Blog/Almacén y stock

Order-picking errors: how to avoid them

Order-picking errors: types, causes and how to avoid them with unique SKUs, fixed locations, scanning and double-checking at packing.

Order-picking errors: how to avoid them

Most order-picking errors aren't caused by a careless warehouse operative. They're caused by a process that forces people to interpret: two SKUs with almost identical names, a product stored wherever there was space, a paper pick list, or an "eyeball" check before sealing the box. Asking the team to pay more attention works for a week. Designing the process so the error simply can't get through works every time.

In practice, error-free order picking rests on five pillars: a catalogue with unique SKUs and barcodes, fixed and logical storage locations, scanning rather than reading at the picking stage, a second independent check at the packing station before the label is printed, and a log of every incident so you can tackle the root cause. Below you'll find each step, a worked example with figures, and a checklist you can apply tomorrow.

What a badly prepared order really costs

Sending a size M instead of an L doesn't just cost you a T-shirt. It adds up the original shipment, the return collection, a second shipment, another picking job, customer service time and, often, a discount code to smooth things over. Using rough industry figures (a domestic shipment at €3-4 and picking at €1.5-2), a single error can come to around €10-15 in direct cost, without even counting the product that comes back and may not be resellable.

A rough example: a shop with 800 orders a month and a 1% error rate has 8 incidents a month. At around €12 each, that's nearly €100 a month and over €1,000 a year. With a 3% error rate, that figure triples.

And there's a cost that never shows up on the invoice: the customer doesn't distinguish between a warehouse mistake and a brand mistake. If the first order arrives wrong, it's likely to be the last. At 100 orders a month an error can seem trivial; at 1,000, the same percentage becomes a fixed line in the budget.

Types of order-picking errors and where they come from

Before changing anything, it helps to put a name to what's going wrong. In almost any ecommerce warehouse, errors fall into these categories:

  • Wrong variant: incorrect size, colour, format or fragrance. The most common issue in fashion and cosmetics.
  • Wrong quantity: three units were ordered but two go out, or two go out from a 2-for-1 when it should have been one.
  • Wrong product due to similarity: near-identical packaging stored in neighbouring locations.
  • Shortfall from phantom stock: the shop sold a unit that doesn't physically exist.
  • Mix-ups during consolidation: when several orders are picked at once, an item ends up in another customer's box.
  • Forgotten extras: the insert, sample or promotional gift that depended on someone remembering.
  • Inadequate packaging: an unprotected bottle or a box too large that arrives crushed.
  • Mismatched label: the right parcel with another order's label.

Each type has a different cause and a different fix. That's why treating "errors" as one single issue rarely works.

Step 1: a unique SKU per variant, with its own barcode

Many errors start outside the warehouse, on the product page. The commercial name sells, but the SKU runs operations. If the warehouse has to work out what "Black Basic T-Shirt" means when there are five sizes and two collections, sooner or later it'll get it wrong.

  • A unique SKU per variant, with no duplicates or reused old references.
  • A readable barcode on every unit. If your supplier doesn't provide one, label it on receipt, not later.
  • An operational description with the attributes that distinguish the variant (size, colour, capacity) at the start of the name, not the end.
  • Well-defined packs and kits: decide whether they're a reference with its own stock or assembled from components during picking. Mixing both criteria is a sure route to discrepancies.
  • If you change a product's contents or packaging, update the product page before the new stock arrives.

Step 2: controlled goods-in and a fixed location for every reference

Next week's picking error usually starts today at the loading bay. If a box of size L arrives labelled as M and nobody checks it on receipt, the warehouse will pick it wrong with total precision. At goods-in, count against the delivery note, open a sample of each reference, label anything without a code, and don't shelve anything that isn't logged in the system.

After that, every reference needs a registered location with a logical zone, shelf and level. Storing wherever there's space works for the first few days; after that it leads to long picking runs, improvised restocking and the classic "it was around here somewhere".

