Order Accuracy and Fulfillment Error Rates: What to Expect from a 3PL

TL;DR

A well-run eCommerce operation should reach about 99.5% order accuracy or better, measured per order. Before you compare providers, ask what each figure measures: picks, order lines or whole orders, and over what period. A figure per order line can look higher than a figure per order, so ask to see the breakdown, not only the headline number. Our own pick accuracy at Gonini is 99.9%, measured across every order, and we can give any brand the breakdown by order line for any period.

Key Takeaways

Accuracy figures are easy to quote and hard to compare, and the points below summarise what sits behind them.

• The same performance can read as 99.0% by order or 99.7% by unit, so always ask which denominator a provider uses
• A rate built only from customer complaints measures complaints, since many errors never get reported
• Published benchmarks such as WERC's DC Measures and technology figures from consultancies and vendors give reference points, though none of them are audited 3PL data
• Accuracy often falls during peak as temporary staff arrive, so an annual average can hide a November problem
• Providers self-report accuracy figures, and clients find them hard to verify independently. Our pick accuracy is 99.9%, measured across every order, and we scan a barcode for each pick

The section on what to do when accuracy slips sets out the evidence to gather before you escalate.

Order Accuracy and Fulfillment Error Rates from a 3PL

A 99% order accuracy rate sounds excellent until you work out what it means at your volume. At 20,000 orders a month, 99% means 200 customers a month receive the wrong item, the wrong quantity or an incomplete parcel. Each of those orders can trigger a reshipment, a refund, a support ticket and possibly a poor review. This guide is for eCommerce brands using or choosing a 3PL who want to understand what an accuracy figure actually measures and what to expect.

Accuracy is one metric among several, and our guide to the KPIs worth tracking with a provider covers the wider set. This guide goes deeper on accuracy alone: how providers measure it, what published benchmarks say, where errors come from and how warehouses reduce them. Separate Gonini guides cover the contract side, such as service level agreements, and the process of moving to a new provider.

If you are still choosing a provider, our guide to evaluating a fulfillment partner covers the full selection process. The questions in this guide fit inside that process, since a provider's answer to "how do you measure accuracy?" says a lot about its operation.

A word on our position before going further. Gonini is a 3PL publishing a guide on how to hold 3PLs to account, so we have an obvious interest in how this topic is framed. Accuracy figures are self-reported and hard for clients to verify, so we give ours in the terms this guide asks you to check: our pick accuracy is 99.9%, measured across every order, and we scan a barcode for each pick. Benchmarks in this guide come from named sources checked on 15 October 2026, or carry an illustrative label.

Quick Summary: The Accuracy Metrics

Several different metrics get called accuracy, and they can produce very different numbers from the same performance. The table compares five common measures by what each one measures, how providers calculate it, and how it can make performance look better than it is. Use it as a checklist when a provider quotes an accuracy figure.

‍

Order accuracy is usually the strictest of the three picking measures, because one wrong unit makes the whole order count as an error. OTIF and perfect order rate widen the view to delivery and condition, and stock availability measures such as how fill rate is calculated sit alongside them. Whichever metric a provider quotes, ask for the formula and the source of the error count.

How We Sourced the Benchmarks

Every benchmark in this guide falls into one of three groups: published survey data with a named source, vendor-reported figures we label as such, or illustrative numbers built to show the arithmetic. We checked each source on 15 October 2026. Our guide to logistics KPIs explained covers how these measures fit into wider performance reporting.

The most cited survey source is WERC, the US Warehousing Education and Research Council, whose annual DC Measures study ranks distribution centres by quintile. Its full reports sit behind a paywall, so we rely on published summaries and say so where we do. Technology accuracy ranges come from equipment vendors and consultants, who have an interest in the results, so we treat them as indicative rather than independent.

Some figures circulate widely without any traceable source, such as a fixed cost per mis-pick quoted as an industry average. Where we found figures like that, we left them out rather than repeating them. The firm figures here are the ones with a named source, and everything else is indicative or illustrative.

How Accuracy Is Actually Measured

Accuracy depends on three choices: the unit of measurement, which errors count, and who decides when an error happened. Change any one of them and the same operation produces a different number. The sections below take each choice in turn.

Per Order, Per Line, Per Unit, and Why the Denominator Matters

The denominator decides how big an error looks. Take a month of 1,000 orders containing 2,500 order lines and 3,500 units, with 10 errors, each one a wrong unit on a separate order. Measured per order, accuracy is 99.0%, since 990 of 1,000 orders shipped correctly. Measured per line, it is 99.6%, and measured per unit, it rises to about 99.71%.

Nothing about the operation changed between those three figures. Only the denominator moved, which explains why two providers quoting 99.7% can be performing very differently. A provider measuring by unit will almost always report a higher figure than one measuring by order, and neither is wrong as long as both say which they use.

