Accounts Payable KPIs
Accounts Payable KPIs: definitions, formulas, benchmarks and targets
In summary: Accounts payable KPIs measure performance across six categories: speed, cost, accuracy, cash and working capital, supplier relationships and governance. The six most important AP KPIs are cost per invoice, cycle time, touchless rate, exception rate, on-time payment rate, and human override rate. The current best-in-class figure for processing an invoice is $2.65 per invoice in 2.9 days. Governance KPIs and developing methods to measure them are currently on the rise. Always measure KPIs against your own baselines first.
The metrics that accounts payable teams are measured by have barely changed in twenty years. However, thanks to technology, the work behind them has evolved significantly. For instance, an AP clerk who used to devote their day to keying invoice headers and chasing PO mismatches is now supervising a platform that does both and spending their time on exception management instead.
And while this shift has not changed the definitions, it has changed what a good number looks like and what it takes to maintain it. Finance leaders are invited to use this free guide to assess their AP operations. We’ll be covering the KPIs that run a modern AP function including the formulas, current benchmarks, and the governance questions that arrive once AI is making the bulk of decisions.
What do accounts payable KPIs measure?
Accounts payable KPIs belong to six categories of performance, each measuring AP performance a different way:
- Speed measures how quickly an invoice moves from arrival to payment readiness, through cycle time, approval time and throughput per person. A slow process shows up as missed discounts and suppliers chasing payment.
- Cost is what processing an invoice takes once labor, software and overhead are counted. Cost per invoice is the headline figure, but it’s really the culmination of everything else on this list rather than something a team can tackle directly.
- Accuracy covers how often invoices are wrong, duplicated or need a person to intervene, measured through exception rate, first-time match rate, and duplicate payment rate and cancellation/reversal rate. These numbers quietly determine what processing costs.
- Cash and working capital describe how long the organization holds its money before paying, and what it gives up or gains by holding it. Days payable outstanding and discount capture sit here, and both relate closely to supplier goodwill. It is worth emphasizing that AP teams can not influence these KPIs but rather sourcing and procurement.
- Supplier relationships measure whether suppliers are paid when promised and get a straight answer when they ask, through on-time payment rate and query volume. Approvers outside of AP also influence these KPIs by delaying or speeding up approvals in the workflow.
- Governance asks whether controls are held and whether the reasoning behind a decision can be produced on request, covering clear approval history, as well as decision logic/reasoning and records. This is the category that changed most once machines began making decisions.
AP KPI Benchmarks at a glance
KPI | Formula | Industry average | Best-in-class |
|---|---|---|---|
Cost per invoice processed | Total AP cost/invoices | $9.84
| $2.65 |
Invoice cycle time | Days from arrival to the posting/payment readiness | 8.2 days
| 2.9 days |
Exception rate | Invoices needing manual intervention / all invoices | 18.4%
| 11.1% |
Touchless rate | Invoices with zero manual intervention / all invoices | 35.4%
| 51%+ |
Staff time on supplier queries | AP staff hours on supplier enquiries / total AP hours | 21.9%
| Roughly half |
These are market-wide figures from Ardent Partners’ survey population and do not reflect the performance of Springtime customers. Most surveyed teams are still partly manual, so treat the average column as a baseline rather than a target.
Source: Ardent Partners, The State of ePayables 2025: AP’s Unfinished Journey, as published in State of ePayables (Part Nine): AP Benchmarks and Best-in-Class Performance, Andrew Bartolini, 22 January 2026.
Core AP KPIs: definitions and formulas
Together, this list of core AP KPIs can paint a vibrant picture of your organization’s accounts payable performance and help leaders quickly understand whether invoices move through the process efficiently, accurately and cost-effectively.
Speed
Invoice cycle time
Invoice cycle time is how many days it takes to move an invoice from receipt to posted and ready for payment, or to approved – some organizations define the “last day” differently.
Formula: average number of calendar days between invoice arrival and the point you measure to.
Approval cycle time
Approval cycle time is how long it takes to move an invoice from workflow start to approved and scheduled for payment.
