SKU generator

What a SKU is, and what it is not

A SKU (stock-keeping unit) is an identifier you own. It exists so your own systems — a spreadsheet, a shop till, a warehouse — can tell one thing you sell apart from another. Because it is yours, there is no authority to apply to and no standard to conform to. That freedom is the whole point, and it is also where schemes go wrong.

It is not a retail barcode. GTIN, UPC and EAN are the globally unique numbers that go in a barcode on a product, and they are issued by GS1 under rules you cannot invent. A SKU never leaves your business; a GTIN appears on shelves worldwide. People say "SKU" when they mean "barcode" often enough that it is worth keeping them apart, because the two have completely different rules.

This tool generates the SKU strings themselves and nothing else. It does not render a barcode and does not emit GS1 data, so what you get is a list of codes to use in your own systems.

How the segments combine

A scheme is built from three kinds of segment, and they combine as a product rather than a sum:

That multiplication is the thing to internalise before you design anything. Two choices of three values and a sequence of ten thousand is not thirteen items — it is sixty thousand codes, from a form that took a minute to fill in. The tool refuses to build past fifty thousand and tells you the true total as you go, so the arithmetic is visible rather than a surprise at export time.

Designing a scheme that survives growth

The mistake almost everyone makes is designing for the catalogue they have now. A scheme is cheap to get right at the start and expensive to change once codes are printed on labels, listed in a shop and quoted on invoices.

• Keep each segment's width fixed. If colour is always three characters, then TSH-BLU-M-0001 can be read by eye and split by a spreadsheet at a known position. A scheme where one segment sometimes has two characters and sometimes four cannot be parsed without guesswork.
• Pad the sequence. This is the single most common oversight. Unpadded numbers sort as text: 1, 10, 100, 2. Four digits gives you 0001 through 9999 in the order a human expects, in every file listing and every spreadsheet column.
• Leave room for the category you will subdivide. If you might split "M" into "M" and "MW" for a wider fit, decide now whether that is a new value in the same segment or a new segment entirely. Changing it later means either a gap in the sequence or reissuing codes.
• Decide where the sequence goes and leave it there. A trailing sequence keeps codes grouped by variant; a leading one groups by nothing useful. Order is never rearranged for you.
• Do not encode anything that changes. Price, supplier and season are all things that will differ from what you wrote down. A SKU identifies the thing, not its current situation.

Separators, case and padding

The separator is a readability choice more than a technical one. A hyphen is the convention most people expect. A slash reads well in print and badly in a URL, which matters if the code will ever appear in one. Underscores survive systems that dislike hyphens. No separator at all produces the shortest code, and the hardest to read aloud or copy by hand.

The case rule is applied to literals and generated parts alike, so one scheme cannot end up mixing conventions. Uppercase is the usual choice: it reads clearly at small sizes on a label, and it removes an entire class of mismatch where someone types tsh-blu-m-0001 into a system that stored TSH-BLU-M-0001.

Characters to avoid on a label

If a code will ever be read by a person rather than a machine — typed into a search box, read down a phone line, written on a returns form — some characters cost you support time. The classic pairs are 0 and O, and 1, I and l. At small print sizes they are genuinely hard to tell apart, and a mistyped code looks like a missing product rather than a typo.

You do not have to avoid them, but it is worth knowing you chose them. A range built entirely from unambiguous characters is noticeably easier to live with at scale.

Checking the list before you commit

Two things are worth doing before you use a generated list anywhere real:

1. Look for duplicates. A scheme can produce the same code twice — BLU as a colour choice and BLU as a size value will collide. The tool reports collisions it finds rather than handing you a list with a hidden repeat in it.
2. Read a few codes out loud. Pick three from different parts of the sequence and read them as you would to a colleague. If you stumble, your customers will too.

The preview materialises the first thousand codes so a large scheme does not lock up the browser, but the total count shown is always the true one. Export gives you the whole list as CSV for a spreadsheet or TXT for a plain list, and you can copy it to the clipboard directly.

SKUs and QR codes

The two do different jobs and are often used together. A SKU identifies a product inside your business; a QR code is a way to hand something to a phone. On a product label you might do both — print the SKU as text for your own staff and a QR code for the customer.

If you want a code that carries the SKU itself, a plain text QR code does that, and a website link code is better if the customer should land on a product page. If the destination might change — a range that gets renamed, a URL that moves — a dynamic code lets you repoint it without reprinting the label.

Frequently asked questions