3-Way Matching in Accounts Payable: Why It Still Matters
Three-way matching is the least glamorous and most effective control in accounts payable. It asks a simple question before any invoice gets paid: does what was ordered (the purchase order), what was received (the receipt), and what was billed (the invoice) actually agree? When the answer is yes, the invoice flows to payment. When the answer is no, something needs investigation before the check clears.
In every survey of AP fraud and payment error over the past decade, the organizations with disciplined three-way matching consistently report the lowest rates of duplicate payments, overbilling, and unauthorized spend. And yet, three-way matching is also the control that gets compromised first when AP volume grows faster than the team — usually through well-intentioned shortcuts that each sound reasonable in isolation.
0%
Invoice error rate for top-quartile AP teams using automated 3-way matching vs. 1.29% for manual or 2-way matching teams (APQC)
Each of the three documents in a three-way match represents a separate person or system committing to a separate fact. The control only works because the three sources are independent — a mismatch signals a real problem, not a data-entry inconsistency in a single system.
What each document attests to
- Purchase Order — what buyer authorized and committed to pay for (unit count, unit price, ship-to location, terms)
- Receipt (goods receipt / packing slip / service acceptance) — what was physically delivered or completed, confirmed by a receiving party who is not the buyer
- Invoice — what the vendor is billing for, prepared by the vendor's AR team
The power of the match comes from independence. If the same person who signed the PO is also the person who confirms receipt and the person who approves the invoice, three-way matching collapses into a single-source attestation — which is how most invoice fraud and duplicate payment errors are able to occur in small or understaffed AP shops.
The three documents rarely agree character-for-character. A disciplined matching process defines tolerance thresholds per field, so that real discrepancies are caught while immaterial rounding or vendor-formatting quirks don't clog the exception queue.
Fields compared in a proper 3-way match
- Vendor identity — PO vendor must equal invoice vendor (after alias normalization)
- PO number — invoice must reference the PO; blank-PO invoices are an exception class
- Line-item quantity — receipt quantity should meet or exceed invoiced quantity
- Line-item unit price — invoice price must not exceed PO price beyond a tolerance band
- Line-item extended amount — (qty × price) reconciles within tolerance
- Tax and freight — compared against PO terms, with shipping-term-aware handling
- Total invoice amount — against the cumulative authorized spend of the PO
- Ship-to or service-location — invoice service address must match PO ship-to
Define your tolerance thresholds deliberately. A common starting point is zero variance on quantity, ±1% or $25 on unit price, and ±2% or $50 on total. Too tight and the exception queue drowns the team; too loose and overbilling slips through.
Two-way matching compares only the purchase order and the invoice — skipping the receipt. It confirms that what the vendor is billing matches what the organization committed to, but it never confirms that anything was actually delivered. That gap is the single most common vector for both external invoice fraud and internal expense shifting.
Two-way matching is appropriate for a narrow band of spend categories: genuine services where no physical delivery exists (insurance premiums, SaaS subscriptions billed from a signed contract, professional service retainers), recurring utility charges, and pre-paid arrangements. Everything tangible — materials, equipment, field supplies, subcontractor progress billings — should live on the three-way side of the line.
0%
Of total invoice volume successfully matched on two documents but invalid on three — the share of invoices that get paid under 2-way but would have been caught under 3-way (IOFM benchmark)
Three-way matching in a textbook is clean. Three-way matching in a $150M construction GC with 4,000 invoices per month is messier. Four failure modes drive almost every false match and missed exception.
A vendor ships 500 units in two deliveries of 250. The first delivery is received and invoiced at the half total. The second delivery arrives three weeks later. During the gap, a naive matching rule either over-pays on invoice one (matching it against the full PO) or fails to match either invoice because neither equals the PO total. The fix is running three-way matching at the line-item and quantity level, with running totals that track cumulative delivered quantity against the cumulative invoice quantity.
The PO orders 100 CWT (hundredweight) of rebar. The invoice bills 10,000 LB. The receipt records 45,359 KG. All three represent the same steel, but naïve string matching fails because the units don't agree. A proper matching engine either canonicalizes units or records an explicit unit-conversion for each spend category.
Get AP insights in your inbox
Get our weekly roundup of AP automation tips and industry news. No spam, ever.
No spam. Unsubscribe anytime.
3. Tax and Freight Rounding
The PO authorizes tax at the quoted rate. The vendor's invoice calculates tax at their exact rate, which rounds to the penny differently on a 47-line invoice than on a single-line PO. Without a tax tolerance band, every invoice with sales tax lines gets flagged as an exception. The tolerance band is usually a few dollars or a fraction of a percent — small enough to catch real tax errors, big enough to not drown the team.
The receipt for a shipment of materials is obvious: the packing slip gets signed at the gate. The receipt for two weeks of site supervision is less obvious. Mature AP organizations define per-category 'service acceptance' workflows — daily log signoff for construction labor, deliverable acceptance for professional services, uptime confirmation for managed services — so that every PO type has a matching receipt convention.
Most AP teams do not fail three-way matching because they don't believe in it. They fail because the manual version does not scale past a few hundred invoices per month, and the workarounds that keep things moving — paying on receipt-only, paying against an approved PO without checking receipt status, paying a recurring invoice without a PO — become the norm.
Automated matching rebuilds the discipline at scale. A well-implemented system extracts line items from both the PO and invoice, ingests receipts in near real time from the receiving or field team, and runs the match within seconds of the invoice arriving. The job of the AP analyst shifts from matching to exception handling — reviewing only the 5-15% of invoices where tolerances fail, and using the full audit trail (PO, receipt, invoice, match log, tolerance result) to resolve each one.
Practical benchmarks for an automated matching program
- 70-85% of invoice volume auto-matches on first pass within 24 hours of invoice receipt
- Top 5 exception reasons account for 80% of all holds (rebuild the rules to address these first)
- Average time from invoice arrival to payment-approved drops from days to hours
- Full audit log: every comparison, tolerance result, override, and approver — stored against the invoice record
Automating three-way matching works best when the receipt side is also digitized. In construction, the receipt is often a delivery ticket, a daily log, or a sub's pay application — ingesting those alongside invoices is what makes the match meaningful for services and progress billings, not just materials.
Every organization has spend categories where three-way matching is genuinely the wrong tool. Pre-approved utility bills on a known account don't need a PO. A $12 office-supply invoice doesn't justify the overhead of a receipt workflow. The right pattern is a tiered policy: explicit three-way match above a dollar threshold (e.g. $2,500), two-way match with approved vendor and budget below that, and category-based exclusions for utilities and recurring services.
What is never acceptable is skipping three-way matching on large invoices because the team is behind. The error rate on unchecked large invoices is orders of magnitude higher than on small ones — both because the stakes are larger and because attackers deliberately target the size range where they expect matching to be bypassed.
Three-way matching is not a legacy control. It is the foundational AP fraud defense, and the organizations that keep it operating at scale are also the ones that pay fewer duplicates, catch more overbilling, and close the books faster at month end. The question is rarely whether to do three-way matching — it's how to do it at invoice-volume velocity without drowning the team. That's exactly where automation earns its keep.
Written by
Sarah Blake
Head of Product
Former AP Manager at a $200M construction firm, now leads product at Covinly. Writes about what AP teams actually need from automation — beyond the marketing promises.
View all posts