Used in duplicate detection to catch invoice numbers or vendor names altered by one or two characters to defeat exact-match controls.
- Method
- Edit distance or phonetic similarity comparison
- Contrast
- Exact character-matching
- Use case
- Catching altered invoice numbers or vendor names
Common questions
How does fuzzy matching catch fraud that exact-match rules miss?+
A fraudster can create a near-duplicate vendor record by transposing two digits or adding a stray character, defeating a rule that checks for exact string equality, but fuzzy matching flags records within a small edit distance of an existing entry so the near-duplicate still surfaces for review.
What is the main operational drawback of fuzzy matching in continuous monitoring?+
It generates more false positives than exact matching because legitimately similar but distinct names or numbers also get flagged, so investigators need a review workflow and a tuned similarity threshold to keep the alert volume manageable.
Related terms
- ACL / Galvanize HighBond
- A purpose-built audit analytics platform (originally Audit Command Language) that imports financial data, executes statistical and rule-based tests, and produces exception reports...
- Benford's Law
- An empirical regularity in naturally occurring numerical datasets: the leading digit follows a logarithmic distribution, with 1 appearing about 30% of the...
- Continuous Monitoring
- An automated control framework that applies fraud indicator tests to transactions as they are processed or on a frequent scheduled basis, generating...
- IDEA
- Interactive Data Extraction and Analysis: an audit data analytics tool that supports Benford analysis, duplicate detection, stratification, and custom query filters across...
- Network Link Analysis
- A technique that represents entities (vendors, employees, bank accounts, addresses) as nodes in a graph and shared attributes as edges, enabling investigators...