Search for a cost-per-invoice benchmark and you will find a dozen confident figures, most of them uncited, several of them contradicting each other, and a few widely-repeated pairings that do not appear in the reports they are attributed to. Before quoting one to a CFO it is worth knowing which ones survive being checked.
Two do. Below are the figures we verified against the primary documents, with the sources listed at the foot of this article so you can check them yourself.
The verified numbers
What both datasets agree on is the shape: a long right tail, with the best performers clustered several times below the median rather than marginally below it. That shape is the interesting part, because a five-fold spread is not explained by staff cost.
What is actually inside the number
All-inclusive cost per invoice covers labour, systems, overhead and the cost of managing exceptions, divided by volume. Two organisations can report wildly different figures while doing very similar work, because the denominator and the boundaries move:
- Whether PO matching sits inside AP or in procurement.
- Whether supplier onboarding and master data maintenance are counted.
- Whether the payment run itself is included or treated as treasury.
- Whether credit notes and self-billed documents count as invoices.
This is why an externally quoted benchmark is a weak management tool and an internally consistent baseline is a strong one. Measure your own, define the boundary explicitly, and compare against yourself over time — the trend is reliable in a way the absolute comparison is not.
The lever that separates the quartiles
Three further figures from the same Ardent report explain most of the gap, and none of them is about the price of labour.
Put together, the mechanism is visible. An invoice that flows straight through costs almost nothing to process. An invoice that stops — because the PO does not match, the receipt is missing, the approver is on leave, the price differs by a rounding error — consumes a person for several minutes and often several days of elapsed time, plus follow-up. The average cost is therefore not the cost of processing an invoice. It is the cost of the clean ones spread across the cost of the broken ones.
That reframes the improvement question entirely. The lever is not making exception handling faster. It is stopping invoices from becoming exceptions.
Where exceptions come from
In practice the causes are boringly consistent, and almost all of them originate before the invoice arrives.
| Exception | Where it was actually caused |
|---|---|
| No purchase order | Requisition process was slower than buying directly |
| Price mismatch | PO raised from an outdated price, or contract price never loaded |
| Quantity mismatch | Goods receipt never entered, or entered against the wrong line |
| Wrong entity or address | Supplier master record has stale or duplicated details |
| Approver unavailable | Delegation rules not maintained |
| Duplicate invoice | Supplier resent because the first one was never acknowledged |
This is the uncomfortable conclusion of any serious look at AP cost: the AP team is where the cost is measured, and almost nowhere the cost is caused. Squeezing AP harder therefore has a low ceiling. The improvements sit upstream, in requisitioning, in PO discipline, in goods receipting and in supplier master data.
A sequence that works
- Baseline your own cost per invoice with an explicitly stated boundary. Without this, nothing afterwards is measurable.
- Categorise a month of exceptions by cause, not by symptom. Expect the top three causes to account for most of the volume.
- Fix the largest upstream cause first, even though it lives in another team's process. This is where the money is.
- Only then look at automation. Capture technology applied to a process with a 20% exception rate automates the easy 80% and leaves the expensive part untouched.
- Re-baseline. If the cost has not moved, the exception mix has not moved either.
Step four is where most programmes are inverted, and it is worth being blunt about why. Automation is procurable — there is a vendor, a demo and a business case. Fixing goods receipting discipline in three business units is not procurable; it is a series of unrewarding conversations. The first is much easier to start and the second is what actually moves the number.
Common questions
What is a typical cost per invoice?
Ardent Partners' 2025 study puts the average all-inclusive cost at $9.40, with Best-in-Class enterprises at $2.78 against $12.88 for all others. A separate APQC dataset reported by CFO.com gives $2.07 at the 75th percentile, $5.83 at the median and over $10 at the 25th percentile. The two studies use different populations and definitions, so figures should not be mixed between them.
Why do published cost-per-invoice benchmarks vary so much?
Partly because the process boundary differs — whether PO matching, supplier onboarding, master data maintenance or the payment run are counted — and partly because several widely-repeated figures are miscited. Your own consistently-measured baseline, with the boundary stated, is a far more useful management number than any external benchmark.
What causes the gap between top-quartile and average AP performance?
Exception rate, more than anything else. Ardent reports a 14% average exception rate and 53% of respondents naming exceptions as their top challenge, with touchless processing at 32.6% overall against 49.2% for Best-in-Class. A clean invoice costs almost nothing to process; the average is really the cost of the clean ones spread across the broken ones.
Where do invoice exceptions actually originate?
Almost always upstream of accounts payable: no purchase order because requisitioning was slower than buying direct, prices raised from stale data, missing goods receipts, stale or duplicated supplier master records, unmaintained approval delegations, and duplicates caused by suppliers resending unacknowledged invoices. AP is where the cost is measured, not where it is caused.
Will automation reduce our cost per invoice?
It will reduce the cost of the invoices that were already going to flow through cleanly, which is not where the money is. Applied to a process with a high exception rate, capture automation handles the easy majority and leaves the expensive minority untouched. Categorise and fix the top exception causes first, then automate — the same tooling delivers substantially more against a clean process.
Sources
Want this run for you?
We take on the transactional half of procurement — invoices, purchase orders, supplier data and indirect spend — inside your own systems and under your approval rules. Start with a free spend audit: we measure your volumes, cycle times and exception rates, and the report is yours whether or not you go further.
Book a free spend audit
