A picking error — shipping the wrong product, wrong size or wrong quantity — costs 3 to 5 times the value of the original shipment, because you're not paying for one extra delivery: you're paying for the product in limbo, two shipments, support time, an inventory mismatch and, often, the negative review.
Picking is simply the moment someone physically pulls the products of an order from where they're stored. It's the most repetitive step in the whole operation and, for that reason, where a small error rate turns into a large number of problems.
The cost breakdown
Take a 500-peso product with 120-peso shipping:
| Item | Cost |
|---|---|
| Original shipment (already paid, unrecoverable) | 120 |
| Pickup of the wrong product | 120 |
| Shipment of the correct product | 120 |
| Customer support time (~20 min) | ~80 |
| Warehouse reprocessing | ~50 |
| Direct subtotal | 490 |
| Wrong product, if it comes back damaged or incomplete | up to 500 |
That single-order error just ate the margin of several good sales. And that's before what never gets invoiced: the customer who doesn't come back.
The math that should worry you: your error rate
The industry standard sits around 99.5% accuracy, with the best operators above 99.9%. It sounds like a negligible difference. It isn't:
| Accuracy | Errors per 10,000 orders | Approx. cost (at 490 each) |
|---|---|---|
| 98% | 200 | 98,000 |
| 99.5% | 50 | 24,500 |
| 99.9% | 10 | 4,900 |
Between running at 98% and running at 99.9% there's 93,000 pesos a month at a volume of 10,000 orders. That's the real size of the problem, and why it deserves to be treated as a design issue rather than a discipline issue.
Why errors happen
Nearly every picking error fits into five causes, and none of them is "the operator wasn't paying attention":
1. Similar products stored together. The same shirt in three adjacent sizes, or two presentations of the same product differing only in weight. If the only visible difference is in the fine print, they'll eventually get mixed up.
2. Poorly defined locations. If a product lives "somewhere in aisle 3," everyone looks for it differently. If it lives in a coded location, it gets pulled from the same place every time.
3. A dirty catalog. Two codes for the same product, or a code that doesn't match what the label says. The error is inherited from the system, not made on the floor.
4. No final verification. If nobody confirms that what was packed matches the order, the error isn't caught until the customer catches it.
5. Rushing during peaks. Peak season spikes volume and brings in temporary staff. Without a process that holds, errors grow exactly when they're most expensive.
The controls that actually work
Mandatory scanning at two points. When the product is picked and when it's packed. The scanner checks against the order and won't let it close if they don't match. By a wide margin this is the control that eliminates the most errors, because it doesn't depend on anyone's memory.
Uniquely coded locations. Every product has an address and the system tells the operator exactly where to go. It removes searching — which is where you grab the item next to the right one.
Physical separation of confusable products. If two references look too similar, they don't get stored together. It's the cheapest fix and the least applied.
Weight verification. If the order should weigh 800 grams and the box weighs 500, something's missing. It catches shortfalls that scanning alone won't see.
Measure per person, without punishing. Logging who picked each order isn't for assigning blame: it's for spotting whether someone needs more training, or whether an error repeats on a shift or in a zone of the warehouse. Used to punish, people hide errors and you lose the information.
The inflection point
At 20 orders a day, manual verification is enough. From roughly 100 orders a day, accuracy stops depending on how careful people are and starts depending on the system: scanning, locations and automatic verification.
That, in practice, is one of the moments when outsourcing the operation starts to make sense — not because of the space, but because building the quality control from scratch costs more than renting it already built.
Frequently asked questions
What's an acceptable error rate?
Below 99.5% accuracy you have a problem that's already showing up in your reviews and your support load. Above 99.9% you're in best-operator territory. What matters is measuring it: if you don't know your rate, you can't know whether it's improving.
Doesn't scanning slow picking down?
For a few days, yes. After that it speeds it up, because the operator stops second-guessing and mentally double-checking. The time spent scanning is a fraction of the time spent resolving a mis-shipped order.
How do I count an order that shipped incomplete but with the right product?
It counts as an error. To the customer, an order missing a piece is a mis-shipped order, and it costs the same to resolve. Counting only "wrong product" gives you an artificially flattering error rate.
Want to know your real error rate and what it's costing you? See Ecommex fulfillment: we pick with two-point scanning and weight verification, and give you the accuracy metric live — not in a monthly report.