An invoice goes out with the wrong number on it. The client finds it, raises it, and after some back and forth you issue a credit. The credit lands in the ledger, the invoice gets paid, and the event closes.
Ask what that error cost and almost everyone answers with the credit. The credit is the only part of it that has an account code.
It is also the smallest number in the event.
TL;DR
- A billing error costs more than the amount you billed wrong. The credit is one of four costs, and the only one that appears in the ledger.
- The other three are the hours spent resolving it, the carrying cost of the money sitting unpaid while it is argued, and the assurance cost of an invoice you cannot trace back to its events.
- All four can be priced from numbers you already hold, using one formula and a worked example.
- On plausible inputs, a $2,400 overbill costs about $3,204 to resolve. The error costs more than it is worth.
- No credible public benchmark exists for how often client invoices are wrong. We checked six widely-repeated figures and could not source one of them. Your own rate is the only honest input.
- Under-billing costs the same way and never generates a complaint, so it never enters anyone's model.
Short answer
A wrong client invoice costs the credit you issue, plus the loaded cost of the hours spent resolving it, plus the carrying cost of the invoice value while collection is delayed, plus the assurance cost of holding a receivable you cannot trace to its underlying events. Total cost of one billing error = credit issued + (resolution hours × fully loaded hourly rate) + (invoice value × cost of capital × days delayed ÷ 365) + assurance cost. On plausible inputs, a $2,400 overbill on a $40,000 invoice costs roughly $3,204 to resolve, about 1.3 times the value of the error itself. Under-billing carries the same resolution and assurance costs while producing no complaint at all, so it is the half of the problem no dispute log has ever recorded.
One clarification before the arithmetic, because the term is used two ways. On the payables side, invoice accuracy means checking that the supplier invoices you receive match what was agreed. On the receivables side it means the invoices you send your clients are right when they leave. Most published writing covers the payables side. The receivables side is where a wrong client invoice takes money out of your business. Different problems, different owners, and the money moves in opposite directions.
What the credit actually covers
Issue a credit for $2,400 and you have returned $2,400. That is the whole of what the credit does.
It does not pay for the analyst who pulled the transaction detail, or the account manager who took the call, or the second review before the corrected invoice went out. It does not compensate for the twenty-two days the client held the full invoice rather than the disputed portion, which is what clients do. And it does nothing about the fact that the same pricing logic that produced this error is still running.
The credit is the receipt for the mistake. It is not the cost of the mistake.
What does it cost to resolve one dispute?
Resolution is labour, and labour is the line most models understate, because they price it at salary.
Salary is not what an hour costs. In March 2026, US Bureau of Labor Statistics data put private-industry employer compensation at $46.60 per hour worked, of which wages and salaries were $32.60. Wages accounted for 69.9 percent of what employers actually paid; benefits made up the remaining 30.1 percent. So the fully loaded cost of an hour is the salary rate divided by 0.699, which is roughly 1.43 times the number on the payslip.
Price a finance analyst at $60 an hour and the real figure is $85.84. Six hours of dispute work, spread across the analyst, the account owner and the reviewer, costs $515. Not $360.
Six hours is not a heavy dispute. It is one that gets found, explained, credited and reissued without escalation.
What does the delay cost?
A disputed invoice is not a partially paid invoice. It is an unpaid one.
Clients rarely pay the undisputed portion and hold back the rest. The whole invoice sits while the line item is argued, which means the carrying cost applies to the full value, not to the error. This is standard receivables arithmetic: the capital cost of a receivable is the average accounts receivable balance multiplied by your cost of capital, applied across the days it is outstanding and annualised.
For a $40,000 invoice held twenty-two days past terms at a 12 percent cost of capital, that is $289.
