Order picking errors: what a mis-pick costs and which checks catch it

A mis-pick looks cheap on the warehouse floor: one wrong item, back to the shelf. It only stays cheap if someone catches it before the box is sealed. Once the customer opens the box, the same mistake costs a support contact, two more labels, a second box and a restock. This article puts numbers on each step and then lines up the shift checks by which errors each one catches. Benchmarks and prices were read on 28 October 2026.

The short answer, from our cost model below: a mis-pick costs about $0.53 if a scan catches it at the shelf, $3.68 if the pack station catches it, and about $30 if the customer finds it. That puts a customer-found mis-pick at more than 50 times the cost of one caught at the shelf. A warehouse shipping 50,000 two-line orders a year at a paper-list error rate ships about 300 wrong orders, which costs about $9,000 a year, before counting any customers who don't come back.

What counts as a mis-pick

Researchers at the Technical University of Munich (TUM), who studied ten manual picking operations (Rammelmeier, Galka and Günthner, 2012), sort picking errors into four types. The type matters because each check below catches some types and misses others:
Wrong item: a different article goes in, instead of or as well as the right one. This includes look-alikes: the same shirt in M instead of L, or the 2 m cable instead of the 1 m one.
Wrong quantity: the right article, but too many or too few units.
Omission: a whole line is skipped. It is really a quantity error with zero units, but it has its own causes, such as a page turned too fast or a line left off the screen.
Wrong condition: the right article, but damaged or wrongly labelled.
At the pack station, two more errors show up: the right item goes into the wrong customer's box, or the box gets the wrong shipping label. Neither is strictly a picking error, but the customer sees the same result: they got the wrong order.

How accurate warehouses are: the WERC benchmarks

The Warehousing Education and Research Council (WERC) surveys distribution centres every year for its DC Measures report. In 2019, order picking accuracy (percent of orders picked without error) was the most widely used metric in the survey. WERC reports each metric in five bands, with the top 20% of respondents as best in class. The last band table published in full is the 2019 one:

Two things to note. First, the 2025 best-in-class threshold (99.68%, as summarised by Yale from the 2025 report) is lower than the 2019 one (99.89%). Thresholds move with whoever answers the survey that year, so don't read the change as warehouses getting worse. Second, WERC defines the metric as accuracy of the picking process, where errors can still be caught and corrected before shipment. A 99.4% picking score with a good pack check can still ship at 99.9%. The rest of this article is about that gap.

The TUM study gives the per-line rate for classic paper-list picking: about 0.3%, or one line in every 300. Orders have more than one line, so the order error rate is higher than the line error rate:

Worked through: order accuracy = (1 − 0.003)^lines. At two lines per order that is 0.997 × 0.997 = 0.99401, or 99.40%, which matches the 2019 WERC median exactly. In other words, a typical two-line operation behaves like paper picking with no checks at all. At ten lines an order, the same picker on the same day scores in WERC's "disadvantage" band.

What one mis-pick costs, by where it is caught

Published per-error figures range from about $22 to $100. They are mostly vendor estimates and none of them say what they count. So we built our own from public rates, and every line can be swapped for your own figure:
Warehouse labour: $22.09 an hour, the May 2025 BLS mean wage for stockers and order fillers in warehousing and storage. Wages were 70.0% of private-industry employer costs in June 2026 (BLS Employer Costs for Employee Compensation), so the loaded cost is $22.09 / 0.70 = $31.56 an hour.
Customer service: $22.40 an hour (May 2025 BLS mean, customer service representatives), loaded $32.00 an hour.
Label: $8.74, USPS Ground Advantage commercial, 1 lb, zone 5, on the price list in force since 26 April 2026.
Our assumptions: minutes per step as listed, $0.75 for a second box and void fill, $12 goods cost for the item when the customer keeps it.

If the customer keeps the wrong item instead of sending it back, the return label and restock ($11.90) are replaced by the $12 goods cost, so the total is $29.98, almost the same. Our ~$30 sits close to the one survey figure that gives a sample: about $30 per mis-pick, from 300 supply chain and distribution managers, as reported by WarehouseOS. It leaves out the hardest cost to price, a customer who doesn't order again. TUM's paper lists "customer loss" at the end of the same chain, after the express re-delivery.

A year of mis-picks: 50,000 orders

Take a small e-commerce warehouse: 200 orders a day, 250 days a year, two lines an order, picking from paper lists at the TUM rate. That gives 50,000 orders and 300 orders with a pick error. What those errors cost depends on how many get through to the customer:

Now add checks. The catch rates in the next table are our assumptions, not measured figures: 85% of errors stopped by a scan at the shelf, and 80% of the rest stopped by a pack-station weight check. Swap in your own from a week of tallies. The point is the shape: checks don't remove errors, they move them to where they cost $0.53 or $3.68 instead of $29.87.

