Barcodes: the cheapest accuracy improvement in a stockroom

Most stock errors begin as a keystroke: a transposed pair of digits, a dropped zero, a supplier's code typed from a carton that spent a week in a van. A scanner removes that whole class of error, and it costs less than any other change of comparable effect — a printer at the door and a device in a hand. It also fails in a specific way: when the label encodes something the system cannot resolve, scanning is slower and less accurate than typing.

What actually goes on the label?

Three facts matter about a carton — the item, the batch, the place it belongs — and only one of them belongs on the carton.

The item code is the identity, and it is printed when the goods enter your control. The batch is a property of that arrival, so its label is produced at the same moment, not later. The location is a property of a shelf, and a shelf does not move; putting the location on the box means that moving the box makes the label wrong twice, once for the box and once for the shelf.

So the scanner reads the item first, always, and the batch only when the process asks for it. A label that leads with the location teaches the receiver to scan a shelf code into an item field, and the resulting record is confident and wrong.

What it identifies	Where it belongs	Read when
Item	Carton label, printed at the door	Every scan
Batch or lot	Carton label, printed on arrival	Traceability, expiry, a complaint
Location	Shelf edge, its own label	Put-away, picking, the count

What happens when the carton already carries a supplier code?

Nothing good, unless you decide which code is authoritative before the first delivery. The supplier's number is a property of the supplier, not of the item: two suppliers can print the same number for different products, and one supplier can change its number when the packaging changes.

So the internal code is the identity and the supplier code is a mapping. The scan resolves the supplier's number through a table to your item, and the table records which supplier, since the same digits from two suppliers are two facts. An identifier that changes when the packaging changes is not an identifier: stability and recorded origin are the two things to ask of any code you adopt, which is the same demand the W3C makes of published data.

When both codes are physically on the carton, choose one of two honest options: cover the supplier's barcode with your own label, or scan the supplier's and let the mapping do the work. What does not work is leaving both readable in front of a receiver with a queue behind him — he will scan whichever is nearer, and the stockroom will hold one item under two identities.

Print at the door, or at the shelf?

At the door, where ownership begins and where the goods are still in one place. That single decision settles most of the argument.

Printing at the shelf means labelling after put-away, so goods move through the stockroom unlabelled, and the label is applied by whoever is least busy rather than by the person who checked the delivery. It also splits one job into two: someone receives, someone else labels, and the gap between them is where a carton gets lost.

A printer at receiving has its own costs, and they are the ones people forget: a device that jams, labels that peel in heat, a queue at the door at eight in the morning. That is why the label size and the printer position get decided by watching a real delivery, not by reading a specification.

Does the scanner work with gloves on?

That is the question that decides the purchase, and it is physical rather than technical. The people scanning are holding a box, wearing gloves, standing in a cold room or in dust, and they have one hand free at most.

What matters: a trigger reachable with a glove, an engine that reads a scuffed label at an angle through a plastic sleeve, a battery that lasts a shift rather than an afternoon, a screen legible in daylight, and a body that survives being dropped on concrete. What does not matter: the brand, the operating system, the colour.

Two consequences follow. A device that needs both hands gets put down, and a device left on a shelf records nothing for the rest of the shift. And the software half has to work without a signal: queue the scans, show how many are waiting, and never let a lost connection look like a successful scan.

What does nobody budget for?

Relabelling the stock you already own. It is the part that turns a tidy pilot into a project, and it is almost always missing from the quote.

Everything on the shelves today predates the internal code, or carries an older one. Someone has to count it, print a label for it, stick the label on it and enter it — by section, on foot, with a trolley. Do it at the same time as a stock count, so the same walk pays for both, and start with the fastest-moving section, because that is where a scanning error costs the most.

There is also a trap in the old labels. If the old barcode still scans and still looks like the new one, an operator will use it, and the record will resolve to nothing — or, worse, to a different item that inherited the number. Retire the old labels deliberately: cover them, or change the length or format so an old label fails loudly.

The order to do this in

1. Fix one internal code per item, and freeze it: no reuse, no reissue after a rename, no renumbering when a supplier changes.
2. Map supplier codes to it in a table that records the supplier, and never let a supplier's number become an item.
3. Decide where ownership begins — the door — and put the printer there.
4. Keep the shelf label separate from the carton label, with its own format.
5. Choose the device by how it is held, not by its datasheet: gloves, one hand, a full shift, a drop.
6. Make the scanning app work offline, with a visible count of what has not synced yet.
7. Relabel the existing stock by section, alongside a stock count, starting where the volume is.
8. Only then remove the keyboard, or leave it behind a supervisor's account for the exceptions.

Where this goes wrong

• Two labels, one carton. Both readable, and the receiver picks the nearer one; the item is now recorded twice under two identities.
• The label is printed before the goods are checked. Damaged and short units get labels, and the label becomes evidence that they arrived in good condition.
• The location label is stuck on the box. Move the box once and two records are wrong.
• The item code contains the price or the batch. A price change creates a new item, and the history splits in two.
• The scanner types into whatever field has focus. Without a deliberate flow, a scan lands in the quantity box and the count is silently overwritten.
• Nobody tests a worn label. The label that scans cleanly in the office is the one that fails on a dusty carton at the back of the van.

What this does not cover

This is about labelling and scanning as a stockroom practice, not about which symbology or standard to print: that follows your printer, your trading partners and any requirement your customers impose, and it changes with them. It does not cover radio-frequency tags, automated storage or robotics, all of which solve a different problem at a different scale. And it assumes you sell countable units — a business selling by weight needs the label to carry a measured quantity, and the arithmetic changes. Nor does scanning repair a stock model that is wrong: barcodes make a correct model faster, and they make a wrong one faster too.

We build the scanning step around one code per item: a label printed at the door, a mapping for the supplier's number, and an app that keeps recording when the signal does not.