Both are accounts payable controls that verify an invoice before you pay it. The difference is how much evidence each one demands — and that difference decides which is right for a given type of spend.
Both are validation steps inside the broader invoice processing workflow. They differ only in how many documents must agree before an invoice is cleared for payment.
Two-way matching compares two documents: the supplier's invoice and the purchase order. The check confirms that the price and quantity being billed are the price and quantity that were ordered. If they agree within tolerance, the invoice passes; if not, it becomes an exception. Because it requires no goods-receipt record, two-way matching is fast and works even when there is nothing physical to receive.
Three-way matching adds a third document — the goods receipt — to the invoice and purchase order. Now three things must agree: what was ordered, what was actually received, and what is being billed. This closes the gap that two-way leaves open, because it refuses to pay for items that were ordered and invoiced but never delivered. The trade-off is a dependency on timely, accurate receiving data; if receipts are not recorded promptly, legitimate invoices stall as exceptions.
The classic failure that three-way matching prevents is paying for goods that never arrived — whether through fraud, a billing error, or a shipment that was short or lost. Without the receipt in the comparison, an invoice that matches the purchase order on price and quantity looks perfectly payable even if nothing was actually delivered. Adding the receipt forces the system to confirm delivery before money moves, which is why three-way matching is the standard control for inventory, direct materials, and capital purchases. The catch is that the control is only as good as the receiving discipline behind it: if your warehouse records receipts days late or inconsistently, three-way matching will generate a steady stream of false exceptions that have nothing to do with bad invoices and everything to do with slow data entry upstream.
For a deeper conceptual walkthrough of the receipt-based control, see our companion explainer on two-way versus three-way matching, which focuses on the document mechanics in more detail.
A helpful way to frame the choice is to ask what question each method is built to answer. Two-way matching answers: "Are we being billed for what we agreed to buy?" It guards the commercial terms — price and quantity — against a supplier who invoices for more than the purchase order authorized. Three-way matching answers a second, harder question on top of that: "Did we actually receive what we are being billed for?" That extra question is the entire reason the goods receipt joins the comparison, and it is the reason three-way matching is the stronger control for anything physical. Where there is nothing to receive, the second question is meaningless, and insisting on a receipt only invents an exception with no offsetting control benefit.
Where the two methods diverge on control, speed, and fit.
| Dimension | 3-Way Match | 2-Way Match |
|---|---|---|
| Documents | PO + receipt + invoice | PO + invoice |
| Verifies receipt | ✓ Yes | ✗ No |
| Control strength | Higher | Lower |
| Speed | Slower (needs receipt) | Faster |
| Fraud / error protection | Strong | Moderate |
| Best-fit spend | Physical goods | Services, subscriptions, utilities |
| Key dependency | Accurate goods receipts | Accurate PO data |
| Exception driver | Missing or late receipts | Price / quantity variance |
See how leading AP tools automate both matching methods and lift straight-through rates.
STP Rate BenchmarkThe decision turns on one question: is there a physical receipt to verify, and is the spend material enough to warrant the control?
You are buying physical goods of material value, receiving is recorded in your ERP, and you need assurance that what was billed was actually delivered. This is the default for direct materials, inventory, and capital purchases.
There is no goods receipt to capture — services, software subscriptions, utilities, professional fees — or the spend is low-value and low-risk, so requiring a receipt would cost more control effort than it saves.
You operate in a regulated or quality-critical environment where an inspection or acceptance step must be confirmed before payment. The fourth document is the inspection record.
In practice, the strongest AP functions do not standardize on a single method. They apply matching rules by category, supplier, and value threshold — three-way for inventory, two-way for the SaaS bill, manual review above a ceiling. This is the same risk-proportionate logic that shapes good invoice and AP automation design, and it depends on the policy being encoded in the system rather than living in an AP clerk's memory.
It also helps to think about the decision in terms of cost versus control rather than right versus wrong. Every additional document you require to match is an additional control, but also an additional dependency that can stall an invoice and an additional piece of data that must exist and be accurate. Three-way matching buys you proof of receipt at the cost of depending on timely receiving; two-way matching buys you speed at the cost of that proof. For a high-value inventory purchase, the control is clearly worth the dependency. For a low-value services invoice, the dependency would cost more in friction than the control could ever save. The skill is not in preferring one method but in drawing those lines deliberately — setting value thresholds and category rules that put each invoice through exactly as much scrutiny as its risk justifies, and no more.
Two practical reminders round this out. First, the choice of matching method is a control decision and should involve finance and audit, not just AP, because it determines how the organization protects its payments. Second, whatever rules you set must be documented and enforced consistently; a policy that exists only in the heads of experienced clerks evaporates the moment they leave. Encoding the rules in the platform makes them durable, auditable, and uniform — which is itself a meaningful control improvement over even a well-run manual process.
The method matters less once software is doing the matching — what matters is the rules and the data.
Modern accounts payable platforms support two-way and three-way matching natively and apply the correct method automatically based on configurable rules. They extract invoice data, pull the matching purchase order and receipt, compare within set tolerances, and route only the exceptions that genuinely need a human. The result is a higher straight-through processing rate and a consistent control that does not depend on which clerk happens to be reviewing.
