Accounts payable documents and invoices on a desk being reconciled
Reference Guide

Three-Way Matching: The Complete Reference

By Fredrik Filipsson
Published February 21, 2026
Updated February 21, 2026
Reading time 11 min

Definition

Three-way matching is an accounts-payable control that compares three documents before an invoice is paid — the purchase order (what was ordered), the goods receipt (what was received), and the invoice (what the vendor is billing) — and clears the invoice for payment only when quantities, prices and descriptions align within tolerance across all three.

It is the single most important financial control in procure-to-pay, because it is the mechanism that ensures every dollar leaving the business corresponds to something the business actually ordered and received at the agreed price. This reference defines the concept precisely, explains the documents and tolerances, distinguishes it from two- and four-way matching, catalogues the exceptions that arise in practice, and explains where AI changes the picture — without rehashing the deep benchmark analysis in our companion article on how smart AI 3-way matching really is.

The Three Documents

Matching works because each document is created by a different party at a different stage, so agreement between them is meaningful evidence that a transaction is legitimate.

  • Purchase order (PO): created by the buyer when an order is placed. It records the agreed items, quantities, unit prices and terms. It is the statement of intent — what the organisation committed to buy.
  • Goods receipt (GR): created by the receiving function when items arrive. It records what was actually delivered and accepted. It is the evidence of fulfilment.
  • Invoice: created by the supplier, requesting payment. It states what the supplier believes it is owed.

When the PO, GR and invoice agree, the chain is complete: the organisation ordered it, received it, and is being billed for the same thing at the same price. The discipline of requiring all three is what separates a controlled payables process from one that simply pays whatever invoices arrive.

How the Match Works, Step by Step

At the line level, a match compares the same attributes across the documents:

  1. Locate the PO referenced by the invoice (via PO number or, with AI, by inference when the number is missing).
  2. Compare invoice quantity against PO quantity and against the goods-receipt quantity.
  3. Compare invoice unit price and line total against the PO price.
  4. Confirm the item description and unit of measure correspond.
  5. Apply tolerances; if every attribute is within tolerance, clear the line for payment. If not, raise an exception.

A clean match needs no human involvement — this is the "touchless" or "straight-through" case that modern AP teams try to maximise. The volume of clean matches a team achieves is a core efficiency metric, tracked in our companion AP automation straight-through rate benchmark.

Two-Way, Three-Way and Four-Way Matching

"Three-way" is one point on a spectrum of control. The right level depends on the risk and nature of the spend.

VariantDocuments comparedTypical use
Two-wayPO + invoiceServices or non-receiptable spend where a goods receipt is not meaningful
Three-wayPO + goods receipt + invoiceThe standard control for physical goods
Four-wayPO + receipt + inspection + invoiceRegulated or quality-critical goods requiring acceptance testing

Two-way matching is common for services, subscriptions and other spend where there is nothing to physically receive. Four-way adds a quality-inspection or acceptance step and is used in environments — pharmaceuticals, aerospace, food safety — where receiving goods is not enough; they must pass a check before payment is justified. Each additional document strengthens control but increases the number of potential exceptions, which is the trade-off every AP policy has to balance.

Tolerances: The Control Dial

Rarely do three documents agree to the penny and the unit. Suppliers round, ship slightly different quantities, or apply minor price changes. Tolerances define how much difference is acceptable before a line becomes an exception — for example, accepting an invoice price within 2% of the PO price, or a quantity within a defined percentage or unit count.

Tolerances are the control dial of the whole process. Tight tolerances catch more discrepancies and prevent more leakage, but generate more exceptions and more review work. Loose tolerances reduce manual effort but let more variance — and more potential overpayment — through unexamined. Setting tolerances is therefore a deliberate risk decision, not a technical default, and good AP functions revisit them by category and supplier rather than applying one global figure.

"A tolerance is not a setting you configure once and forget. It is a continuous statement of how much variance the business is willing to pay without asking a question."

