AP cost and benchmarks
What does it actually cost to process one invoice?
Most AP cost-per-invoice numbers are built to sell software. Here is the arithmetic with the assumptions visible, including the lines vendors leave off the slide.
Vendors will tell you an invoice costs about $2 to process. Build the number from the bottom for a portfolio doing it by hand, and labor alone comes to $8.67, before the software bill.
The gap is two costs that almost never make it onto the slide: the approver, and the invoices that go wrong.
Here is the whole calculation, worked at 2,500 invoices a month, with every assumption visible enough to argue with.
Start from 2,500 invoices a month, 10 minutes each, $35 an hour
Those three are inputs, and they are the defaults our ROI calculator starts from. Every line below is arithmetic on them, so redo it with your own counts.
The number to watch as it builds is the gap between invoices handled and invoices a person actually touched. The whole cost model lives in that gap.
Clerk time alone is $5.83 an invoice
Start with the part everyone agrees on.
A clean vendor invoice, keyed by hand into a PMS, takes about 10 minutes end to end. That is not just typing. That is opening the email, saving the PDF somewhere, finding the vendor in the supplier master, keying the header, keying the lines, picking the GL account, picking the property, and then chasing the thing that does not match.
At a fully loaded AP cost of $35 an hour, 10 minutes is $5.83 per invoice.
2,500 invoices at $5.83 is $14,583 a month, or $175,000 a year, in clerk time alone.
Ten minutes is probably not exactly right for your shop. It is low if your vendors send handwritten field invoices, and high if you are mostly utility bills on a template. Time your own team over a morning and substitute the number. The structure does not change.
The approver adds $2.40 an invoice, and nobody counts it
Here is where most cost-per-invoice models quietly break.
In multifamily, the invoice does not sit in AP. It sits with a regional. A property manager approves, then a regional approves, then anything capital goes to a CapEx director. Every one of those is a person more expensive than the AP clerk, and every one of them is doing approvals between property visits.
Say each invoice averages 1.6 approval touches, at 90 seconds of actual attention each, at a blended $60 an hour. That is $2.40 per invoice, or $6,000 a month.
You will notice the attention number is small and the cost is not. That is the point. Approval cost is driven by how many approvals there are and who has to do them, not by how long each one takes.
An exception invoice costs 2.5x a clean one
Not every invoice costs the same, and averaging hides it.
Say 5% of the 2,500 need a chase. That is 125 invoices: supplier code mismatch, property code mismatch, price mismatch against the PO, a missing invoice number, an extraction that just failed. An exception runs 25 minutes rather than 10, because it involves a second person and usually an email to the vendor.
Five percent is an input. If your team chases one invoice in ten, use 10% and the exception line doubles.
So split the population:
| Population | Count | Minutes each | Cost each | Monthly |
|---|---|---|---|---|
| Clean | 2,375 | 10 | $5.83 | $13,854 |
| Exception | 125 | 25 | $14.58 | $1,823 |
| Approvals (all) | 2,500 | 1.6 touches | $2.40 | $6,000 |
Straight labor lands at $21,677 a month, which is $8.67 per invoice.
That is the all-in labor number for the worked example, and it is more than four times the $2 figure a lot of vendors put on a slide.
The per-page tool added close to $10,000 a month at ResProp
Now add the tools, and this is where property management gets its own particular problem.
ResProp Management runs about 24,000 units out of Austin. Their legacy invoice solution charged per page, and at the portfolio's volume it was tracking toward nearly $10,000 a month. Per page, not per invoice. If a vendor sends a 6 page statement with one invoice buried in it, you are billed for 6. If a vendor attaches their W-9 and a certificate of insurance to be helpful, you are billed for those too.
We wrote about why that pricing model punishes exactly the documents you cannot control in why per-page invoice pricing is a tax on your W-9s.
For comparison, BillRoute is $99 a month plus $1 per invoice uploaded. At the worked example's 2,500 uploads that is $2,599 a month. Treat that comparison with the suspicion it deserves, because we sell one of these. The narrower point stands on its own: per page and per invoice are different units, and the difference is not a rounding error at portfolio scale.
Labor alone is $8.67 an invoice, not $2, and the software line goes on top
At 2,500 invoices a month, labor runs $21,677, or $8.67 per invoice. Your current tool is the line to add, in the same unit. If it bills per page, that line can be the largest one on the sheet.
That is more than four times what a cost-per-invoice benchmark would have told you, because the benchmark left out the regional and left out the exception tail.
Automation removes the touches, not the minutes
This is where AP marketing usually gets loose.
Automation does not make the 10 minutes go to zero. It makes most invoices stop being touched at all, and leaves you the ones that genuinely need a person. ResProp runs at a 98% automation rate, meaning 98% of invoices require no human touch. The 2% that reach a person are the system doing its job, which is to hand you the invoices that are actually wrong.
The measured outcome at ResProp was 40 hours a month saved per team and more than $150,000 a year in cost. Roughly $120,000 of that is the replaced per-page tool at nearly $10,000 a month.
What automation does not fix:
- A vendor who sends the wrong invoice. Extraction reads what is on the page. If the page is wrong, you get a correct reading of a wrong invoice.
- An approver who does not open their queue. Routing is a technology problem, urgency is a management one.
- A chart of accounts nobody agrees on. If two properties code the same expense differently, an AI trained on your history learns to be inconsistent too, faithfully.
The third one is the quiet killer, and it deserves saying plainly. Coding automation is only as good as the coding history you feed it. If your GL is a mess, automating it produces a faster mess.
Four inputs get you within 10% of your own number
You do not need a calculator to get within 10% of your real figure. You need four things.
First, your monthly invoice count, split into clean and exception. Pull the exception count from however you currently track "stuff AP had to chase."
Second, your fully loaded AP hourly cost. Salary plus burden, not salary.
Third, your average approval touches per invoice. Count the approval steps in your workflow and multiply by the share of invoices that take each path.
Fourth, your current software bill, converted to the same unit. If you are billed per page, get last month's page count, not last month's invoice count. Those two numbers are usually further apart than people expect.
Multiply it out. The worked example used 2,500 invoices, 10 minutes, $35 an hour, and a 5% exception rate. Yours differ on all four. Then compare it to whatever your AP software costs, and to whatever a vendor's ROI calculator told you.
What did your per-invoice number come out to?
Keep reading
AP cost and benchmarks
Per-page invoice pricing is a tax on your W-9s
If your AP tool bills per page, you pay for every certificate of insurance, every multi-page statement, and every document a vendor attached to be helpful. Here is the arithmetic.
AP cost and benchmarks
Slow vendor payments cost the vendor first and your AP team second
A vendor offering 2/10 net 30 prices 20 days of waiting at 2%. That is their number for payment speed. Your AP team pays for the same wait in phone calls.
AP cost and benchmarks
Slow AP costs you twice: the discounts you miss and the late fees you pay
A 2/10 net 30 discount is a 36% annualized return on paying early. Most portfolios cannot take it, because approval takes longer than ten days.