Formula: average number of calendar days from invoice receipt to approval, stopping at the approval date and excluding the days between approval and actual payment.
Invoices per FTE
Invoices per FTE is how many invoices one full-time equivalent processes in a defined period (some organizations choose to calculate per year, others per day).
Formula: total invoices processed in a year divided by the number of full-time equivalents in AP.
Published figures for this metric disagree more than for any other AP KPI. APQC puts the cross-industry average at 10,853 invoices per FTE per year, rising to roughly 23,000 for organizations receiving at least 70% of invoices electronically.
Other surveys of the same top cohort report figures ranging from 6,900 to 42,000. Most of that spread is definitional rather than real, which makes this a metric to track against your own baseline rather than against anyone else’s number.
Cost
Cost per invoice
Cost per invoice is what it costs you to handle one invoice from the moment it arrives to the moment it’s paid.
Formula: total cost of running AP for a period divided by number of invoices processed in that period.
Different organizations will measure the total cost using different criteria. Leaving out some items for others can change the cost by 50% or more. Therefore, this list should be considered an example of how you can structure the total cost, as opposed to a hard rule. Many of Invoicetrack’s users calculate the total costof running AP to be the sum of:
- Wages and time of everyone who processes invoices (more accurate if you include a share of their manager’s time)
- AP software licenses and any per-invoice or per-transaction fees
- A fair share of overhead (office space, IT, general admin) for the AP function
- The cost of chasing and fixing errors and exceptions (rework, duplicate recovery, supplier queries)
The performance of a team’s accounts payable process can be measured across several categories, including speed, cost, accuracy, cash and working capital and more. We go into significant detail on the topic of accounts payable KPIs in this article.
Accuracy
Touchless rate
Touchless rate is the share of invoices that go all the way through on their own, with nobody from AP touching them. Some organizations also refer to this as their straight-through processing rate.
Formula: invoices processed with zero manual interventions divided by total invoices processed.
Touchless rate figures remain one of the more elusive AP KPI figures to benchmark accurately because many published figures contradict one another. Ardent Partners’ State of ePayables 2025 puts the market average at 35.4% while SSON’s State of Accounts Payable Market Report 2026 paints a worse picture from practitioner polling, with most organizations reported to be running below 25% touchless and only 7% above 75%. Read together, they suggest a market average somewhere in the mid-twenties to mid-thirties depending on how the question is asked, leading teams reaching roughly half, and genuinely touchless operations above 75% remaining rare.
Exception rate
Exception rate is the share of invoices that get stuck and need a person to sort them out.
Formula: invoices requiring manual intervention divided by total invoices processed.
This metric is particularly popular when it comes to demonstrating performance, because a stuck invoice can cost five to ten times more to handle than one that seamlessly flows through a workflow.
First-time match rate
First-time match rate is the share of invoices that line up correctly against the purchase order, and the receipt for a two or three-way match, on the first attempt with no manual correction.
Formula: invoices matched successfully on first pass divided by total PO-backed invoices.
Cash and working capital
Days payable outstanding
Days payable outstanding are the average number of days the business takes to pay a supplier after recording the bill. It shows how long cash stays in the business before it goes out to suppliers.
Formula: average accounts payable balance divided by cost of goods sold multiplied by 365 days.
Discount capture rate
Discount capture rate is the share of early-payment discounts you catch when suppliers offer them.
Formula: value of early-payment discounts captured divided by value of discounts offered.
Teams tend to miss discounts for two reasons, and only one is about speed. The obvious one is timing because slow processing means the early payment window closes before the invoice is even approved. The less obvious one, yet more costly, is visibility. When terms sit inside a PDF, a non-PO invoice, or a contract that never reaches the person scheduling the payment, the option is lost before the clock even starts running. Closing this metric means surfacing the discount at the point of ingestion, then moving fast enough to act on it.
Supplier relationships
On-time payment rate
On-time payment rate is the share of invoices you pay by their due date.
Formula: invoices paid by the due date divided by total invoices due in the period.
