The freight billing error percentage of spend is commonly quoted as 2 to 5 percent, and the conversion is simple: at $3 million of annual freight spend that is $60,000 to $150,000 a year, at $6 million it is $120,000 to $300,000, and at $12 million it is $240,000 to $600,000. That is the arithmetic, and it is the reason the range gets quoted so often.
It is also a range you should handle carefully, because it is softer than it looks. Nobody has published a rigorous, independently sampled measurement of billing errors as a percentage of freight spend. What exists is a set of related figures, measured on different denominators by parties with a commercial interest, that happen to cluster in roughly the same neighborhood.
This post does two things. First, it says exactly where the 2 to 5 percent comes from and what it does and does not support. Second, it converts it into your dollar figure and shows you how much of that figure is realistically recoverable, which is a much smaller number.
Where the 2 to 5 percent actually comes from
Three separate figures get blended into “2 to 5 percent of freight spend,” and they are not measuring the same thing.
The circulating vendor range. The 2 to 5 percent figure is the number this post is named after, and it does not have a study behind it. We went looking for one. The trail runs from a vendor primer to a page published by a competing vendor, which cites nobody at all. We used to carry the range on our own homepage and have taken it down for exactly that reason. The full trace is in nobody has measured the freight invoice error rate.
Recovery benchmarks by mode. A consultant-authored guide from Darrigo Consulting gives recovery benchmarks of 2 to 5 percent of spend for parcel, 3 to 8 percent for LTL, and 1 to 3 percent for truckload. Note three things about it: it is authored by a consultancy that sells audit services, it is recovery rather than error (recovery is what gets credited back, which is always less than total error), and it is split by mode because mode drives most of the variance.
Self-reported invoice inaccuracy. The figure behind “1 in 5 freight bills contains an error” is not a measurement either. Its most traceable origin is the Journal of Commerce’s 2020 Freight Payment Benchmark Study, in which 77 shippers self-reported an average invoice inaccuracy rate of 20.8 percent. No invoices were audited and the study never defined an inaccurate invoice. Set aside the sourcing and it is still a count of invoices rather than a percentage of dollars. A 20 percent invoice error rate and a 3 percent dollar error rate are entirely compatible, because most billing errors are small relative to the invoice they sit on.
So the range is not a measurement of your operation, and it is not a measurement of anyone else’s either. The honest position is that the freight invoice error rate is unmeasured, which is a different claim from low. Use the range to size the question, then measure. The method for that is in how to benchmark your freight invoice error rate.
The dollar conversion at your spend
Run the range against your annual freight spend before you decide whether the topic deserves your time.
| Annual freight spend | At 2 percent | At 3.5 percent | At 5 percent |
|---|---|---|---|
| $1,000,000 | $20,000 | $35,000 | $50,000 |
| $3,000,000 | $60,000 | $105,000 | $150,000 |
| $6,000,000 | $120,000 | $210,000 | $300,000 |
| $12,000,000 | $240,000 | $420,000 | $600,000 |
| $25,000,000 | $500,000 | $875,000 | $1,250,000 |
The spread between the low and high column is the honest uncertainty. At $6 million the answer is somewhere between $120,000 and $300,000, and anyone who narrows that for you without looking at your invoices is guessing.
What the table is genuinely useful for is a threshold decision. Below roughly $1 million of freight spend, even the high end of the range is a number most operations will not staff against. Above $6 million, the low end of the range exceeds the fully loaded cost of a person spending part of their week on it. That is the actual decision the percentage informs.
Cross-checking the range with the per-invoice numbers
Two per-invoice numbers multiply into a percentage of spend, which is a useful sanity check on whether 2 to 5 percent is even plausible. Neither input is a published fact. Both are assumptions, and in your own version of this you should replace both with counts from your own invoices.
Take an illustrative shipper: 400 loads a month, 4,800 loads a year, at an average invoice of $1,250. That is $6,000,000 of annual freight spend.
