BENCHMARKS · VARIANCE BY REASON
Vendor invoice variance, reason by reason.
When a carrier, co-loader, trucker or agent invoice doesn't match the job file, the useful question is not only how often — it's why. This page sorts the citable figures by the variance reasons the desk names: duplicate charges, detention and demurrage, buy-rate differences and tax-line defects. Sources named on every row, vendor claims labelled as vendor claims, the gaps left visible, and one famous statistic marked do-not-cite.
SOURCES: FMC · ARDENT PARTNERS · APQC · SAP CONCUR · UPDATED JULY 2026
The defensible band for freight invoices carrying a billing error is 3–8% (industry estimates); vendor figures of 20%+ are marketing. Split by reason, two figures can be cited: duplicates run at about 1.29% of invoices (SAP Concur, secondhand), and roughly 22% of the detention and demurrage nine major carriers billed in 2020–2022 was never collected — waived or successfully disputed (FMC-derived). Buy-rate and tax-line variances have no citable public figure, so this page publishes none.
| Variance reason | What it looks like on a forwarder's desk | Best citable figure |
|---|---|---|
| Buy-rate variance | The invoice line doesn't match the buy rate on the job file | None we can source — no row published |
| Duplicate charge | The same charge on two documents, often weeks apart | 1.29% of invoices are duplicates (SAP Concur, secondhand) |
| Detention & demurrage | Free time, clock start and the event history in dispute | ~22% of billed D&D never collected (FMC-derived) |
| Tax-line defects | Wrong GST treatment or missing GSTIN; VAT/TRN defects | None we can source — no row published |
REASON CODES AS THE DESK NAMES THEM · EACH FIGURE DETAILED, WITH ITS SOURCE, IN THE SECTIONS BELOW
How often a vendor invoice carries a variance
Before splitting variance by reason, the headline rate: how many freight invoices contain a billing error at all. This is where marketing goes to exaggerate. Here is the honest band, the inflated band, and the one number you should never repeat.
| Figure | What it measures | Source |
|---|---|---|
| 3–8% | defensible band for freight invoices containing billing errors | Industry estimates (Shipware, ICC at the low band) |
| 20%+ | upper figures circulated in the market | Vendor claims — label them as vendor claims when citing |
| 80% | “of carrier invoices include discrepancies” — widely attributed to FreightWaves | DO NOT CITE — we fetched the cited article; the figure is not in it |
LOW BAND: INDEPENDENT-LEANING ESTIMATES · HIGH BAND: VENDOR MARKETING · 80% ROW: DEBUNKED, KEPT HERE SO YOU DON'T REPEAT IT
Even the bottom of the honest band is expensive. At 3%, a desk processing 500 vendor invoices a month is paying roughly 15 wrong ones — every month, at forwarder margins. If a statistic seems designed to sell software, check the footnote before you borrow it. We did; that is why one row above is struck through.
Reason: the duplicate charge
Duplicate-payment figures circulate widely, and both of the standard citations are secondhand — repeated across the AP industry without a verifiable primary study in public view. We label them exactly as strong as they are.
| Figure | What it measures | Source |
|---|---|---|
| 1.5% | duplicate or erroneous disbursements as a share of annual disbursements | APQC — secondhand citation; primary study not independently verified |
| 1.29% | of invoices are duplicates, on average | SAP Concur — secondhand citation |
| $2,034 | average value of a duplicate invoice | SAP Concur — secondhand citation |
BOTH SOURCES SECONDHAND — WIDELY REPEATED, PRIMARY STUDIES NOT IN PUBLIC VIEW · TREAT AS ORDER-OF-MAGNITUDE
What is not in doubt is the mechanism. In forwarding, duplicates rarely arrive as the same invoice twice — they arrive as the same charge on two documents weeks apart: the THC on the carrier invoice and again on the co-loader’s SOA. Header-level dedupe never sees them. Line-level matching against the job file does.
