A 99% order accuracy rate is the standard line in a fulfillment quote. The denominator almost never comes with it, and the denominator is the whole number. Ninety-nine percent of orders, of order lines, or of units shipped are three different measurements, and the same warehouse scores differently under each one. This post shows you how each is calculated, which one your customers actually feel, what the published benchmarks say, what a wrong order costs you in dollars, and the six questions that get a straight answer out of any provider, including us.
Why a 99% order accuracy rate is an incomplete number
Accuracy is a fraction. Everyone publishes the top half and leaves out the bottom half.
There are three bottom halves in common use, and a warehouse can report any of them while calling the result “order accuracy.”
Order level. An order is correct only if every single thing in it is correct. One wrong item in a six item box fails the entire order. This is the harshest denominator and the only one that matches what your customer experiences. Nobody opens a box and feels 83% satisfied.
Line level. Each line on the packing slip is scored on its own. Five correct lines and one wrong line counts as five wins and one loss, not a failed order.
Unit level. Each individual unit is scored. A six item order that goes out with one wrong unit scores 83% on that order and barely moves the monthly number.
Here is one month of shipping from the same operation, scored three ways.
| Measure | Denominator | Defects counted | Reported accuracy |
|---|---|---|---|
| Order level | 10,000 orders shipped | 60 orders with at least one defect | 99.40% |
| Line level | 25,000 order lines | 72 incorrect lines | 99.71% |
| Unit level | 40,000 units shipped | 95 incorrect units | 99.76% |
Same month. Same 60 customers who got something wrong. Three numbers, 36 basis points apart. A provider quoting you the unit figure is not lying to you. It is choosing, and you were not told there was a choice.
The spread matters more as your orders get bigger. If you ship single item orders of one SKU, all three measures are mathematically identical and the distinction is academic. If you ship a six item subscription assortment that changes every month, the order level number can sit a full point below the unit level number, and the order level number is the one your inbox will reflect.
When we quote pick and pack fulfillment for a brand with multi item orders, this is the first thing we establish, because a brand comparing our order level figure against a competitor’s unit level figure is comparing two different measurements and does not know it.
What counts as an error, and how errors actually get caught
The denominator is the first gap. The defect list is the second.
Most warehouses count these as errors:
- Wrong item shipped
- Wrong quantity shipped
- Item missing from the order
- Damage caused during packing
Most warehouses do not count these, and will not volunteer that they do not:
- A missing insert, gift note or promotional sample
- The wrong carrier service level, so a two day order ships ground
- Carrier loss or damage after the package leaves the dock
- A bad address that came through from your store, not from the label
- A short ship because the unit was not on the shelf, which gets filed as an inventory problem rather than a pick problem
None of those exclusions are unreasonable on their own. The problem is that a 99.4% built on a four item defect list and a 99.4% built on a nine item defect list are not the same 99.4%, and the shorter list always produces the better number.
How the error gets into the count
There are three detection methods, and they are not equally honest.
- Customer complaint. Somebody emails support, support logs it, the warehouse gets a ticket.
- Internal audit. A sample of packed orders is pulled and re checked before shipping.
- Scan data at pack. Every unit is scanned against the order before the box is sealed, and the system refuses to close an order that does not match.
If a provider’s accuracy number comes only from method one, it is not measuring accuracy. It is measuring complaint rate, and complaint rate understates errors every time. A customer who receives an extra unit keeps it quietly. A customer who gets the wrong shade sends it back through your normal returns flow and ticks “not as described” or “changed my mind,” and the mispick disappears into ordinary return volume.
That volume is substantial. The National Retail Federation and Happy Returns put the 2025 US ecommerce return rate at 19.3% of online sales. A wrong item has a lot of cover to hide in.
The only way to pull those errors back out is to read return reason codes against the original pick record, which is work somebody has to decide to do. On our floor it sits inside returns handling rather than inside the fulfillment report, because that is where the evidence physically lands. Ask your provider whether anyone does that reconciliation, or whether a mispick that comes back as a normal return simply never enters the accuracy number.
What the published benchmark actually says
The closest thing this industry has to a neutral reference point is the DC Measures study run by the Warehousing Education and Research Council, a trade body rather than a vendor. It surveys distribution centre operators and publishes performance in quintiles.
Two things make it useful here. It measures order picking accuracy as a percent by order, which is the harshest of the three denominators. And it publishes the whole distribution, not a single flattering figure.