How the last row works out: 300 × 85% = 255 stopped at the shelf (255 × $0.53 = $135). That leaves 45, of which 80% = 36 are caught at packing (36 × $3.68 = $132). The other 9 reach the customer (9 × $29.87 = $269). Total $536, against $8,962 with no checks. Every input is listed above, so the sums take a few minutes to redo with your own wages, labels and tallies.

Which check catches which error

No single check catches everything. A weight check misses look-alike variants. A scan of the first unit misses a short quantity. Nothing that reads a barcode can see a crushed box. Here is how the common checks line up against the error types, based on how each one works and on the TUM paper's notes on scanning and weighing:

Three things follow from the table:
Scan the item, not just the location. A location scan proves the picker stood at the right slot. It doesn't prove the right item was in that slot, because a put-away error months earlier puts the wrong stock behind the right label.
Scan every unit when quantities are more than one. A first-unit scan checks the item but not the count. TUM notes that scanning gives protection against omission, wrong-item and/or quantity errors depending on whether each unit or only the first one is scanned, and that small parts often carry no barcode at all.
Weigh to back up the scan, not to replace it. The companies TUM interviewed said weighing still lets quantity errors through when items vary in manufacture or the scale doesn't suit the job. The next section shows where that limit sits.

Worked example: when a weight check sees a missing unit

A weight check compares the packed box with the weight the system expects (item weights plus box). It only works if every item has a weight in the item master. It flags anything outside a tolerance band, and the band has to cover normal variation, or packers learn to ignore the beep. Here is one order, with our assumed variations:
Order: a phone case (45 g) and a 2 m USB-C cable (60 g) in a mailer (20 g). Expected weight: 125 g.
Scale: reads in 2 g steps; allow two steps: ±4 g.
Mailer and fill: vary by ±3 g.
Items: vary by ±3% of their weight: 3% × 105 g = ±3.2 g.
Tolerance band: 4 + 3 + 3.2 ≈ ±10 g, so anything from 115 g to 135 g passes.

The rule: a weight check only sees a unit heavier than the band, and to be reliable it should be at least twice the band. The band also grows with the order. At 3% item variation, a 2 kg order has a ±60 g band from the items alone, so a 45 g phone case can go missing unnoticed. In practice: pull a list of SKUs lighter than twice your usual band, and make those scan every unit items at packing.

The shift checklist that goes with it

Most of these checks are habits, not hardware. These are the items that keep the cheap catches cheap, sorted by when in the shift they happen. Our order picking shift and pack station check checklists carry most of them, ready to start and send to the team:
Before the shift: scanners charged and logged in; pack scale zeroed and checked with a known test weight; new SKUs since yesterday have a weight and a barcode in the system; look-alike SKUs (sizes, colours, lengths) not in neighbouring slots.
At every pick: scan the location, then the item; scan every unit when the quantity is more than one; report a short pick on the spot instead of "fixing it at packing".
At the pack station: scan the order, then every item; weight check on every box; items under twice the band scanned one by one; photo of the open box before sealing; scan the label against the order.
End of shift: count the pack-station catches by error type (wrong item, quantity, skipped line, wrong tote) and by picker or zone; damaged stock reported; scanners on charge.
Once a week: compare pack-station catches with customer complaints. If customers report error types the pack station never flags, the check has a hole. Use the table above to find it.

A list only helps if people tick it rather than skim it; our piece on what the surgery and ICU checklist studies measured covers why short, specific items work better than long ones.

Sources

Warehousing Education and Research Council (2019). DC Measures 2019, Snapshot in Time (handout), Table 3: quintile performance classification. werc.org handout PDF
Hyster-Yale (2025). Top 12 distribution center metrics, from the 2025 WERC DC Measures report. yale.com PDF
WERC (2025). WERC releases 2025 DC Measures report. werc.org
Rammelmeier, T., Galka, S. & Günthner, W. A. (2012). Fehlervermeidung in der Kommissionierung (Error prevention in manual order picking). Logistics Journal: Proceedings. doi:10.2195/lj_Proc_rammelmeier_de_201210_01
U.S. Bureau of Labor Statistics, Occupational Employment and Wage Statistics, May 2025: 53-7065 Stockers and Order Fillers (warehousing and storage, NAICS 493), 43-4051 Customer Service Representatives. bls.gov/oes
U.S. Bureau of Labor Statistics (9 September 2026). Employer Costs for Employee Compensation, June 2026. bls.gov
Ship.com (2026). USPS Ground Advantage commercial vs retail, price list effective 26 April 2026. ship.com
WarehouseOS. The cost of picking mistakes for small businesses (survey of 300 supply chain and distribution managers). warehouseos.com
Cost model: our own, inputs as listed above, run 28 October 2026.