Automation also delivers a control benefit that is easy to overlook: uniformity. In a manual process, the rigor of matching varies with the experience, attention, and workload of whoever is reviewing — a tired clerk at month-end may wave through what a fresh one would catch. A rules engine applies the same logic to every invoice, every time, and records exactly what it checked. That auditability is itself valuable to finance and audit teams, because it turns matching from a judgment that lives in people's heads into a documented, repeatable control. The combination of higher throughput and stronger, more consistent control is why matching automation has become a default expectation in any serious accounts payable function rather than a nice-to-have.
Our analysis of buyer-reported deployments and the patterns documented in our invoice and AP automation market analysis point to the same lesson: matching automation succeeds or fails on data quality. Clean PO data lifts two-way match rates; promptly recorded receipts lift three-way match rates. Platforms such as Tipalti, Stampli, and Vic.ai differ in how they capture documents and surface exceptions, but none can overcome a messy item master or missing receipts.
This reframes the whole debate. Once software is applying the rules, the question is no longer "which matching method should we adopt?" but "have we configured the right method for each kind of spend, and is our data good enough for it to work?" A capable platform will run three-way matching on inventory, two-way on the SaaS subscription, and a value-threshold escalation to manual review on anything unusually large — all without an AP clerk deciding case by case. The constraint that remains is the data feeding it: a receipt that is never recorded will defeat the best three-way engine, and a purchase order missing line items will defeat the best two-way one. That is why teams evaluating tools should weigh integration with their receiving and ERP systems at least as heavily as the matching features themselves, a theme we return to throughout the invoice and AP automation category.
Matching is never perfectly exact — the question is how much variance you allow before an invoice stops.
Neither method demands that documents agree to the penny. Both run within configured tolerances — small, deliberate margins that let routine, immaterial differences pass without human review. A two percent price variance on a commodity, a few cents of rounding, a minor quantity difference within an agreed range: tolerances absorb these so your team is not buried under trivial exceptions. The art is in the setting. Tolerances that are too tight turn every invoice into an exception and erase the efficiency you bought; tolerances that are too loose weaken the control and let real overcharges slip through. The right level varies by category, supplier, and value, which is why it should be configured rather than assumed.
When an invoice falls outside tolerance, it becomes an exception that someone resolves. In two-way matching, exceptions cluster around price and quantity variances against the purchase order. In three-way matching, the most common exception is the missing or late goods receipt — a timing problem rather than a real discrepancy. Recognizing that distinction matters: a missing-receipt exception is usually solved by tightening receiving discipline upstream, not by reviewing the invoice. The proportion of invoices that become exceptions, and how quickly each clears, is the metric that ultimately decides whether your matching process is fast or slow, a point we develop further in our guide to invoice processing.
Tolerances are best set deliberately and by segment rather than as a single global figure. A commodity bought from a stable supplier may safely carry a slightly wider price tolerance, because small, frequent fluctuations are normal and reviewing each one wastes time. A high-value or volatile category warrants tighter tolerances, because a small percentage variance represents real money and real risk. The same logic applies to quantity: a supplier known to ship in partial batches needs different handling from one expected to deliver complete orders. Getting tolerances right is therefore not a one-time configuration but an ongoing tuning exercise, revisited as suppliers, prices, and categories change — and it is one of the highest-leverage levers an AP team has, because it directly governs how many invoices need a human at all.
The pitfalls that quietly degrade either matching method.
The most expensive mistake is standardizing on a single method for everything. Forcing three-way matching onto a software subscription manufactures an exception on every invoice, because there is no receipt to match; applying two-way matching to high-value inventory removes the proof-of-delivery control where you most need it. The fix is rule-based matching that varies by spend type and value, encoded in the platform rather than left to a clerk's discretion.
A second common error is neglecting the data the method depends on. Three-way matching collapses when goods receipts are recorded late or not at all, and two-way matching collapses when purchase-order data is incomplete. Teams sometimes blame the matching software for low match rates that are really a symptom of upstream data quality. A third mistake is setting tolerances once and never revisiting them as suppliers, prices, and categories change. Finally, many organizations under-invest in supplier compliance — failing to insist that every invoice carries a valid PO reference — which guarantees a steady stream of avoidable exceptions regardless of which method is used.
Avoiding these turns matching from a source of friction into a quiet, reliable control. The platforms in our AP automation market analysis all support rule-based matching across methods; the differentiator is how cleanly they surface exceptions and how well they integrate with your receiving and ERP data.
This is not a contest with a single winner — it is a fit decision. Three-way matching is the stronger control and the correct default for physical goods, because it is the only method that proves you received what you are paying for. Two-way matching is not a weaker version to be avoided; it is the right tool for spend where no receipt exists, and forcing a receipt onto a software subscription only manufactures exceptions.
The mature position is to run both, governed by clear rules that match the control to the risk. Encode those rules in your AP platform, keep your purchase-order and goods-receipt data clean, and let automation handle the routine while your team works the exceptions. To go deeper on the surrounding workflow, read our guide to invoice processing and browse all procurement comparisons.
Common questions from AP managers and controllers.