Why Picking Errors Happen, and What They Actually Cost

Every merchant who ships their own orders has sent the wrong thing at least once. Most treat it as bad luck, or as a bad day. It is neither. Picking errors follow patterns so consistent that you can predict which orders in your shop will generate them.

And they are far more expensive than the item you shipped by mistake, which is the part that usually gets missed when deciding whether to do anything about it.

The real bill for one wrong parcel

Start with what a single mistake costs, because the number changes the decision.

You pay the shipping twice, out and often back. You pay the correct item's shipping a second time. You pay for the time: reading the complaint, finding the order, working out what actually happened, apologising, re-picking, re-packing. On a small team that is rarely under twenty minutes, and it lands in the middle of something else.

Then the inventory goes wrong in two places at once. The item you shipped by mistake is now missing from a stock count that still says it is there, and the item you should have shipped is sitting on the shelf marked as sold. You discover this weeks later as a mysterious discrepancy.

And the part nobody puts in the spreadsheet: the customer now checks their next parcel. Some of them just do not order again, and they do not tell you why. A refunded order at least gets recorded. A customer who quietly stops buying does not.

Add it up honestly and one wrong parcel on a small store rarely costs less than several times the value of the item. Which means an error rate of a couple of percent is not a rounding error, it is a line item.

The five patterns

In practice, mistakes cluster around the same five situations.

Variants that look alike. Same product, different size, different colour, different capacity. The picker reads the product name, which is identical across the variants, and the distinguishing attribute is the smallest text on the screen. This is the single most common error in stores that sell apparel, components or anything with options.

Quantities above one. An order for three of something gets one, because the line was read as a line rather than as a quantity. Multi-quantity lines are read fastest and checked least.

Multi-item orders, late in the pick. Errors do not spread evenly across an order. The first item is almost always right. The last item in a six-line order is where the short-shipment lives, because by then the picker is mentally already taping the box.

Interruptions. A phone call, a customer in the shop, a supplier delivery. The pick resumes from memory, and memory reports that it is further along than it is. Any process that relies on holding state in your head fails exactly here.

Two people, one order. On busy days, the same order gets picked twice, or half-picked twice with each person assuming the other finished. This one is rare and expensive, and it is purely a tooling problem.

Notice what these have in common. None of them are caused by carelessness, and none are fixed by telling people to be careful. They are caused by holding state in your head while being interrupted, and by information that is present but not prominent.

What actually reduces them

Confirm each line as it goes in the box. This is the highest-value change available and it is almost embarrassingly simple. The pick stops being a memory exercise and becomes a list with checkmarks. An interruption costs you nothing because the state is on the screen, not in your head. The last line of a six-line order gets the same attention as the first because it is the only one still unticked.

Make the distinguishing attribute the biggest thing on screen. For variants, the size or the colour matters more than the product name, which is identical across the options anyway. Most picking screens render them in the same weight, which is exactly backwards.

Make quantity impossible to skim past. Anything above one should be visually loud.

Lock the order to the person picking it. Not as an ownership policy, as a physical constraint: two people cannot take the same order, so the failure disappears rather than being managed.

Pick in a fixed order. Same sequence every time, ideally matching your shelf layout. Predictability is what lets a tired person be accurate.

Do not verify with a second full pass. It sounds rigorous and it does not work: checking a box you just packed suffers from the same blind spot that filled it. Confirmation during the pick catches what a re-read does not.

What to measure before you change anything

Be careful with this, because most stores measure the wrong thing.

Counting complaints undercounts badly. Customers who receive an extra item rarely write in. Customers who receive the wrong size sometimes keep it. Your complaint rate is a floor, not a measurement.

Two numbers are more honest. Stock discrepancies found at your next count, because a picking error always leaves two wrong counts behind. And the proportion of after-sales cases that trace back to a shipping mistake rather than a product fault, delivery damage or a change of mind. If you attach every case to its order, you get that split without any extra work.

Where the tooling fits

Order Prep Pro is built around the confirmation idea rather than around a nicer order list. The picker ticks each item as it goes in, an order being picked is locked to the person picking it, and the after-sales cases stay attached to their original order so the pattern is visible later instead of anecdotal.

It also runs on the phone or tablet in the picker's hand, which matters for a reason that only shows up in practice: a picking list you have to walk back to a computer to update does not get updated during the pick. It gets updated afterwards, from memory, which is the thing we were trying to eliminate.

If you are setting this up for the first time, the practical starting point is not a tool at all. It is asking which of the five patterns above produced your last three mistakes. Most stores find they all came from the same one, and that fixing it takes an afternoon.