It is a small number on one invoice. It is not a small number on a book of them, and the relevant context is that overdue is the normal state of B2B trade rather than the exception. Atradius reports that 43 percent of credit-based B2B sales in the US are overdue, with around 5 percent of long-overdue invoices written off as bad debt. Atradius attributes the delays primarily to customer cash-flow pressure. Billing errors are one contributor sitting inside a much larger overdue pool, and they are the contributor you control.
The cost that only shows up in diligence
The fourth cost has no invoice attached and no line in the monthly close. It arrives later, when somebody outside the company is looking.
Revenue is the account auditors are trained to distrust by default. PCAOB Auditing Standard 2110, paragraph .68, requires that "the auditor should presume that there is a fraud risk involving improper revenue recognition and evaluate which types of revenue, revenue transactions, or assertions may give rise to such risks." A note to paragraph .71 adds that "a fraud risk is a significant risk."
Those standards govern audits of US public companies and SEC-registered brokers. Most fintechs reading this are private and not directly subject to them. That is a scope point. It is not an exemption. It is the standard your future auditor is trained on, and it is the bar you inherit the moment you raise a priced round, sit through acquisition diligence, or take on a client whose own auditors want comfort over the numbers you bill them.
The practical form the exposure takes is simple. An invoice you can trace to its underlying events is evidence. An invoice you can only reproduce by rerunning the system that generated it is an assertion. The difference costs nothing until somebody asks, and then it costs whatever the delay in the diligence timetable costs.
The fix is a build decision rather than a control you can add afterwards, which is the argument in how to build an audit-traceable client invoice.
The cost you cannot put a number on
There is a fifth cost, and honesty requires saying that it does not price cleanly.
A client who finds one billing error starts checking. That is rational, and it is permanent. The checking has a cost to them, and it has a cost to you, because every subsequent invoice now moves through a verification step on their side before it moves through approval. Your collection cycle lengthens by however long that takes, on every invoice, indefinitely.
That is the real trust cost, and it is measurable in days added to collection rather than in dollars of goodwill. Anyone offering you a dollar figure for damaged client trust is estimating.
The formula
Four costs, one line:
Total cost of one billing error = credit issued + (resolution hours × fully loaded hourly rate) + (invoice value × cost of capital × days delayed ÷ 365) + assurance cost
Worked through, with illustrative inputs a finance lead can swap for their own:
| Line | Input | Cost |
|---|---|---|
| Credit issued | Overbilled amount | $2,400 |
| Resolution labour | 6 hours × $85.84 loaded ($60 salary ÷ 0.699) | $515 |
| Carrying cost | $40,000 × 12% × 22/365 | $289 |
| Assurance cost | Not per-invoice; a function of whether the invoice is traceable | Not priced here |
| Total | $3,204 |
A $2,400 error costs about $3,204 to put right, roughly 1.3 times its own face value. The multiplier is the finding. It holds whatever numbers you substitute, because three of the four lines are driven by the invoice value and the resolution effort rather than by the size of the error. The expensive case is a small error on a large invoice held through a long argument.
Substitute your own loaded rate, your own cost of capital, and your own average days to resolve. The arithmetic does not care about the industry.
Why nobody publishes the number you actually want
The obvious next question is what proportion of client invoices go out wrong, so you can multiply.
We went looking for that figure and could not find a defensible one. Six candidates surfaced repeatedly across the field, and all six failed a source check.
The most-repeated is a cost-per-deduction figure attributed to the Credit Research Foundation. Fetching the page it is usually cited from returns a page that does not contain the figure at all. CRF's own deduction survey is member-gated, and the benchmark tool behind it is built on a 2015 dataset. Two other widely-quoted dispute rates carry no attribution and no year on the pages that publish them. One characterisation of the Atradius findings, that one in four B2B invoices are late or disputed, is contradicted by the Atradius page itself, which reports 43 percent overdue and does not name disputes as the cause.
So there is no benchmark here to borrow, and we are not going to invent one. Your own error rate is the only input to this model that has to be real, and it is the one input you already have. Count the credit memos you issued last quarter. Divide by the invoices you sent. That number is worth more than any published average, because it is yours.