Common Exceptions

When documents fall outside tolerance, a line becomes an exception requiring resolution. The recurring types are worth knowing because they recur in every AP function:

  • Price variance: the invoice price differs from the PO price — legitimately (a renegotiated rate) or not (an error or overcharge).
  • Quantity variance: invoiced quantity differs from ordered or received quantity, often from short or partial shipments.
  • Missing goods receipt: the invoice arrives before the receipt is recorded — a timing problem rather than a real discrepancy.
  • Partial shipment: a supplier ships and invoices in tranches, so any single invoice matches only part of the PO.
  • No PO (maverick spend): an invoice arrives with no purchase order to match against, indicating an off-process purchase.

Each exception type has a different root cause and a different correct resolution, which is precisely why exception handling — not the matching itself — is where AP staff time goes, and where automation earns its keep.

Three-Way Matching as Fraud Control

Beyond catching honest errors, three-way matching is a primary defence against payment fraud. Common schemes — duplicate invoices, fictitious vendors, inflated quantities, billing for goods never delivered — all rely on an invoice being paid without verification against an order and a receipt. By requiring that an independent purchase order and an independent goods receipt corroborate every invoice, the control denies these schemes the unverified payment they depend on. This is why auditors treat the integrity of the matching process as a core internal control and why weakening tolerances or bypassing matching for "trusted" vendors is a recognised risk.

Where AI Fits

Traditional matching is rules-based and brittle: it needs a clean PO number, exact references, and pre-configured tolerances, and it flags anything it cannot reconcile. AI changes three things. It matches without perfect references, inferring the right PO when the number is missing or malformed. It learns which variances are routine — recognising that a given supplier always ships in three tranches, for instance — so it stops flagging benign patterns as exceptions. And it prioritises genuine anomalies, routing the cases that actually warrant a human to a person while clearing the rest.

The net effect is a shift in where work happens: from manual line-by-line matching to focused exception handling, which raises touchless processing rates and frees AP staff for higher-value work. The realistic limits of that automation — accuracy ranges, what breaks, and why vendor claims overstate it — are covered in our companion analysis of AI 3-way matching, and the tools that do this best are profiled in our invoice & AP automation category, including platforms such as Stampli and Tipalti.

Compare AP Automation Tools

See how the leading invoice and AP platforms handle matching, exceptions, and touchless processing.

Practical Best Practices

Three habits separate a matching process that works from one that frustrates everyone. First, fix data before tooling: the biggest driver of match rates is clean, complete PO and receipt data, not the sophistication of the engine — a point reinforced across the AP analysis in our State of Procurement AI report. Second, set tolerances by category and supplier rather than globally, so high-risk spend gets tight control and low-risk spend does not drown reviewers in trivial exceptions. Third, treat exception handling as the real workflow: design clear ownership and resolution paths for each exception type, because that is where time and risk concentrate. For procurement leaders building the broader business case for AP automation, our CPO strategic guide places matching within the wider source-to-pay control framework.

Frequently Asked Questions

What is three-way matching?

An AP control that compares the purchase order, goods receipt, and invoice before payment. When quantities, prices, and descriptions align within tolerance, the invoice is cleared; if not, it becomes an exception.

What is the difference between two-way, three-way, and four-way matching?

Two-way compares PO and invoice; three-way adds the goods receipt; four-way adds an inspection or quality-acceptance document. More documents mean stronger control and more potential exceptions.

What is a matching tolerance?

The allowable difference between documents before a line is flagged — for example an invoice price within 2% of the PO price. Tight tolerances catch more discrepancies; loose tolerances reduce review work but allow more variance.

Why is three-way matching important?

It ensures money is only paid for goods actually ordered and received at the agreed price, prevents duplicate and erroneous payments, and is a primary defence against invoice fraud.

How does AI help with three-way matching?

AI matches at scale even without perfect references, learns which variances are routine, and routes only genuine exceptions to humans — shifting AP work from manual matching to exception handling and raising touchless rates.