For an eCommerce brand, per-order accuracy usually comes closest to the customer's experience, since a customer with one wrong item in a five-item order still has a problem. Ask providers to report accuracy per order, or to report all three measures side by side. If they only report units, convert the figure using your own average lines and units per order before comparing.

Perfect Order Rate and What It Bundles Together

Perfect order rate combines several measures into one. A common version multiplies the on-time, in-full, damage-free and accurate-documentation percentages together, so weakness in any one area pulls the whole figure down. For example, four respectable-looking rates of 96%, 97%, 99% and 94% multiply to a perfect order rate of 86.66%, which means about thirteen orders in every hundred fail somewhere.

The multiplication method has a known flaw. Where failures cluster in the same orders, multiplying component rates can count one bad order as two failures, which produces a perfect order rate that is lower than the truth. Counting perfect orders directly is the more precise method when the data allows it, so ask which method a provider uses before comparing figures.

Perfect order rate also mixes responsibilities. Under many 3PL contracts, the warehouse answers for packing and documentation, while the carrier answers for on-time delivery and damage. That split means a poor perfect order rate can reflect a carrier problem rather than a warehouse one, so ask for the component rates as well as the total.

Who Counts an Error, and When

The source of the error count matters as much as the formula. Some providers count errors caught internally, such as at a packing check, while others count only errors that customers report. If a provider counts errors only when a customer complains, its measured rate is really a complaints rate, since many customers who receive the wrong item never report it.

Industry benchmarks use their own definitions too. WERC's order picking accuracy measure tracks picking accuracy against errors caught before shipment, such as during packing, which captures internal catches rather than customer complaints. A provider comparing its complaint-based rate with that benchmark is comparing two different things.

Ask any provider three questions: who records errors, at what stage, and whether customer-reported errors get checked and added to the count. The most useful accuracy figure combines internal catches with verified customer reports, so neither source hides the other.

What Good Looks Like

Accuracy benchmarks exist, and they come with caveats about who measured what. The sections below set out the published ranges, explain how volume and SKU complexity shift expectations, and look at what accuracy costs to achieve.

Benchmark Ranges and Their Sources

WERC's DC Measures study defines best in class as performance within the top 20% of all respondents. RF-SMART, a warehouse software vendor, summarised the 2020 study and put best-in-class order picking accuracy at 99.89% or higher, with anything below 98% flagged as an area to improve. Those figures come from a 2020 summary, so check WERC's current edition before relying on them.

Consultancies publish ranges by picking technology. MWPVL International, a supply chain consultancy, puts voice picking at 0.3 to 3 gross errors per 1,000 units picked, equal to about 99.7% to 99.97% pick accuracy. It gives 99.5% to 99.7% for pick to light and 99.3% to 99.5% for RF scanning. These are pick accuracy figures, counted per unit picked, and they are not independent audits of 3PL order accuracy.

Context changes what counts as good. Cisco-Eagle, a US materials handling supplier, argues that picking a $2,000 laptop correctly matters more than picking a $2 packet of sweets, and that the same applies in regulated sectors such as pharmaceuticals and jewellery. Our reading of these sources is that well-run eCommerce operations should aim for 99.5% order accuracy or better, measured per order, with the strongest performers higher still.

How Volume and SKU Complexity Shift Expectations

The same provider can deliver different accuracy for different clients. A brand with 50 SKUs that look nothing alike gives pickers little chance to confuse items, while a brand with 2,000 SKUs in near-identical packaging, sizes and colours creates constant opportunities for mistakes. Apparel, cosmetics in different shades and spare parts all fall into the harder group.

Order profile matters too. Single-item orders give one chance for error per order, while ten-item orders give ten, so per-order accuracy naturally falls as basket size grows. When you compare a provider's figure with a benchmark, ask what SKU count and order profile sit behind it.

More orders and more SKUs mean more chances to pick the wrong item, but that risk is not fixed. If you work with your 3PL on how your orders are handled, you can design some of it out. 

For example, a pet supplement brand we work with sells monthly subscriptions. We know the number of orders and the recipients in advance, so we pack every box before the orders arrive. When the orders come in, we only print and attach the shipping labels. Every subscription order goes out on the same day, at a lower cost than picking each order when it arrives, and there is very little picking left to get wrong. Ask your provider which of your orders are predictable enough to prepare in advance.

What Accuracy Costs to Achieve

Accuracy costs money, and each improvement usually costs more than the last. Moving from paper picking to scanning removes a large share of errors at modest cost, while chasing the final fraction of a percent can mean double-checking every order. The right level depends on what an error costs you.

