Stockertim Dictionary · Artificial intelligence · Accounting
Fuzzy matching
Also called approximate matching, fuzzy match
In short
Fuzzy matching is matching records that are similar but not identical, such as a slightly different payee name, reference or amount, usually producing proposed matches a person approves.
- Intermediate
- Process
- Checked
In detail
Fuzzy matching pairs records that refer to the same thing but do not agree exactly. A bank line reading "HARBOUR FDS LTD INV 4471" and a ledger entry for Harbour Foods invoice 4471 are the same payment, yet an exact match on the text fails. Fuzzy matching measures how close the two are and proposes a pair.
It sits between exact and rule-based matching. Exact matching needs identical values; rule-based matching uses set rules, such as same amount within three days. Fuzzy matching scores similarity in names, references, dates and amounts using string distance measures or machine learning, then ranks candidate pairs by score.
It matters in reconciliations because real data are messy: abbreviated names, typing errors, bank charges deducted from payments, and one payment covering several invoices. Fuzzy matching raises the share of items matched automatically and saves review time. It also creates a risk: two similar but different items can be paired wrongly, hiding a duplicate payment or an error.
For that reason fuzzy matches usually go to a person for approval above a set level of difference, and approved matches are recorded with who approved them.
Worked example
Lagoon Logistics reconciles a supplier payment that does not match exactly.
- Bank line: "LAGN LOGSTCS PAY REF 2210" for 4,985.
- Ledger: invoice 2210 to the same supplier for 5,000.
- Difference: 5,000 - 4,985 = 15, matching the supplier's usual bank charge.
- The tool proposes the pair with a high similarity score, and a person approves it and posts the charge.
The payment is matched and the 15 difference explained, rather than left unmatched or matched silently.
In practice
Bank reconciliation, intercompany matching, supplier deduplication and customer screening all use fuzzy matching, with thresholds that decide which proposed pairs need a person.
Watch out
- A high similarity score is not proof the two records are the same transaction.
- Fuzzy matching can hide duplicates if two near-identical payments are both paired to one invoice.
Learn it properly
Official sources
- Technology · International Auditing and Assurance Standards Board
- AI Risk Management Framework · National Institute of Standards and Technology
Facts last checked on 3 November 2026.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.