Duplicate payment rate
Duplicate payment rate is the share of payments where the same invoice was paid twice.
Formula: duplicate payments identified divided by total payments issued.
Staff time on supplier queries
Staff time on supplier queries is the share of AP hours spent answering suppliers rather than processing invoices. It reads as a service metric and behaves as an exception metric, because most queries trace back to an invoice that stalled.
Formula: AP staff hours spent on supplier enquiries divided by total AP hours.
Why two companies calculate the same KPI differently
Comparison is a useful exercise to understand if you clear a baseline. Where it quickly becomes unhelpful is when you trust benchmarks blindly, without applying the specific context of your own organization or when you fail to consider that your contemporaries may be measuring the same figures a bit differently.
Accounts payable KPIs are the textbook example of this problem playing out in real life. Here are six examples of why two companies may calculate the same KPI differently:
Touchless rate depends on the denominator: non-PO invoices auto-posting through coding rules, or only true three-way-matched invoices? Including the former inflates it.
Cost per invoice depends on how much you count. A thin version counts only labor while a rigorous one adds software, overhead, payment fees, rework and missed-discount leakage.
Cycle time depends on where you start and stop the clock. Receipt-to-approval and receipt-to-payment are different metrics, and the start point, invoice date or arrival in AP, shifts the number.
Exception rate comes down to what each company logs as an exception. Price and quantity mismatches only, or missing POs, tax errors and coding rejects too?
Invoices per FTE vary on both sides of the ratio: invoices or line items, and processors only or the full AP team (leads, back office and approvers) too? Business partners should be excluded because folding them in makes the metric unable to focus on AP performance.
DPO is largely shaped by sampling. A balance checked on the last day of the quarter, just after a big payment run, reads low. Averaging across the period is the honest method. Tax and payroll sitting in the same account will inflate it.
AP KPI cheat sheet: what to measure and what to aim for
Use this to set your own numbers. Pin down the boundary before the target. A cycle time that stops at approval will always beat one that runs to payment, and that gap says nothing about performance.
KPI | Boundary to identify | Working target |
|---|---|---|
Cost per invoice | Whether software, overhead, payment fees and rework sit inside the cost | Under $3.00
|
Invoice cycle time | Receipt to approval, or receipt to payment readiness | Under 5 days
|
Approval cycle time | Stops at approval, excludes approval to payment readiness | Under 2 days
|
Touchless rate | Whether non-PO invoices auto-posting through coding rules count or invoices approved by business partners in automated approval workflows count | 70%+
|
Exception rate | What the team logs as an exception | Under 10%
|
First-time match rate | Two-way or three-way; invoice level or line level | 90%+
|
Invoices per FTE | Invoices or line items; processors only or approvers too | 20,000+ |
On-time payment rate | Paid on the due date, not before it | Above 95% |
Days payable outstanding | Average payables not period-end; trade payables only | Agreed terms, paid on time |
Discount capture rate | By value, not by count | Above 80% |
Duplicate payment rate | Detected before payment or recovered after | As close to zero as controls allow |
Tip: set your own against your baseline and sector rather than a figure from a competitor’s page.
Governance metrics for an AI-driven AP function
Of course, AI is not a new lever for AP automation by any means, but its capabilities and usage prominence exploded in 2026. Global enterprises and the tools they use to work are under pressure to fully leverage everything modern AI has to offer. But alongside this call to arms is a burning question that requires a bit of forward thinking to fully answer: “How much can we trust it?” followed by, “How do we measure that trust?”
Here’s a list of the most useful governance metrics to track the effectiveness and behavior of your AI-driven AP processes:
Human override rate
Human override rate is the share of AI decisions a person reverses, calculated by the total number of overturned automated decisions divided by the total automated decisions. This metric helps quantify how much the system can be trusted to operate independently.
Decision traceability
Decision traceability is the share of automated decisions for which a complete decision-making rationale can be retrieved (not reconstructed) on demand. This means what the AI observed, which rule or model fired, and why it was approved or flagged. You can calculate this by dividing decisions with a complete audit trail by total automated decisions.