From the 2022 DC Measures Snapshot in Time, based on 240 individual responses:
| Band | Order picking accuracy, percent by order |
|---|---|
| Major opportunity, bottom 20% | Below 97% |
| Disadvantage, 20th to 40th percentile | 97% to below 99% |
| Typical, 40th to 60th percentile | 99% to below 99.5% |
| Advantage, 60th to 80th percentile | 99.5% to below 99.85% |
| Best in class, top 20% | 99.85% and above |
| Median of all respondents | 99.20% |
Read that table against the sales conversation you have been having.
A 99% order level accuracy rate sits at the bottom edge of the typical band and below the 99.20% median. 3PL marketing treats 99% as a selling point. The survey treats it as average. Both can be true at once, which is exactly why the denominator question is not pedantry.
The 2025 edition moved the best in class threshold to 99.68% or better, according to Hyster-Yale’s published summary of the top twelve metrics. The full 2025 quintile tables sit behind WERC’s paid report, so the 2022 bands above are what we can show you in full.
Two caveats before you take this into a vendor meeting. WERC’s respondents are distribution centres across retail, wholesale, manufacturing and third party warehousing, not a pure sample of ecommerce 3PLs, and the figures are self reported. Use the table as a reference point for what good looks like, not as a contract term.
What one wrong order costs you
Work out your own cost per error before you negotiate anything about accuracy. Without it you cannot tell whether a provider’s accuracy programme is worth paying for.
The hard costs stack up like this:
- The replacement unit, at your cost of goods
- Outbound postage, paid a second time
- A return label for the wrong item, if you pay for returns
- Warehouse labour to receive, inspect and restock the returned unit
- Support labour across two or three emails
- The returned unit itself, which often comes back opened and is no longer first quality
Put placeholder numbers in to see the shape. A $12 cost of goods, a $7.40 ground label out, a $7.40 return label back, six minutes of warehouse labour and twelve minutes of support lands somewhere near $30 per error. Those figures are arithmetic to show the method. They are not our data and they are not industry data. Replace every one of them with yours.
Now scale it against the accuracy table from earlier.
At 10,000 orders a month and 99.40% order level accuracy, you are eating 60 errors a month, 720 a year, roughly $21,600. At 99.90% you are eating 10 a month, 120 a year, roughly $3,600. The gap is about $18,000 a year.
The point where paying for accuracy stops paying
That $18,000 gap is the entire budget available for an accuracy upgrade. Spread across 120,000 orders a year it is $0.15 per order.
The formula is short enough to do in your head:
\text{max justifiable fee per order} = \frac{\text{percentage points of error removed}}{100} \times \text{cost per error}| Your cost per error | Removes 0.10 points | Removes 0.25 points | Removes 0.50 points |
|---|---|---|---|
| $20 | $0.02 | $0.05 | $0.10 |
| $30 | $0.03 | $0.08 | $0.15 |
| $45 | $0.05 | $0.11 | $0.23 |
| $60 | $0.06 | $0.15 | $0.30 |
Read the columns before the rows. Half a percentage point of error is a large amount of error to have available. If you are already running at 99.4%, there are only 0.6 points in the building, and no verification step removes all of them.
That is the answer to why the last 1% costs the most. Moving from 97% to 99% is process discipline and barcode scanning at the pick face, and it pays for itself easily because there is so much error to remove. Moving from 99.5% to 99.9% means a second verification touch on every order you ship, paid for on every order, including the 99.5% that were already correct. You are buying insurance against four orders in a thousand and paying for all one thousand.
This is the same arithmetic that makes per order add ons expensive in aggregate, which is worth reading alongside the charges that show up after you sign. A fee of fifteen cents sounds like nothing until you multiply it by your annual volume and compare it against the problem it solves.
When accuracy should not be your deciding factor
If you ship single unit orders of one SKU, stop optimising for this. All three measures collapse into the same number, the picker has one decision to make, and almost any competent operation clears 99.5%. Accuracy is not where providers differ for you. Choose on cutoff time, transit days to your customers, storage rate, or whether a person picks up the phone. Paying a premium for a verification programme on a one item order is paying to solve a problem you do not have.
If you ship variable multi item boxes, subscription assortments, apparel with size and colour variants, or bundles whose contents change monthly, accuracy is exactly where you should push. Push on us too.