Apply two assumed per-invoice figures, chosen only to make the arithmetic visible: about 1 bill in 5 contains an error, and the average error on a bad bill is $212.
- Bad bills per year: 4,800 divided by 5 = 960.
- Total error dollars: 960 times $212 = $203,520.
- As a percentage of spend: $203,520 divided by $6,000,000 = 3.4 percent.
That lands inside the 2 to 5 percent range, which confirms nothing about freight and quite a lot about arithmetic: assumptions picked from that neighborhood produce an answer in that neighborhood. What the exercise is genuinely good for is testing other people’s numbers for internal consistency. A 20 percent invoice error rate at a couple of hundred dollars an error does produce a low-single-digit percentage of spend. So if someone quotes you 10 percent of spend, run this same multiplication against their invoice error rate and their average error, and see whether it closes.
Run it on your own numbers too. If your average invoice is $400 rather than $1,250, the same 20 percent error rate and the same $212 average error produce a much higher percentage of spend, because the denominator shrank and the errors did not. Small invoices carry proportionally more error. That is one reason LTL and parcel benchmarks sit above truckload.
Mode mix moves this more than anything else
The Darrigo ranges cited above split by mode for a reason, and the reason is invoice complexity.
| Mode | Cited recovery range | Why it sits there |
|---|---|---|
| Truckload | 1 to 3 percent | Few line items, high invoice value, mostly linehaul plus fuel |
| LTL | 3 to 8 percent | Classification, reweigh, discount sequencing, dense accessorial stack |
| Parcel | 2 to 5 percent | Very high volume of small surcharges, low value per shipment |
Those are the source’s recovery figures, attributed, not our measurement. But the ordering is what to take from it. A truckload invoice has three or four things that can be wrong. An LTL invoice has a class, a weight, a discount, a fuel percentage and often four or five accessorials, each an independent chance for a variance. A drayage invoice has more line items than either.
So a shipper that is 90 percent truckload should expect to sit at the bottom of the range, and a shipper with heavy LTL should expect to sit above it. If your mix is mostly LTL and your measured error rate is 1 percent, the likely explanation is that you are not catching classification and accessorial errors, not that your carriers are unusually accurate.
What the errors actually are
The percentage is abstract until you know which line items produce it. In practice the dollars concentrate in a short list.
Duplicate invoices. The highest dollar value per instance by a wide margin, because the error is the entire invoice rather than one line. They survive AP because the second submission carries a different invoice number, which is why catching duplicate freight invoices across rebills and corrected bills needs a dedicated check rather than a glance at the invoice number.
Fuel surcharge computed off the wrong index week or peg. High frequency, mechanical, and individually small. This one is fully recomputable from published data, which makes it the easiest category to prove, using the method in how to calculate a truckload fuel surcharge from the DOE index. The weekly on-highway diesel numbers come from the EIA’s On-Highway Diesel Fuel Price Survey, published Tuesday mornings from a Monday 8:00 a.m. capture, so both parties are working from the same public series.
Accessorials that were never authorized. The highest count. Individually $50 to $150, which is exactly the size that makes them not worth chasing one at a time and very much worth chasing as a pattern.
Detention billed without supporting timestamps. Mid-size and highly disputable, because the evidence standard is documentary rather than arithmetic.
Linehaul rate variances. Rare and large. Usually a rate loaded incorrectly on the carrier’s side after a renewal, which then repeats on every load in that lane until someone notices.
The mix matters for what you do next. If your dollars sit in fuel and duplicates, the fix is a recurring mechanical check. If they sit in accessorials and detention, the fix is contract language and documentation discipline at the dock.
The recoverable number is smaller than the loss number
This is the part that gets skipped, and skipping it is how audit programs end up over-promised and under-delivered.
Three deductions sit between the error percentage and the cash you get back.
