Transaction matching is the automated process of pairing two or more records — typically a bank statement entry (debit or credit) and a corresponding record in a business system (invoice, order, mandate, or GL entry) — based on shared attributes such as amount, date, and reference number. In South African payment reconciliation, transaction matching is the core engine that drives automated reconciliation: it eliminates the need for humans to manually compare bank statement lines to accounting records. High match rates (typically 95%+ for well-structured payment data) mean that only exceptions need human attention, drastically reducing reconciliation time for businesses processing hundreds or thousands of transactions per day.

Automated transaction matching works by applying a hierarchy of matching rules to two datasets (typically a bank statement and a payment or invoice dataset): (1) Exact match — same amount, same reference, same date: auto-confirmed match. (2) Amount match with partial reference — same amount, reference is a substring or partial match: flagged for review or auto-confirmed with a lower confidence threshold. (3) Tolerance match — amount within a defined tolerance (e.g. ±R1 for rounding), same reference: auto-confirmed with tolerance noted. (4) Date tolerance — amount and reference match but transaction date differs by 1–3 days (timing difference): confirmed as timing difference. (5) No match — no rule produces a match: exception flagged for manual investigation. Some modern reconciliation platforms use machine learning to improve matching accuracy over time.

To maximise automated matching rates, South African businesses should: (1) Use a unique, immutable reference per invoice or order in all payment instructions (EFT, DebiCheck mandate, PayShap request, RTP request). (2) Ensure the reference appears on the bank statement exactly as instructed — confirm reference format with your bank. (3) Avoid generic references like "payment" or "deposit" — these create unmatched exceptions that require manual resolution. (4) For EFT collections, use the debtor account number or invoice number as the collection reference. (5) For PayShap and RTP, embed the order ID in the payment request reference field. (6) For bulk salary or creditor payments, use the employee or supplier code as the payment reference. Consistent reference discipline is the single most impactful action to improve transaction matching rates.

Transaction matching is a component of reconciliation — it is the step that pairs individual records from two datasets. Reconciliation is the broader process that includes: data import and normalisation, transaction matching (the automated pairing step), exception investigation and resolution, posting to the accounting system, and producing a reconciliation statement that proves two sets of records agree. A business can have a 99% transaction match rate but still need to "reconcile" to confirm that the unmatched 1% are all legitimate timing differences or correctly identified exceptions. Transaction matching software is typically embedded within a broader automated reconciliation platform.