Cost per invoice: how to calculate it for your AP team

Cost per invoice is the total cost of your AP process in a period divided by the number of invoices processed in that period. The total should include staff time at a fully loaded rate, the extra time exceptions take, software and bank fees, and money lost to late payments. The number only holds up if you measure your own handling times instead of borrowing someone else's benchmark.

#Why the number matters

Cost per invoice is the baseline for any AP automation business case. It tells you what a unit of work costs today, so you can judge what automation might save and how long payback takes. It also shows where the money goes. When teams measure it properly, they often find that routine invoices are cheap and a minority of exceptions eat most of the labor.

We don't quote an industry average here. Published figures vary widely by source, invoice mix and what each study counted, and many don't state their method. Your own figure, measured the same way each quarter, is more useful than a benchmark you can't reproduce.

#The cost components

Count each of these. Teams that count only salaries understate the cost.

#Loaded hourly rate

The loaded rate is what an hour of someone's time costs the company, which is more than what appears on a pay slip. Start with annual salary, then add benefits, payroll taxes, equipment, software seats and a share of office and management costs. Divide by productive hours in a year. That's fewer than 2,080 once you take out leave, training and meetings. Your finance team probably has a standard loaded rate already. Ask for it, because a made-up one invites an argument later.

#Approver and requester time

Approver time is easy to overlook. A department head who spends 20 minutes a week approving invoices is part of the cost even though they never open the AP system. For each approval step, estimate minutes per invoice and multiply by that person's loaded rate. It also hides a benefit of good automation: if approvers get a clean summary on their phone, that time drops.

#How to calculate it, step by step

1. Choose a period. A month or a quarter, ideally not a close or holiday outlier.
2. Count invoices. Total invoices received, including credit notes, and the number you actually processed. If PO and non-PO invoices are handled differently, count them separately.
3. Time a sample. Have clerks log minutes for 30 to 50 invoices from capture to posting, split into simple and exception cases. Use the team's own timing, not an estimate from a meeting.
4. Convert time to cost. Minutes ÷ 60 × loaded rate, for every role in the chain.
5. Add the non-labor items. Software, payment fees, late fees, and the monthly share of overhead.
6. Divide. Total cost ÷ invoices processed.
7. Break it out. Show the figure for the simple group, the exception group and the blended average. The split tells you more than the headline.

Written out:

Cost per invoice =
  (Labor cost + Exception cost + Rework cost + Software
   + Payment costs + Late fees and lost discounts + Overhead)
  ÷ Invoices processed in the period

#A worked example (illustrative numbers)

These numbers are invented to show the arithmetic. They don't describe any real company, and yours will differ.

Suppose a team processes 2,000 invoices a month, with a loaded AP rate of $40 an hour.

Simple invoices. 1,500 invoices (75%) take 6 minutes of AP time on average.

• 1,500 × 6 ÷ 60 × $40 = $6,000

Exception invoices. 500 invoices (25%) take 25 minutes of AP time each.

• 500 × 25 ÷ 60 × $40 = $8,333

Approver and requester time. Each invoice takes 3 minutes of approver time at a $60 loaded rate.

• 2,000 × 3 ÷ 60 × $60 = $6,000

Other costs per month.

Total. $6,000 + $8,333 + $6,000 + $2,500 + $800 + $1,200 + $1,000 + $1,500 = $27,333

Cost per invoice. $27,333 ÷ 2,000 = about $13.67

Now read the breakdown. Exceptions are 25% of the invoices but about 30% of total cost. The 1,500 simple invoices account for only about 22%, and approver time is just as large. A project that only speeds up data entry goes after that 22%. In this example the better targets are fewer exceptions and a lighter approval step.

Key point: Cost per invoice is a diagnostic, not a goal. The useful output is the split: how much goes to exceptions, approvals, rework and late fees. That split tells you what to fix first.

#Using the number to judge automation

With a baseline in hand, model what automation changes. Keep the assumptions explicit and conservative.

• Simple invoices need fewer touches, with the system handling capture and coding and a person confirming.
• Exceptions fall as matching improves and PO discipline tightens, though some will always remain.
• Approvals speed up when requests arrive with the match result attached.
• Rework shrinks when validation catches duplicates and arithmetic errors before posting.

Then subtract what automation adds: software or model usage, hosting, monitoring, and the review time that remains. Our ROI of AI automation method covers one-time costs, ramp-up and the payback calculation.

If the new cost per invoice is only slightly lower than the old one, say so and ask whether the project is worth doing. At low volumes it often isn't.

#Common mistakes

• Using base salary instead of the loaded rate, which understates labor cost.
• Averaging simple and exception invoices. The blend hides where the cost sits.
• Ignoring approvers and requesters. Their time is real even when it isn't on the AP budget.
• Counting only invoices received. If you reject or return some, decide how to treat them and stay consistent.
• Comparing against a benchmark with an unknown method. If you can't see how it was calculated, you can't compare.
• Forgetting the cost of mistakes. A duplicate payment clawed back six weeks later is expensive even when rare.

#Where to go next

The cost of each stage depends on how invoices flow. Our guide to accounts payable automation walks through the five stages and shows where AI, rules and people each fit. If matching is your main cost driver, read about three-way matching.

We can help you build the baseline and tell you honestly whether automation pays back. Book a strategy call and bring a sample of invoices and your current process notes. You can also see how our finance operations work is structured.

#Frequently asked questions

#How do you calculate cost per invoice?

Add up all costs of processing invoices in a period: labor at a loaded hourly rate, exception handling, rework, software, payment fees, late fees and overhead. Divide by the number of invoices processed in that period. Measure handling time on a real sample instead of estimating it.

#What is included in invoice processing cost?

Include the time of AP staff, approvers and requesters, software and system costs, bank and payment fees, late payment penalties, missed early-payment discounts, rework and a share of overhead. Many teams leave out approver time and the cost of exceptions, which understates the total.

#What is a good cost per invoice?

There's no single right answer, because it depends on invoice mix, volume, approval steps and what you count. A better question is whether your figure is falling over time and where the cost sits. Track the split between simple invoices and exceptions, and compare against your own history.

#How do I find my loaded hourly rate?

Add salary, benefits, payroll taxes, equipment, software seats and an allocation of management and office costs. Divide by productive hours per year, which is fewer than total working hours once leave and training are removed. Your finance or HR team usually has a standard rate to use.

#Does automation always reduce cost per invoice?

No. It tends to help most at higher volumes, with many layouts, PO matching or several approval steps. At a few dozen invoices a month, the software and maintenance cost can outweigh the time saved. Calculate your baseline and the post-automation estimate before committing.

#How often should we recalculate it?

Quarterly is a sensible rhythm, using the same method each time. Recalculate after any major process change, such as a new ERP, a new approval policy or an automation go-live, so you can attribute the change to its cause.