Findings below your dispute threshold. Most operations will not file on a $22 variance, and they are right not to. Whatever your threshold is, the dollars beneath it are a real loss that you have consciously chosen not to pursue. Disclose them rather than pretending they were never there.
Denials. Some findings are wrong, some are right and lose anyway. A program with a zero percent denial rate is not measuring itself honestly.
Expired windows. Under 49 U.S.C. 13710, you must contest a bill within 180 days of receiving it to preserve your right to challenge it, and a carrier billing additional charges beyond what it originally billed must do so within 180 days of your receipt of the original bill to preserve its collection rights. If you have already paid, the process is an overcharge claim instead, and under 49 CFR 378.8 the processing carrier must pay, decline or settle a written overcharge claim within 60 days of receipt absent a written agreement to extend. Errors found on invoices that aged past the contest window are a measurement of what your process cost you, not a pipeline. The complete set is in every freight billing deadline that can cost you money.
Be careful about what these rules do. They govern deadlines and claim handling. They do not decide whether a given accessorial was owed, which is almost always a question of what your contract and the carrier’s rules tariff say. Overstating the regulatory position is the single fastest way to lose credibility with a carrier’s billing department.
After those three deductions, a realistic recovery is a meaningful fraction of identified error, not all of it. Which is exactly why the reporting structure in tracking freight recovery savings in a ledger finance will accept separates identified, agreed and realized into three columns.
Turning the percentage into a decision
The percentage only earns its keep if it changes something. Two decisions it should drive.
Whether to review at all, and how. Compare the low end of your dollar range against the fully loaded cost of reviewing. If your range is $120,000 to $300,000 and reviewing costs a fraction of the low end, the decision is made and the only remaining question is which review model. Those models and their costs are compared in what freight audit actually costs across three pricing models.
Where to point the review. Do not audit everything evenly. Spend concentrates: a small number of carriers and a small number of charge types will produce most of your error dollars. Start with your two highest-volume carriers and the four charge categories above, and expand only when that stops producing.
One warning about first measurements. A first audit pass reliably runs hotter than the steady-state rate, because it clears an accumulated backlog: a fuel peg that has been wrong for eight months, an accessorial schedule that never got loaded, a rate that has been off since renewal. Those are real dollars and worth recovering, but do not annualize the first pass. The second quarter is your baseline.
The checklist
- Multiply your annual freight spend by 2 percent and by 5 percent. That is the size of the question, not the answer.
- Cross-check it: your invoice count times your error rate times your average error should land in the same neighborhood. If it does not, one of the inputs is wrong.
- Adjust for mode mix. Heavy LTL sits high, heavy truckload sits low.
- Measure your own rate on at least 100 invoices before quoting any number internally.
- Deduct sub-threshold findings, expected denials and expired windows before promising a recovery figure.
- Check the 180-day contest window on anything you plan to dispute.
- Report the first pass and the steady state separately.
The 2 to 5 percent range is worth exactly one thing, and it is not accuracy: it tells you whether this is a rounding error or a budget line at your spend. For most operations above a few million dollars of freight, it is a budget line. Then you stop using the industry number and start using yours.
Sources
- Freight and shipping audit guide, recovery benchmarks by mode (Darrigo Consulting, consultant-authored, attribute when citing)
- EIA On-Highway Diesel Fuel Price Survey methodology (U.S. Energy Information Administration)
- 49 U.S.C. 13710, billing and contesting windows (Cornell LII)
- 49 CFR 378.8, overcharge claim resolution within 60 days (Cornell LII)
- 49 CFR 378.2, definitions of overcharge, duplicate payment and overcollection (Cornell LII)
- Nobody has measured the freight invoice error rate (CheckMyFreightBill source audit: the 2 to 5 percent of spend range traces to a circular vendor citation, and the “1 in 5 bills” figure traces to 77 shippers self-reporting 20.8 percent in a 2020 Journal of Commerce study that never defined an inaccurate invoice)