.svg)
Every finance team processes invoices. But not every finance team has a reliable way to verify that what they are paying for is what they actually ordered and received. That gap is where duplicate payments, overpayments, and invoice fraud tend to live.
3-way matching is the control that closes that gap. It is one of the most effective processes in accounts payable automation, and one of the most time-consuming when done manually. This guide covers how it works, where it breaks down, and what changes when you automate it properly.
3-way matching is a verification process that compares three documents before a supplier invoice is approved for payment: the purchase order, the delivery note, and the invoice itself.
The logic is straightforward. You ordered something (purchase order), you received it (delivery note), and now someone is asking you to pay for it (invoice). Matching all three confirms that the payment request is legitimate, accurate, and tied to a real transaction.
When the three documents agree, the invoice moves forward. When they do not, it stops until someone investigates.
The purchase order is the document your team raises when committing to buy goods or services from a supplier. It sets the agreed price, quantity, and terms. It is your reference point for everything that follows.
The delivery note is the confirmation that the goods or services were actually delivered. Without it, there is no evidence that what was ordered actually arrived.
The invoice is the payment request. It comes from the supplier and should reflect what was ordered and delivered. The problem is that it often does not, whether due to errors, different pricing, partial deliveries, or, in some cases, deliberate manipulation.
Mismatches between these three documents are more common than most finance teams realise. And the consequences go beyond a delayed payment.
An undetected mismatch can mean paying for goods that never arrived. It can mean paying a price that was never agreed. It can mean a duplicate payment going through because the same invoice was entered twice under slightly different reference numbers. And in more serious cases, it can mean a fraudulent invoice from a supplier that does not exist.
The financial impact compounds over time. A team processing 500 invoices a month that misses even 1% of errors is absorbing those losses quietly, often without realising the scale.
The process starts before the invoice arrives. When your team commits to purchasing from a supplier, a purchase order is raised in your ERP or procurement system. It records the supplier, the items or services, the agreed unit price, and the quantity.
This PO is the contract. Everything downstream should match it.
When the goods or services arrive, someone in the business confirms receipt. This creates the delivery note. In manufacturing or retail environments, this step is usually tied to a physical delivery. In service businesses, it might be a project sign-off or a delivery confirmation.
The delivery note is critical. Without it, you have no documented evidence that the transaction actually took place. Many AP problems trace back to invoices being approved before the delivery note is created or recorded.
When the invoice arrives, the matching process begins. The system, or a member of the AP team, compares the invoice against the purchase order and the delivery note across three dimensions:
If all three match within the defined tolerance thresholds, the invoice is approved and moves to payment. If any field is outside tolerance, it becomes an exception.
This is where most of the manual work lives in teams that have not automated the process.
An exception means someone needs to investigate. Is the price difference a supplier error or a contract update that was not reflected in the system? Was part of the delivery short-shipped? Was this invoice already processed under a different reference?
In a manual environment, resolving exceptions involves chasing the supplier, cross-referencing spreadsheets, and sending emails. In an automated environment, the exception is routed to the right person with full context, the relevant documents attached, and a clear resolution path.
This is the most frequent exception. The supplier invoices at a price that differs from the purchase order, sometimes because of a legitimate price update, sometimes because of a billing error, and occasionally because of something more serious.
The right response depends on the cause. A well-designed matching system will flag the discrepancy, show the delta, and route the exception to the person who can confirm whether the new price was agreed. Without that visibility, the AP team is either approving payments blind or chasing answers manually.
The invoice claims 100 units were delivered. The delivery note records 85. This is a short-shipment situation, and it is common in high-volume supply chains.
The correct action here is a partial payment or a request for a credit note from the supplier. Teams without proper matching controls often pay the full invoice amount and attempt to reconcile later, which creates a significant amount of downstream work and, frequently, overpayment that is never recovered.
The invoice arrives before the delivery note is created. This happens often in businesses where the team receiving goods and the team processing invoices are not well connected. The AP team has an invoice but no confirmation of delivery.
Approving the invoice at this point is a risk. Holding it and flagging the missing delivery note is the correct approach, but it requires a system that surfaces this clearly rather than just letting the invoice sit in a queue.
A supplier delivers in two shipments. They invoice in two parts. The purchase order was for the full order. Matching partial invoices against a single purchase order requires a system that can track what has already been matched and what is still outstanding. Manual matching struggles with this. Teams often end up with duplicate payments or missed liabilities because the tracking is not robust enough.