Policy violation rate
Policy violation rate measures how often an automated action breaches defined controls like missing a tax validation, bypassing an approval threshold, or failing to adhere to the rules of a mandate (e.g., your automation processes an invoice that doesn’t meet the required e-invoicing format or clearance for that country)
Model drift on match accuracy
Where first-time match rate reveals how accurately the AI matches invoices today, the governance metric investigates whether that number is holding. An AI model can decay quietly as suppliers, PO formats, and coding rules change out from under it, so nothing breaks, the system keeps running and reporting confident answers, it’s just wrong more often than it used to be. The metric worth instrumenting is the trend: is match accuracy steady, or slipping against where it started? Unlike a stuck invoice, this failure doesn’t announce itself, which is exactly why it belongs in the governance category rather than the accuracy one.
AP Governance KPIs: what to measure and what to aim for
Metric | Formula | Working target | How it fails |
|---|---|---|---|
Human override rate | Reversed automated decisions/all automated decisions | Falling and stable; a sudden rise is the signal
| A very low rate can mean nobody is checking rather than the system being right |
Decision traceability | Decisions with a complete retrievable rationale/all automated decisions | 100%, available on request rather than rebuilt afterwards | Reconstructed logs pass an audit and cannot answer a supplier in real time |
Policy violation rate | Automated actions breaching a defined control/all automated actions | Zero, with every instance investigated
| Violations only surface if the control is instrumented in the first place |
Match accuracy drift | Change in first-time match rate against its own baseline over time | Flat against baseline
| The system keeps running and reporting confidently while getting slowly worse |
Staff time on supplier queries | AP staff hours on supplier enquiries / total AP hours | 21.9%
| Roughly half |
No published benchmark exists for these figures at time of publishing. Measure them against a baseline taken over a full quarter, then against the trend, and treat the trend as the number that matters.
How to benchmark AP performance across multiple countries
Mature global organizations centralizing their invoice processing into a set of shared service centers need to set their benchmarks with this single idiom in mind: the devil’s in the details.
Every country is affected by a list of unique nuances that can impact their ability to process invoices quickly and touchlessly – ignore them and you risk building a list of vanity metrics that’s really just fiction.
Global enterprises should follow these four steps to benchmark AP performance across their multi-country mix accurately:
- Fix the definitions before comparing anything. Cost per invoice has to include the same ingredients everywhere, and cycle time has to start and stop on the same events. If two countries measure the same KPI with different recipes, every comparison after that is worthless. Write one documented formula per KPI and apply it identically across the estate.
- Group like with like before ranking. A country drowning in non-PO invoices looks worse on match rate and cost because its work is harder, not because its team is weaker. Comparing its blended number against another country’s tells you nothing unless you know what’s inside each. So measure PO and non-PO invoices separately, and compare like against like, PO against PO, before you rank anyone.
- Normalizing structural metrics. Splitting PO from non-PO removes the largest source of noise, but what remains is still not comparable. A shared service center runs one team, one system and one pay scale, so a cost gap between Germany and Italy is most likely a difference in the invoice mix instead of labor cost related. Expect heavier tax and regulatory regimes to run higher cycle times and exception rates, because that’s just compliance working, not AP failing. Supplier master data is the factor most often missed: incomplete bank details, duplicate vendor records and missing tax IDs generate exceptions that read as process failure when the cause sits upstream.
- Benchmark internally first and externally second. Your best-run country is your most honest target, it already operates inside your systems, policies and supplier base, so asking why one country hits a number another can’t is a fair, answerable question. Rank your countries against each other and against their own trend over time. Use external benchmarks like Ardent, APQC or SSON only as a floor check: do we clear the baseline, yes or no.
A practical example on multi-country benchmarking
Imagine you are running a shared service center in Krakow with invoices funneling in from both Germany and Italy and the two entities still sitting on separate ERP instances. Your reporting tells you Germany has a first-time match rate of 92% while Italy is barely exceeding 70%. Your first instinct would be to conclude that Italian invoice processing is broken but you’d be doing your AP team a disservice.
