Mon – Sat: 9:00 AM – 7:00 PM IST
Purchasing

What is three-way matching?

The control that stops you paying for goods you did not order, did not receive, or already paid for.

In short

Three-way matching compares the purchase order, the goods receipt note and the supplier invoice before an invoice is approved for payment. The order says what was agreed, the receipt says what arrived, and the invoice says what is being charged. Where all three agree within tolerance the invoice posts automatically; where they do not, it is held for someone to look at.

Why It Matters

Why this matters in practice

Accounts payable fraud and simple duplicate payment both exploit the same gap: an invoice approved because it looks plausible rather than because it was checked against what was ordered and received. Three-way matching closes that gap mechanically, and as a side effect it makes the accrual at period end accurate, because received-not-invoiced is a reportable position rather than an estimate.

Prevents payment for goods never ordered or never received
Catches price and quantity discrepancies before money leaves
Makes received-not-invoiced an accurate accrual rather than a guess
Automates the routine cases so people review only exceptions
How It Works

How it works in practice

Purchase order raised

With agreed price, quantity and delivery terms, approved according to the authority matrix.

Goods receipt recorded

Against the order, at the point of physical receipt, by the person receiving — not retrospectively by finance.

Invoice received and matched

The system compares all three documents on price and quantity.

Tolerances applied

Small differences within a configured tolerance pass automatically. The tolerance should be a deliberate decision, not a default.

Exceptions routed

Mismatches go to a named person with the three documents side by side, rather than into a general queue.

Worked example

A worked example: a 2% tolerance that cost real money

A business sets a price tolerance of two percent on all purchase invoices, on the reasonable grounds that chasing small differences wastes more than it saves. The rule is applied uniformly across every line.

On a 4,000 rupee stationery invoice, two percent is 80 rupees and the rule is sensible — investigating costs more than the exposure. On a 40,00,000 rupee machinery invoice, two percent is 80,000 rupees passing without review. Over a year the business auto-approves roughly 11 lakh of unreviewed price variance, almost all of it on a handful of large invoices, while the exception queue fills with stationery.

Banded tolerances fix it: a percentage with an absolute cap, so small differences clear automatically and large ones stop regardless of how small a proportion they represent. The exception queue gets shorter and more useful at the same time, because the items in it are now the ones worth a person's attention. A single global tolerance is almost always simultaneously too tight for low-value lines and too loose for high-value ones.

Key terms

The vocabulary, briefly

Term What it means
Purchase order What you agreed to buy, at what price. The commercial reference point for the match.
Goods receipt What actually arrived, recorded at the dock rather than reconstructed later.
Supplier invoice What you are being asked to pay. The document being tested against the other two.
Tolerance The permitted variance before a mismatch blocks payment. Set per value band, not as one global figure.
Exception queue Invoices that failed the match and need a human. Its size is the health check on the whole process.
What Goes Wrong

Common mistakes worth avoiding

Receipting in bulk after the fact

Receipts entered by finance to clear invoices defeat the control entirely — the receipt is then just a copy of the invoice.

Tolerances set too wide

A generous tolerance passes exactly the discrepancies the control exists to catch.

No owner for exceptions

Unmatched invoices accumulate until someone approves them in bulk to clear the backlog.

Services treated as goods

Service invoices have no physical receipt. Use a service entry sheet or two-way match rather than forcing the model.

What to measure

Numbers worth tracking

Metric What it tells you What distorts it
First-pass match rate The share of invoices matching without intervention. The core efficiency measure. Counting auto-approvals within tolerance as matches, which flatters the figure.
Exception queue ageing How long mismatches sit unresolved. Ageing exceptions become duplicate payments and missed discounts. Clearing old exceptions in bulk at period end without recording causes.
Mismatch reasons Price, quantity, missing receipt, missing order. Each points at a different broken process. A dominant 'other' category, which means causes are not being captured.
Value passed within tolerance The money flowing through unreviewed. The number that justifies banded tolerances. Reporting the count of invoices rather than their value.
FAQ

Questions we are asked about this

A comparison of the purchase order and the invoice only, with no goods receipt. It is appropriate for services and for low-value purchases where a receipt step adds no control.
Low enough that genuine discrepancies are caught, high enough that rounding and freight variance do not block every invoice. Many businesses set a small percentage with an absolute cap.
Where there is no goods receipt to match against, which mainly means services. Professional fees, subscriptions, rent and utilities are matched to the order and an approval rather than to a receipt. The important thing is that two-way matching is a deliberate configuration for a defined category, not a fallback used whenever a receipt is missing — the second case is how missing receipts stop being a problem anyone notices.
A percentage with an absolute cap, banded by value, and reviewed after a quarter of real data. Small invoices clear on percentage, large ones stop on the cap, and you avoid the trap in the example above. Set it by category too where it matters: a commodity with a volatile market price justifies a different band from a contracted service with a fixed rate.
Match against cumulative receipts rather than against a single one, so an invoice covering two part-deliveries matches correctly. Systems that match invoice to receipt one-to-one will generate exceptions on every split delivery, and the usual workaround — staff manually forcing them through — defeats the control entirely. If your exception queue is full of partial deliveries, the configuration is wrong rather than the suppliers.
It makes one common type much harder: invoices for goods never received. It does not address collusion between a buyer and a supplier, since someone who can raise the order and confirm the receipt can match an invoice for anything. That is a segregation-of-duties control, not a matching one. The two are complementary, and treating matching alone as fraud prevention is a well-documented way to be surprised.
Related

Where to read next

Our ERP software solutions

Read more

What is ERP software?

Read more

ERP with warehouse management

Read more

Need this applied to your own systems?

We build and support the systems this page describes. Start with a free working proof of concept — no cost, no commitment.