Reason: detention and demurrage
D&D is the most-disputed variance reason in freight — and the rare one with a primary source: the US Federal Maritime Commission publishes what carriers actually charged. The numbers reward reading twice.
| Figure | What it measures | Source |
|---|---|---|
| $15.4B | D&D charged by nine major carriers, April 2020 – March 2025 | FMC (fmc.gov) — primary source |
| $8.9B | D&D billed by those carriers, 2020–2022 | FMC-derived |
| $6.9B | D&D actually collected over the same 2020–2022 period | FMC-derived |
| ~22% | of billed D&D never collected — waived or successfully disputed | Derived: ($8.9B − $6.9B) ÷ $8.9B |
SOURCE: FMC — THE PRIMARY SOURCE IN THIS CATEGORY · DERIVED ROWS MARKED
The operational headline is the last row: roughly a fifth of billed D&D was never paid, because somebody pushed back. Disputing works — for the forwarders equipped to dispute with evidence rather than indignation.
The FMC’s 2024 D&D billing rule choreographed US disputes: invoices within 30 days of when charges stopped accruing, mandatory invoice contents, a 30-day dispute window. In December 2025 a federal court set aside the rule’s “properly issued invoices” provision (Federal Register 2025-23920) — weakening the procedural lever that let a billed party refuse a defective invoice. Content citing the rule as fully in force is out of date. Dispute leverage now rests on evidence — free time, clock start, the event history behind the box — more than on procedure.
Reasons without a citable figure: rate and tax
Two of the four reasons the desk names have no row on this page. Buy-rate variance — the quote said one number, the invoice says another — and tax-line defects— GST treatment, a missing GSTIN, VAT or TRN errors — are everyday work on a forwarder’s payable desk, but this page carries no figure for either. Until one can be sourced as cleanly as the rows above, the honest entry is a blank, and they are the first two rows the benchmark below is built to fill.
What finding a variance costs
Every variance has to be found before it can be disputed, and the finding costs money per invoice whether a variance turns up or not. The most-cited figures are Ardent Partners’ all-industry AP benchmarks — not freight-specific, which is worth saying out loud, because a forwarder’s invoices arrive messier than most.
| Figure | What it measures | Source |
|---|---|---|
| $12.88 | average all-industry cost to process a single vendor invoice | Ardent Partners — AP Metrics that Matter research |
| $2.78 | best-in-class cost per invoice — the same work, organised differently | Ardent Partners — AP Metrics that Matter research |
| 17.4 days | average invoice cycle time, receipt to approval | Ardent Partners — AP Metrics that Matter research |
| 3.1 days | best-in-class cycle time | Ardent Partners — AP Metrics that Matter research |
| 2–4 hrs/day | staff time spent on manual invoice audit at forwarders | Expedock — vendor-published claim, flagged as such; checked September 2026 |
| 5–15 min | manual matching time per vendor invoice | Vendor-published claims, flagged as such |
ARDENT FIGURES: ALL-INDUSTRY AP BENCHMARK, NOT FREIGHT-SPECIFIC · VENDOR CLAIMS FLAGGED IN-LINE
The gap between average and best-in-class is ×4.6 on cost and ×5.6 on cycle time — and nothing in that gap is exotic. It is ingestion, matching and exception-handling: precisely the work a reconciliation desk automates.
The benchmark we’re building: variance by reason
No independent study of freight-audit recovery rates exists — every “3–8% of freight spend recovered” figure in circulation traces back to a vendor selling the audit. We searched. Until someone neutral publishes one, our benchmark will do the honest version: the share of vendor invoice lines carrying a variance, split by reason, with match rates and dispute win rates from live, anonymized reconciliation desks — published as they accrue, sources named, our own numbers included.
The method, stated before the first number so it can be checked: a variance is an invoice line that doesn’t match the job file’s buy rate or accrual, and it is counted under the reason the desk named when it queued the line for a person — buy rate, duplicate, D&D or tax. Nothing on this page today comes from that dataset: every figure above comes from a third party, or from arithmetic on one, and is labelled as such.
Turn the benchmarks into your number.
The calculator takes your invoice volume and error assumptions and returns what the leak costs annually — using the defensible band, not the marketing one.
Open the calculator