At low volumes, manual 3-way matching is manageable. A team processing 50 invoices a month can realistically check each one against the purchase order and delivery note without too much friction.
At 500 invoices a month, it is a different problem. The time required scales linearly. The error rate increases because people are cross-referencing more documents under more pressure. And the cost per invoice, in staff time alone, becomes significant.
According to Gartner, the cost of processing a single invoice manually averages between £4 and £25 in the UK, with complex or error-prone processes reaching as much as £50. Automated processing consistently brings that figure below £4. Taking a conservative mid-range estimate of £15 per invoice, a team processing 500 invoices a month spends around £90,000 a year on processing costs alone. Automation at £3 per invoice brings that to £18,000 — a saving of £72,000 annually, before accounting for payment errors or fraud.
Manual matching tends to hold up until it suddenly does not. The two most common breaking points are month-end and staff changes.
At month-end, invoice volume spikes and the pressure to process quickly increases. That is exactly when exceptions get waved through rather than properly investigated.
When an experienced AP team member leaves, the institutional knowledge of how to handle specific suppliers or resolve particular exception types goes with them. Without a system that captures that logic, the team is starting from scratch.
For most finance teams processing more than 200 invoices a month, the ROI on automating 3-way matching is clear within the first year. The saving comes from three places: reduced processing time, reduced payment errors, and reduced fraud exposure.
You can calculate your own numbers with Dost's savings calculator.
Traditional matching tools work on fixed rules. If the price on the invoice is within X% of the purchase order, it passes. If not, it fails. That works for straightforward cases, but real supply chains are not straightforward.
Prices change. Contracts get renegotiated. Suppliers use different reference formats. A rule-based system generates exceptions for all of these situations, many of which are not genuine problems. The AP team ends up spending significant time resolving false positives.
AI-native platforms learn from context. They recognise that a particular supplier consistently invoices under a different reference format, or that a certain category of purchase always has a tolerance of 5% due to weight-based pricing. Over time, the system gets better at distinguishing genuine exceptions from noise.
This is the practical difference between a platform that was built with AI at its core and one that has had matching rules added on top of legacy architecture. The former improves. The latter requires constant manual reconfiguration.
When a genuine exception is identified, the way it is handled matters as much as how it is caught. An approval workflow that routes exceptions to the right person, with the right context, resolves them faster and with less back-and-forth.
Dost's platform attaches the original purchase order, the delivery note, and the invoice to every exception, along with a clear summary of what does not match and by how much. The approver has everything they need to make a decision without leaving the system.
And because every action is logged, the audit trail is complete. Who reviewed the exception, when, and what they decided.
Dost handles the full matching workflow: intelligent data extraction from incoming invoices, automatic comparison against purchase orders and delivery notes, tolerance-based routing, and exception management with a complete audit trail. All of it integrated with your existing ERP in real time.
If you want to see it working against your own invoice formats, book a demo with our team.
It is not mandated by law in most jurisdictions, but it is strongly aligned with financial controls that auditors expect to see in any well-governed finance function. For businesses subject to external audits, having documented evidence that invoices were verified against purchase orders and delivery records is important. It is also a requirement in many procurement frameworks and supplier contracts.
2-way matching compares the invoice against the purchase order only. It confirms that the price and quantity match what was ordered, but it does not verify that the goods were actually received. 3-way matching adds the delivery note, which closes that gap. For businesses where goods or services are received before payment, 3-way matching is the more robust control.
It significantly reduces fraud risk, particularly for the most common schemes: duplicate invoices, inflated quantities, and fictitious supplier invoices. A fraudulent invoice that does not correspond to a real purchase order will fail the matching process immediately. That said, matching is one layer of control, not the only one. A comprehensive fraud-prevention approach also includes supplier verification, segregation of duties, and anomaly detection across your full invoice history.
3-way matching is one of the most reliable controls a finance team can have. It catches errors before they become payments, and it creates the audit trail that auditors and regulators expect.
The challenge is scale. Done manually, matching is accurate enough at low volumes but breaks down as invoice volume grows. The time cost is significant, and the error rate under pressure is higher than most teams realise.
Automated 3-way matching, built into a proper AP automation platform, solves this. Exceptions are caught, routed, and resolved systematically. The team spends less time on mechanical checks and more time on decisions that require human judgement.
Want to see what the numbers look like for your business? Use our savings calculator to estimate the impact of automating your matching process.