The example below is illustrative. Suppose a pack-station scan adds five seconds per order at a loaded labour cost of £18 per hour, which comes to about £0.025 per order, or £25 per 1,000 orders. If it cuts errors from 10 to 3 per 1,000 orders, and each error costs £8 in reshipping, product and handling, it saves £56 for every £25 spent. A full 30-second second check on every order would add about £150 per 1,000 orders to catch one or two more errors, which would cost more than the errors themselves.

Run the same sums with your own error cost, which rises with product value, postage and the customer lifetime value at risk. Brands selling expensive or regulated products justify more checking than brands selling low-cost consumables.

Where Errors Come From

Errors start in four places: picking, goods-in, packing and systems. Each produces a different kind of mistake, and fixing one does little for the others. Errors that reach customers also lead to returns, and our guide to returns and what they cost covers the downstream impact.

Picking Errors

Picking errors are the most familiar kind: the wrong item, the wrong size or colour, or the wrong quantity. They usually come from similar-looking products stored near each other, unclear location labels, or pickers working from paper lists with no scan to confirm. Fatigue and time pressure make all of them more likely.

Quantity errors deserve separate attention. A picker asked for three units may pick two or four, especially with small items picked from loose bins. Scanning each unit, rather than scanning once and keying in a quantity, reduces this kind of error at the cost of a few seconds per line.

Goods-In and Putaway Errors

Some errors start long before picking. If goods-in records the wrong SKU, the wrong quantity or the wrong location, every later pick from that stock inherits the mistake. A picker can follow the system perfectly and still ship the wrong item, because the item in the location never matched the record.

Putaway errors work the same way. Stock placed in the wrong bin, or mixed with a similar SKU in the same location, creates errors that look like picking mistakes. Regular cycle counts catch these early, and our guide to running a stock audit covers how to structure them.

Packing and Labelling Errors

Packing errors happen after a correct pick. Two orders packed at the same bench can swap items or labels, an insert can go into the wrong parcel, or a label can print for the wrong address. These errors look identical to picking errors from the customer's side, so root-cause analysis needs to separate them.

Labelling errors also include carrier and service mistakes, such as a next-day order shipped on an economy service. Those orders arrive complete and still fail the customer's expectation. A pack-station scan that checks both the item and the shipping label against the order catches most of these.

At Gonini, every brand we work with has its own written packing procedure, so each order matches the unboxing that the brand designed. Brands can also see their branded packaging stock in our system. When the packing steps for each brand are written down, and the packaging is tracked like any other stock, there is less room for the wrong insert or the wrong box.

System and Integration Errors

Some errors come from data rather than people. A product listed with the wrong barcode, a variant mapped to the wrong SKU, or an order that syncs twice can produce errors no picker could avoid. Warehouse management systems only work as well as the data flowing into them from your sales channels.

These errors often appear after a change, such as a new product launch, a platform migration or a new marketplace connection. Check the first orders after any change closely, and agree with your provider how system errors get attributed, since they often stem from the brand's data rather than the warehouse floor.

How Warehouses Reduce Errors

Warehouses reduce errors through technology, layout and people. Each lever addresses a different cause, and the best results come from combining them. The sections below explain how each works.

Scanning, Verification and Pick-to-Light

Barcode scanning is the foundation. When a picker scans the location and each item, the system can reject a wrong product before it leaves the shelf, and barcodes and product identification explains how that works. A second scan at the packing station confirms the order before the parcel closes.

Pick-to-light and voice picking guide pickers to the right location and quantity without paper. The consultancy figures quoted earlier suggest both can achieve high pick accuracy, though results depend on how well the systems are set up and maintained. Our overview of warehouse technology in use covers these tools and where each one fits.

Weight checks add a further control. A packing station that weighs each parcel against its expected weight can flag missing or extra items, which suits orders with several units of the same product. None of these tools removes errors entirely, and each adds cost that needs to match the errors it prevents.

Slotting and Layout

Slotting decides where each product sits, and good slotting separates look-alike items. Storing two shades of the same lipstick in adjacent bins invites errors, while placing them in different aisles makes a mix-up far less likely. Our look inside a pick and pack warehouse shows how layout shapes the picking process.

Clear location labels, good lighting and one SKU per location also help. Fast-moving products near the packing area reduce travel and fatigue, which supports accuracy as well as speed. Slotting needs regular review as ranges change, since a layout built for last year's bestsellers can create new confusion this year.

Training, Staffing and Peak-Season Dilution

Trained, experienced pickers make fewer mistakes, and stable teams learn a client's products over time. Induction training, regular refreshers and feedback on individual errors all support accuracy. Providers that track errors by picker can target coaching where it helps most.

Peak season tests all of this. Accuracy often falls when temporary staff arrive, since new pickers do not know the products and work under pressure to clear volume. A provider reporting an annual average can hide a November problem behind eleven stronger months, so ask for monthly figures that include the peak months, and ask how the provider supervises temporary staff.