There is also a case where a national competitor genuinely beats us on this. At high volume with stable SKUs, a heavily automated network running goods to person picking and scan to light pack verification can hold a tighter accuracy band than a manual operation, ours included. If you are shipping 50,000 orders a month of the same forty SKUs, that is a real advantage and you should weigh it honestly. Our argument has never been that we out automate ShipBob. It is that you can reach the person who runs the floor, which is a different thing and worth less to some brands than to others.
One more thing you should know before you take any accuracy figure seriously, including ours.
Nobody audits 3PL accuracy rates. There is no certification body, no required disclosure, no standard definition you can hold a provider to. Every number you see in this industry, on our site and on every competitor’s site, is self reported by the company publishing it. Treat all of them as claims until the provider shows you the calculation and lets you check it against your own order data for your own account.
That is the part of this post that is bad for us commercially. We would rather you knew it, because the alternative is you discovering it in month four. If that kind of plain dealing is what you are looking for, it is roughly what being an owner operated 3PL is supposed to mean.
Six questions that get you a real answer
Send these to your current provider and to anyone you are evaluating. The answers take five minutes to give if the operation tracks accuracy properly, and a week to give if it does not. The delay is itself an answer.
- Is that figure order level, line level or unit level? If the answer is “order accuracy,” that is not a denominator. Ask again.
- What window does it cover, and does it include October through December? A twelve month figure that quietly excludes peak is a different number from one that includes it. Peak is when accuracy breaks.
- What counts as a defect? Ask for the list, in writing. Compare it against the two lists earlier in this post.
- How does an error get detected? Customer complaint, internal audit sample, or scan data at pack. If it is complaints only, the figure is a floor on your errors, not a measure of them.
- Is that number facility wide or for accounts like mine? A facility shipping mostly single unit orders will carry a strong blended figure that tells you nothing about how it handles your eight item kits.
- Who pays to reship, and is that written into the agreement? This is the question that separates a marketing number from a commitment. A provider that owns the cost of its own mistakes has a financial reason to keep the number honest.
How we answer them
We publish 99%+ order accuracy, and we count it at order level, which is the harshest of the three. Every order is scan verified against the order before it leaves the facility. When we get one wrong, we pay for the reship.
Measured against the WERC bands above, 99%+ puts us in the typical range, not best in class. We could pick a softer denominator and publish a prettier number. We would rather you compared us like for like, because a brand that arrives expecting 99.9% and gets 99.4% is a brand that leaves in six months, and switching a 3PL is expensive for both of us.
Our figure is self reported and not independently audited. So is everyone’s. Ask us question four and we will tell you exactly how it is detected.
If your current provider will not answer these six, that in itself is one of the signs a brand has outgrown its 3PL, and it tends to show up before the operational problems do.
Order accuracy FAQ
What is a good order accuracy rate for a 3PL?
Measured at order level, WERC’s 2022 DC Measures survey put the median at 99.20%, with 99.5% and above in the top 40% of respondents and 99.85% and above in the top 20%. Anything below 97% was in the bottom quintile. Those bands only mean something if the provider is measuring by order rather than by unit.
Is order accuracy measured per order or per item?
Both are used, and they produce different numbers from identical performance. Order level treats any defect as a failed order. Unit level scores each unit separately and always reports higher on multi item orders. Ask which one you are being quoted.
Does an order accuracy rate include shipping damage?
Usually not. Most accuracy figures stop at the dock and exclude carrier loss and transit damage, because the warehouse does not control them. That is defensible, but it means your customer facing problem rate is higher than the accuracy rate you were quoted.
How do I verify my 3PL’s accuracy number?
No third party audits 3PL accuracy, so verification is on you. Ask for the denominator, the defect list, the detection method and the time window, then compare the claimed error count against your own support tickets and return reason codes for the same period. A large gap means errors are being detected by complaint only.
What does 6G Logistic’s 99%+ order accuracy rate mean?
It is measured at order level across orders shipped, which is the strictest of the three common denominators. Every order is scan verified against the order before it leaves the facility, and when we ship the wrong thing we pay for the reship. The figure is self reported, as every 3PL’s is.
What to do next
Take the six questions to whoever ships your orders today. If the answers come back vague, send us a recent invoice and a month of your order data, and we will tell you what the same volume would cost here and how we would measure it. Quotes come back in 24 hours.
Talk to us about your fulfillment, or call (949) 468-7759.