A picking error is cheap to make and expensive to fix. It happens in seconds, in the aisle, and it is paid for in customer service, return freight, stock adjustments and, very often, in the trust of the customer who never complains and simply buys less. Almost every warehouse knows it has errors; few measure them with a stable definition. This guide explains what counts as an error, the four families, how the rate is calculated, where each one is detected and which causes are fixed without buying equipment.
What a picking error is
A picking error is any difference between what the order asked for and what actually left the warehouse, provided it is attributable to picking.
That last part matters. If the system allocated stock that did not exist, the error belongs to inventory. If the carrier lost a package, it belongs to dispatch. Mixing all three into a single indicator is the most common reason a warehouse "has a lot of errors" and never manages to bring them down: each family is corrected in a different place, by different people.
Write down what is in scope and what is not, who records it and at what moment. An error caught by the outbound check and one caught by the customer are the same error, but they cost very differently: count them together in the rate and separately in the cost.
The four families of error
Almost every picking error falls into four categories, and it is worth recording them separately because their causes differ.
Wrong reference: a different product shipped. It typically happens between similar items stored side by side, with similar packaging or codes that differ by one digit.
Wrong quantity: the reference is right but more or fewer units went out. It shows up when goods are picked by case and invoiced by unit, or when counting is done from memory.
Short pick: the line was never picked and the order shipped incomplete without warning. Almost always an empty location nobody reported.
Damaged: the product left dented, wet or with the packaging open. It is the only family that usually originates outside the pick itself, in storage or in pallet building.
A board with those four columns, even in a spreadsheet, will tell you within two weeks where your problem is.
How is the picking error rate calculated?
The formula is simple; what matters is choosing the denominator well.
The rate per line divides lines with an error by total lines picked in the same period. It is the most useful measure for warehouse work, because every line is an opportunity to get it wrong.
The rate per order divides orders with at least one error by total orders. That is the one the customer sees, because an order with one error is a badly delivered order, whether it has three lines or forty.
Keep both. The rate per line tells you whether the process is improving; the rate per order tells you what the customer is experiencing. And both are measured over the same period of at least four weeks, so a bad week is not mistaken for a trend.
Where each error is detected
Where the error surfaces determines what it costs and how quickly you can correct it.
At the pick itself, if there is code verification: the cheapest scenario. It is fixed on the spot and never reaches the customer complaint log.
At the outbound check, before loading: it costs the checking time and the rework, but it does not leave the warehouse.
At the customer goods-in: this is where it starts to hurt. There is return freight, a credit note, an urgent replacement and a phone call.
At stock reconciliation, weeks later: the worst case, because you no longer know which order it was and the adjustment is made blind.
Recording the detection point alongside the error family is what later lets you prioritise. Not all errors are worth the same, and the board has to show that.
What each point of the rate costs
Value recovered time using your actual operating costs and measurements of the tasks that change. Distinguish time available for other work, reduced overtime and expenses that can actually be avoided.
The calculation goes like this: for every error, add up detection time, rework time, administrative handling and return freight if it came back. Multiply by the number of errors in the period and you have the direct cost of that point of rate.
What is missing is the indirect cost, which is larger and never invoiced: the customer who does not complain and buys less. You will not measure it precisely, but it is worth naming when an investment is being discussed.
Process causes that are fixed without buying anything
Most errors have process causes, and they are corrected with organising work.
Neighbouring locations holding similar products: separate them. It is the most frequent cause of wrong-reference errors and costs nothing beyond relocating.
Unreadable or missing location labels: a large label, with the full code, at eye level, brings errors down immediately.
Picking from memory: if the picker knows the product, they do not read the code. Barcode verification at the pick cuts that whole family.
Late replenishment: an empty location generates silent short picks. Define a minimum replenishment level and who watches it.
Untrained staff: a new operator repeats the mistakes of whoever trained them. A written travel path and two weeks of shadowing are worth more than any equipment.
What an automated system reaches, and when it is not needed
Inventory accuracy is measured by comparing records with physical stock. Results depend on receiving and dispatch records, location rules, training and controls; software or robots do not guarantee an accuracy percentage.
That said: it is not what you need if your error rate still comes from badly labelled locations, late replenishment or untrained staff. Automating over a disorganised process freezes the disorder and makes it far more expensive to change.
Payback is calculated using the full investment and your operation’s net savings. Separate genuine expense reductions from capacity available for growth; freeing space or time does not automatically save cash.
Measure first, decide afterwards
An error rate is only useful once it is defined: what is counted, over which denominator, in which period and with which detection point. With four weeks of recording split across the four families, most warehouses discover that two causes explain more than half the errors, and that both are fixed by relocating, labelling and verifying. Only once that work is done and the rate has plateaued does the equipment conversation make sense, and only then can it be argued with your own numbers.
Frequently asked questions
Any difference between what the order asked for and what left the warehouse, as long as it is attributable to picking. If the allocated stock did not exist, it is an inventory error; if the package was lost on the truck, it belongs to dispatch. Separating those three families is what allows each to be corrected in the right place.
Divide the number of lines with an error by total lines picked in the same period, and keep the rate per order alongside it, which divides orders with at least one error by total orders. The first measures the process; the second measures what the customer perceives. Both over at least four weeks.
Four families: wrong reference, wrong quantity, short pick and damaged goods. Wrong reference usually comes from similar products stored side by side; wrong quantity, from counting by memory; short picks, from an empty location nobody reported; damage, almost always from storage or pallet building rather than the pick itself.
In most cases yes, and it is worth trying first. Separating similar references, labelling locations with the full code at eye level, verifying by barcode at the pick, defining a replenishment level and shadowing new staff are process measures that bring the rate down without buying equipment.
Inventory accuracy is measured by comparing records with physical stock. Results depend on receiving and dispatch records, location rules, training and controls; software or robots do not guarantee an accuracy percentage.