Brands can help too. Clear product images, distinct packaging for similar variants and accurate barcodes all make a picker's job easier, especially for someone in their first week. Sharing forecasts early also lets a provider plan peak staffing rather than scramble.

What to Do When Accuracy Slips

When accuracy slips, start with evidence rather than frustration. Raise the issue through an account manager who can escalate, and bring data that shows the pattern rather than isolated anecdotes. The list below sets out what to gather.

• The error rate over time, by month or week, using the same definition each period
• Individual error cases with order numbers, dates, SKUs and the nature of each error
• Customer photos or complaints that confirm what arrived
• Any pattern by SKU, product type, day of the week or shift
• Changes on your side, such as new products, barcodes or integrations, that coincide with the rise

With that evidence, ask the provider for a root-cause analysis and a written remediation plan with dates. A reasonable remediation period depends on the cause: a mislabelled SKU can be fixed in days, while a staffing or layout problem may take several weeks. Agree how you will measure improvement and when you will review it.

If accuracy does not recover within the agreed period, the issue becomes a contractual matter. Check what your agreement says about accuracy commitments, service credits and termination rights, which Gonini's guide to fulfillment SLAs covers in detail. Some brands set a fixed improvement window with clear targets before deciding whether to leave.

If you do decide to leave, plan the move carefully, since switching providers carries its own risk of errors during the transition. Our guide to moving to a different provider covers the process step by step. Staying and fixing the problem is often cheaper, so leave only when the evidence shows the provider cannot or will not improve.

If You Move to a New Provider

Ask each new provider three questions about accuracy, and compare the answers side by side.

• What do you measure: picks, order lines or whole orders?
• Over what period, and can I see the breakdown, not only the headline figure?
• How do you check each pick? For example, do you scan a barcode for every item?

Then plan the move itself. The first weeks at a new warehouse carry the most risk, because the team is still learning your products, so check accuracy each week until the new site is stable. 

If you can, do not move all your stock at one time. When one brand moved to Gonini from a large 3PL, we moved half of its stock first and set up its Shopify store to send orders to both warehouses at the same time. Orders at our warehouse grew while the stock at the old warehouse ran down. The brand stayed online through the move, and it did not pay to remove all its stock from the old warehouse at one time. Our page for brands switching 3PL explains how we handle the move.

The Limits of Accuracy Figures

Perfect accuracy is not achievable, and any provider claiming 100% is either not measuring properly or not counting the same things as everyone else. People pick products, systems carry imperfect data, and labels occasionally end up on the wrong parcel. A realistic goal is a low, stable and honestly measured error rate.

Chasing the last fraction of a percent also costs more than the errors do for most brands. The cost section above shows how verification steps pay for themselves up to a point and then stop paying. Set a target that reflects your own error cost rather than the highest number a sales deck can quote.

The uncomfortable truth is that providers self-report accuracy figures, and clients find them difficult to verify independently. You can audit samples, track customer complaints and run test orders, and yet you rarely see every error the warehouse catches or misses. 

Our own figures show what a figure looks like when it says what it measures. Gonini reports the following, and each item counts something different.

• Pick accuracy of 99.9%. We scan a barcode for each pick, and our own software records every order, so the figure covers every order rather than a sample
• On-time despatch of 99.5%. An order counts as on time when it arrived before the cut-off time and we picked it, packed it and handed it to the carrier on the same day
• A breakdown by order line, which we can give each brand for any period
• Trend reviews every 6 weeks and every 6 months

Pick accuracy is a different measure from the 99.5% order accuracy benchmark earlier in this guide, and on-time despatch measures speed rather than accuracy. You can see how picking works day to day on our pick and pack service page.

What you can do is make the figures harder to flatter. Agree the definition, the denominator and the error sources in writing, ask for monthly reporting with peak months included, and cross-check against your own customer service data. A provider willing to share that detail usually has less to hide.

FAQ

What is a good order accuracy rate for a 3PL?

A well-run eCommerce operation should reach about 99.5% order accuracy or better, measured per order.

How is order accuracy calculated?

Error-free orders divided by total orders, times 100. Line and unit versions give higher numbers. See our guide to logistics KPIs explained for related measures.

What is the difference between order accuracy and perfect order rate?

Perfect order rate adds on-time delivery, condition and documentation. See what on time in full means for a related metric.

What causes most picking errors?

Look-alike products stored together, unclear labels and unscanned picks. Read how pick and pack operates for the basics.

Does accuracy drop during peak season?

Often, as temporary staff arrive. See planning for peak trading, and ask for monthly figures.

Can I verify my 3PL's accuracy figures independently?

Only partly. Test orders, sample audits and your own complaint data help. Track them monthly alongside the KPIs to track with a 3PL.

‍