The harder measurement problem is the direction nobody counts. Every dispute log records over-billing, because over-billing is what clients report. Under-billing produces no complaint, no dispute, and no log entry. It carries the same assurance exposure and the same broken pricing logic, and it takes the money out of your revenue rather than out of your margin. A billing error rate built from disputes measures one half of the error population and calls it the whole.
The bottom line
The credit is not the cost. It is the part of the cost that has somewhere to go in the ledger.
Price the other three and one wrong invoice runs at roughly 1.3 times the value of the error, before the permanent lengthening of your collection cycle with a client who now checks everything. The multiplier explains why billing errors feel more expensive than they look on paper, and why a run of small ones does more damage than a single large one.
The four inputs are already in your systems. What is usually missing is the fifth thing, which is a way to know an invoice was wrong before the client tells you. That is the difference between an error you priced and an error you discovered, and it is the same discipline applied to the other side of the ledger in revenue assurance for payment businesses. The failure points that produce these errors in the first place are covered in why client billing breaks at scale.
An error your client finds first is an error you paid full price for.
Frequently asked questions
What is invoice accuracy?
Invoice accuracy is the proportion of invoices that go out correct, requiring no correction, credit or adjustment after issue. The term is used for two different processes. On the payables side it means checking that supplier invoices you receive match what was agreed. On the receivables side it means the invoices you send your clients are right the first time. The mechanics are different, the owners are different, and the money moves in opposite directions.
How do you calculate an invoice accuracy rate?
Divide the number of invoices that required no correction after issue by the total number of invoices sent, then multiply by 100. The measurement problem is what counts as a correction. If you only count invoices the client disputed, you are measuring what clients caught rather than what was wrong, and you will exclude every under-billed invoice by construction. A rate built from credit memos alone reports the over-billing half of the population.
How do you check a client invoice for accuracy before it goes out?
Compare it against the contract, not against the last invoice. Most billing errors survive review because the check is a reasonableness check: the number looks like last month's number, so it passes. The contract is where the tier thresholds, minimums, true-ups and version dates live, and it is the only reference that catches an invoice that is consistently wrong. The check that works is expected versus actual, where expected is reconstructed from the contract terms and compared line by line against what the billing system produced.
What is a red flag on a client invoice?
Any line you cannot trace to the events that produced it. Others worth flagging: a total that is round when the underlying volumes are not, a tier applied that the period's volume does not support, a minimum charged in a period that exceeded it, a rate that does not match the current contract version, and a period with no true-up in a contract that requires one. A line nobody can explain without rerunning the billing system is the one to hold.
What does an invoice dispute cost to resolve?
Price it as hours at a fully loaded rate rather than at salary. US Bureau of Labor Statistics data for March 2026 shows wages accounting for 69.9 percent of employer compensation costs in private industry, so the loaded cost of an hour is roughly 1.43 times the salary rate. Six hours of resolution work at a $60 salary rate costs about $515, not $360. Add the carrying cost of the full invoice value for however long the client holds it.
Does invoice accuracy affect DSO?
Yes, and by more than the disputed amount suggests, because clients generally hold the entire invoice rather than paying the undisputed portion. Days sales outstanding is average accounts receivable divided by net revenue, multiplied by 365, so every day a disputed invoice sits unpaid carries the full invoice value through the calculation. The longer-term effect is larger: a client who has found one error adds a verification step to every invoice that follows, which lengthens collection permanently rather than once.
Can this be verified automatically?
The comparison is deterministic, so yes. Reconstructing what a contract says an invoice should total, and comparing that against what the billing system produced, is arithmetic with a source on both sides. Bluefyn does this for the provider side of the ledger, verifying that providers charge exactly what they agreed to charge, transaction by transaction. Bluefyn never moves, holds, or custodies funds; it only analyses transaction and provider data. The same expected-versus-actual discipline applies to what you bill your clients.