Three checks will help you decide whether the gap is signaling a crisis or is a non-event:
The first is whether both are counted the same way. Match rate can be measured at invoice level or at line level, and the two produce very different numbers for identical work. A 10-line invoice with one price mismatch fails outright at invoice level and scores 90% at line level. Where entities are continuing to report on their own ERP, this convention is usually inherited rather than chosen.
The second is whether the invoice mix is comparable, as Italy may simply be handling more non-PO work, which is harder to match automatically.
The third is what’s structural, because Italy’s tax and e-invoicing rules differ from Germany’s regardless of how the team performs.
If after completing all three checks you discover the gap still holds, then it’s a lead worth investigating more thoroughly.
All three checks come down to the same thing: whether the numbers in front of you were produced the same way. In a multi-entity estate that question is usually unanswerable from the reports themselves, because each ERP calculates and presents its own version. The fix is structural. Move the calculation above the source systems, so one definition is applied once and the working behind every figure stays visible.
How Invoicetrack reports AP KPIs
Invoicetrack reports AP KPIs through two layers: most teams receive a library of standard operational reports from go-live, task-oriented views like open approval tasks with aging, unaccounted documents, and non-compliant invoices. On top of that sits Beachwalk, Invoicetrack’s embedded AP analytics engine, which handles the trending, drill-down, and cross-document pattern analysis that operational reports aren’t built for.
Beachwalk covers AP performance across a set of report families spanning automation, invoice processing, control and compliance, and operational productivity. Natively it reports the core KPIs this guide has walked through, touchless rate, cycle time, exception rate, alongside FTE productivity, posting automation, and SLA adherence. Customers appreciate the ability to drill-down on performance, as every KPI traces back to the individual invoices, workflow events, and approval decisions behind it, so a number on a management view can be followed all the way to line-item evidence. That traceability is what turns a KPI from a figure into something an auditor or process owner can truly investigate.
FAQs
Cost per invoice, invoice cycle time, touchless rate, exception rate and on-time payment rate cover most of what an AP function needs to know. Between them they measure what processing costs, how fast it runs, how much of it happens without a person, how often something goes wrong, and whether suppliers get paid when they were promised. Add days payable outstanding if the CFO is the audience, since it is the metric that turns AP into cash.
The best-performing fifth of AP teams process an invoice for $2.65, against $12.42 for everyone else, a gap of roughly 79%. Ardent's market-wide figure is $9.84. Under $3.00 is a reasonable target for a function that has automated capture, matching and approval routing. The figure only means something if the scope is stated, because a version counting only labor will always look better than one that includes software, overhead, payment fees and rework.
35.4% is the current market average and 51.0% is being reported as best-in-class – though many global enterprises achieve higher touchless rates when supported by sophisticated AP automation platforms like Invoicetrack. Teams with structured invoice data arriving at the source can go higher, and 70% is a defensible target once most spend sits on a purchase order. The number depends heavily on the denominator: counting only PO-backed invoices inflates it, so state whether non-PO invoices auto-posting through coding rules are included.
Average the calendar days between an invoice arriving and the point being measured. That end point is the part that matters because receipt to approval and receipt to payment readiness are different metrics and produce very different numbers, so pin down both ends before comparing against anyone. Ardent puts the market average at 8.2 days against 2.9 days for best-in-class.
Not cost per invoice, even though it is usually the one under pressure. Cost is an output of everything upstream, so it cannot be managed directly. Touchless rate, exception rate and first-time match rate are the levers, and cost per invoice falls on its own once those move, with cycle time following. A rising cost per invoice is a symptom, and the cause is almost always a drop in touchless processing or a spike in exceptions.
Through governance metrics rather than efficiency ones: human override rate, decision traceability, policy violation rate, and drift in match accuracy against its own baseline. Traceability is the one to watch, because it asks whether the reasoning behind a decision can be retrieved on request rather than reconstructed afterwards. No published benchmark exists for any of these yet, so measure against a baseline taken over a full quarter and track the trend.