  • Fast movers close to the packing station and at a comfortable height. If a reference appears in a large share of your orders, it can't sit at the back just because it arrived first.
  • Physically separate near-identical references. Two sizes of the same model in neighbouring slots is inviting an error; put them on different levels or modules.
  • Group items that sell together (shampoo with its conditioner) without mixing different references in the same slot.
  • Review the map before every campaign. Relocating stock the week before costs far less than fixing errors mid-peak.

If your digital stock and your physical stock don't match, no validation system will save you from shortfalls. You'll find the method for counts and adjustments in the guide to inventory control for accurate stock.

Step 3: scan at picking, don't read

A printed list relies on someone reading correctly, finding correctly and confirming correctly on every line, amid noise, rushing and interruptions. Scanning turns that check into a rule: the operative scans the product on picking it up and, if it's not the one on the order, the system flags it there and then, before it reaches the tote or trolley.

  • Validate quantity too. Scanning one correct unit doesn't confirm that three were actually picked. Require a scan per unit or a controlled quantity confirmation.
  • No verbal exceptions. If the code doesn't match, the reason gets logged (wrong location, mislabelled item, missing stock) with user and timestamp.
  • A photo of the product on screen or on the list helps a lot with similar-looking packaging and with new staff.

With a small catalogue, a barcode scanner and a simple workflow are enough. You don't need to automate everything — you need to automate the point where the error keeps repeating.

Step 4: double-check at packing, before the label

The second check has value because it's independent: it happens at a different time and, ideally, at a different station. At packing, the tote is opened, every item is rescanned, and only once everything matches the order are the extras added, the box sealed and the label generated.

Order matters. If the label is printed before validation, there's a lot of pressure to ship an order that hasn't actually been checked yet. That's where mismatched labels come from: the sound rule is one open order per station, with the label generated from that order's scan, not from a stack printed that morning.

You don't need two people checking every order — that just doubles costs without removing the root cause. Scale the level of control to the risk:

  • Single-unit, low-value orders: scanning at packing is usually enough.
  • Multi-line orders, kits and promotions: full scanning of every unit and component.
  • High-value, B2B or international shipments: add a weight check or a photo of the contents before sealing.

Weight works as a third signal. If a T-shirt should weigh around 0.3 kg and the parcel shows 1.1 kg, something's extra. It doesn't replace scanning, since packaging varies, but it does catch duplicates and items from a different product family.

Choose the picking method to suit your orders

A common mistake is treating every order the same way. The picking method affects both speed and errors, and what really matters is how many lines your average order has and how many references you manage.

Single-order picking

One person prepares a whole order from start to finish. It's the easiest to control and the right choice when you're handling a few dozen orders a day or they're complex orders. Its limitation is travel time: you walk the warehouse once per order.

Batch picking

The total quantity needed of a reference is picked at once for many orders and then sorted out. Very efficient when most orders have one or two items, for example after a launch. The risk shifts to the sorting stage: validation at packing is essential there.

Wave and zone picking

Waves group orders that share a dispatch time or carrier; zones assign each person an area, with orders consolidated at the end. Both gain speed with volume and a wide catalogue, but consolidation is the critical point: totes identified by order and a scan every time an item is placed in.

Connect your shop and warehouse in real time

An order copied by hand from the shop into a spreadsheet is an error waiting to happen: a mistyped variant, a customer note that gets lost, an incomplete address. The integration with your platform should bring in the order with its variants, notes and extras, and send stock updates and the tracking number back to the shop.

Packaging instructions need to travel with the order too. If the welcome insert depends on someone remembering, it'll be forgotten the first time things get busy. The same goes for carrier choice: a well-picked order can still fail if it goes out with a service that doesn't suit its weight or destination.

Train the team on the process, not with tips

Training is the most basic thing — and the most neglected. But "be careful with sizes" isn't training. What actually works on the warehouse floor is concrete:

  • On day one, the new starter picks alongside someone else and doesn't touch multi-line orders or kits.
  • A short sheet listing the references that get confused most often and how to tell them apart, pinned up at the station.
  • Explaining why each check exists: someone who understands that the label goes on last won't print it early
Back to blog