Guide · Matching

Reconciliation matching rules: exact, tolerance and fuzzy narration, and when to use each.

Auto-match rates are decided by rules, not software. Good rules match the obvious items safely and leave only real exceptions for people.

Three kinds of rule

RuleHow it matchesUse it forRisk
ExactSame reference, amount and dateClean, high-volume flows with shared IDsLow
ToleranceAmount within a set difference, or date within a set windowCharges, rounding, FX, posting lagsMedium: set limits carefully
Fuzzy narrationSimilar references or narrations, after cleaningBank statements where narrations are truncated or reformattedHigher: needs review and limits

Order matters

  1. Run exact rules first and take those items out of the pool.
  2. Run tolerance rules next, narrowest tolerance first.
  3. Run fuzzy rules last, and send low-confidence matches to review.
  4. Handle one-to-many and many-to-one matches, such as a bulk settlement against many transactions, as their own rules.

Govern the rules

How AutoRecon does it

AutoRecon is built for finance, operations and treasury teams at banks, NBFCs and payment companies. Switch, network, core, GL, bank and MT950 statements and ERP files are loaded as they land and mapped to one schema, with bad rows rejected and the reason given. UPI and IMPS, cards, NEFT/RTGS, nostro and vendor recons are matched daily on rules you write: exact, tolerance and fuzzy narration matching. Each break is classified, aged and given an owner; recurring breaks become a suggested rule, which takes effect only after maker-checker. The daily close is approved by maker and checker, and MIS, board packs and regulatory returns are drafted with every number traced to the query that produced it. It is read-only by default and writes nothing to your core unless you open the path.

Questions

What are reconciliation matching rules?

Rules that decide when two items from different sources are the same: exact matches on reference, amount and date; tolerance matches within a set difference; and fuzzy matches on similar narrations.

What is fuzzy matching in reconciliation?

Matching items whose references or narrations are similar but not identical, after cleaning; low-confidence matches should go to review.

In what order should matching rules run?

Exact first, then tolerance from narrowest to widest, then fuzzy, with each pass removing matched items from the pool.

See it on your own data. AutoRecon — Close the day before it starts. Book a 30-minute working session with an engineer.

General guidance, current as of the date above. Figures and examples are illustrative unless a source is linked.