Three-way match: what to do about the exceptions, not the matches

A three-way match compares three documents before an invoice is paid: the purchase order, showing what was ordered and at what price; the receiving record, showing what actually arrived; and the supplier invoice, showing what is being charged. Where all three agree within tolerance, the invoice can be released for payment. Where they do not, it becomes an exception.

The control exists because the alternative is paying on the invoice alone, which means paying for goods nobody confirmed arriving at prices nobody agreed. It is the standard accounts payable control and it is genuinely effective — but its cost and its failures both live entirely in the exception queue rather than in the matching itself.

Where three-way match exceptions come from

• Quantity differences — partial deliveries, over-shipments, or goods received but not recorded because the warehouse was busy.
• Price differences — a supplier price increase not reflected on the order, freight added on the invoice, currency or rounding.
• Missing purchase order — bought without one, which is a process problem rather than a matching problem.
• Timing — invoice arrives before the goods are receipted, which is not a discrepancy but looks like one.
• Unit of measure mismatches, where the order is in cases and the invoice in units.

Tolerances decide how much work you create

Every three-way match needs tolerances: a percentage and an absolute amount within which a difference is accepted without review. Set them too tight and a team spends its week clearing pennies; too loose and the control stops controlling. The workable approach is a small tolerance by value with a hard cap, reviewed against actual exception data after a quarter — most organisations discover that a large share of their exceptions sit under a threshold that would have been safe to accept.

Two-way match for services. Services have no receiving record, so matching the order to the invoice plus a documented approval by the person who received the service is the correct control. Trying to force a three-way match onto services produces a receipt entry that means nothing.

Who releases an exception is the real control

The matching is arithmetic; the authority to override it is governance. An exception queue where anybody in accounts payable can release a mismatch removes most of the control’s value, and one where only the budget holder can release it slows payment to a crawl. The workable split is tiered: small differences released by accounts payable within a stated tolerance, price and quantity differences routed to purchasing, anything above a threshold to the budget holder — with every release recorded against the invoice and the reason stated.

Keeping the three documents together

Most exception time is spent finding documents rather than deciding anything. Ettex Invoices holds the invoice with the order and receipt attached so the comparison is on one screen, Ettex Sheets tracks the exception queue with reasons and ageing so recurring causes are visible, and the supplier records behind it are the ones covered in accounts payable.

Plainly: this is documents and records, not an AP automation platform. Above a certain invoice volume, automated matching with supplier portals pays for itself; what this covers is designing the control and the exception path, which no platform decides for you.