Almost every warehouse problem traces back to the same thing: the product has no single identity. The same item sits under one code in the warehouse, another on the site and a third on the marketplace. The result is reports that do not reconcile, stock sync that misfires and counting variance nobody can trace.
This guide walks through the order we follow to build the standard from scratch and migrate existing stock onto it. You can apply it yourself.
1. Decide first: what gets a code
An SKU is the smallest sellable form of a product. Different size or colour means a different SKU, because stock is held separately. If a packaging difference changes what is sold, that is a separate SKU too.
What you do not encode is also a decision: batch numbers, serial numbers and expiry dates are not part of the SKU. Those are tracking fields, not identity. Putting a date in the SKU turns the same product into a new product with every shipment.
2. Make the SKU human readable
A good SKU reminds the reader what the product is without becoming an essay. The structure we recommend has four parts:
BRAND-CATEGORY-MODEL-VARIANT
For example AQ-TSH-BASIC-BLK-M: brand abbreviation, category abbreviation, model name, colour, size.
Three rules:
- Use fixed lengths, it makes sorting and searching easier
- No local characters, spaces or slashes; they break when moving between systems
- Keep abbreviations in a dictionary so everyone uses the same ones
When deciding the length, size it for your catalogue three years from now, not today.
3. Barcodes: GTIN or internal
They do different jobs.
A GTIN, the global number for a product, faces outward. Marketplaces and retail chains ask for it. If you manufacture your own product you get it through GS1 membership; if you import, it is already on the item.
An internal barcode works only inside your warehouse. Use it for products without a GTIN, for bundles and for kits you assemble yourself.
The rule is simple: use the GTIN if it exists, otherwise generate your own, and map both to the SKU. A product can have more than one barcode, but only one SKU.
4. Design label templates up front
Three label types are enough:
- Product label: barcode, SKU, product name, variant
- Shelf label: shelf address and its barcode, scanned during picking
- Carton label: contents summary, quantity and dispatch information
Most thermal printers speak ZPL or TSPL. Building templates in that language means a printer change does not force a redesign. Choose label size against the smallest surface of your product; a barcode that will not scan means manual entry.
5. The migration plan for existing stock
This is the step people fear most, and done in order it takes a few days.
- Export the current product list
- Generate new SKUs in a mapping table: old code, new code, barcode
- Review the mapping; this is where you catch the same product sitting on two rows
- Print labels and work shelf by shelf, do not try to convert the whole warehouse at once
- Count each zone as you finish it; migration and counting can happen together
- Do not delete the old codes, keep them in the mapping table for reading historical orders
6. Pushing it to the channels
A new SKU structure that stays in the warehouse is worthless. In order:
- Update product codes on the site and in your panel
- Map the stock code field on every marketplace to the new SKU
- Verify which data the code field carries in shipping and accounting integrations
- Push a test order through and follow the chain end to end
Do not skip the test order. Most integrations keep running quietly while pointing at the wrong field.
7. Keeping the standard alive
After it is set up, there is only one reason a standard degrades: whoever creates a new product does not know the rule.
A single page document solves that. Put the abbreviation dictionary, example codes and the steps for creating a new product on it. New staff get trained from that page.
Run a short check monthly: any codes outside the standard, any product sitting under two codes, anything left without a barcode.
Bundles and kits
If you sell two products together, the bundle gets its own SKU while physical stock comes off the components. This has to be set up as a recipe or bundle in the system; without it, component stock drifts every time a bundle sells and the gap surfaces at counting.
The rule: if the bundle is a sellable unit it has an SKU, but stock is held on the components.
When batch or serial tracking is required
Electronics, cosmetics and food often require batch or serial tracking. That information does not belong inside the SKU. It is held in a separate field and, where needed, carried inside the barcode or added on a second label.
The product identity stays fixed and traceability runs as its own layer. Brands that push batch numbers into the SKU end up splitting one product across hundreds of codes, and reporting becomes unreadable.
Rolling it out on the floor
The standard does not finish at a desk. In the warehouse we go in this order:
- Start with the shelves touched most during picking, so the effect shows on day one
- Print labels in the warehouse, not in the office; a wrong label gets caught before it reaches a shelf
- Verify by scanning at every shelf, visual checks are not enough
- Count each zone as you finish it and record the variance, so migration and counting advance together
This order makes it possible to migrate without stopping work. Brands that try to close the warehouse and convert everything at once usually stall halfway.
Common mistakes
- Putting price or supplier information in the SKU; both change, the code should not
- Using local characters; they break in transfer
- Leaving the variant out of the SKU and only in the product name; stock then shows as one line and size level tracking becomes impossible
- Forcing the label onto the smallest surface of the product; a barcode that will not scan means manual entry every time
Applying this guide takes a few days and the effect is permanent. Stock accuracy, counting speed and channel sync all hang on this one decision.