Name the item your warehouse must actually pick
A product family is not always the item that goes into a parcel. A shoe style may have several colors and sizes. A bottle may be sold individually or in a three-pack. Start with the exact combination that your team must pick, pack, and count. If two options cannot be substituted without changing the customer's purchase, they need to be distinguishable in your operational records.
For a fictional luggage brand, a charcoal carryall and a navy carryall might share photography and a product page, but they remain separate sellable variants. A bundle containing the carryall and a pouch introduces another question: is it stocked as a prepacked unit, or assembled from components? The answer determines how its identity should connect to inventory. Resolve that physical reality before designing a clever code.
Give each identifier one job
Your SKU is an internal convention that helps your business distinguish stock. A platform variant ID identifies a record inside that platform. A marketplace offer ID identifies a particular selling offer. A GTIN identifies a trade item under the GS1 system. These identifiers can coexist without being interchangeable, and keeping them separate makes integrations easier to reason about.
Do not replace a stable internal SKU merely because a marketplace assigns a new listing reference. Likewise, do not invent a GTIN from a product title or treat an internal warehouse code as if it were globally registered. Verify the identifier requirements for the destination and the product. The aim is a clear relationship among identifiers, not one universal string squeezed into every field.
Build a SKU convention people can maintain
A good SKU convention survives new products and ordinary human mistakes. Keep it consistent, avoid easily confused characters where practical, and leave room for the assortment to grow. Encoding a small number of useful attributes can help pickers, but an excessively detailed code becomes fragile when merchandising changes a color name or product category.
Document examples and assign responsibility for issuing new codes. Check uniqueness before a product reaches a channel. Shopify's SKU guidance recommends distinct codes for distinct variants and warns that duplicates can interfere with integrations. If your business intentionally uses a special bundle or shared-stock convention, document the exception and test the tools that depend on it. An unexplained duplicate is much more dangerous than a deliberate, supported rule.
Make the mapping table explicit
For every sellable variant, keep the internal SKU, central platform variant reference, relevant barcode values, channel, seller account, and channel listing or offer reference. Add condition, pack quantity, and location where those affect fulfillment. Avoid relying on a single column called product ID when different teams mean different things by that label.
The table should answer a concrete question: when this channel sends an order for this listing, which exact inventory record should be used? If the answer requires a person to compare pictures or remember that an old title means the new blue model, the mapping is not ready. Keep unresolved matches out of automatic order creation until the responsible person confirms them.
A useful spot check starts at the shelf and ends at the channel. Scan one physical item, find its internal variant, inspect every external mapping, and compare the unit quantity in the listing. Then run the path in reverse from a recent order. This catches errors that a tidy spreadsheet can hide, including a correct barcode attached to the wrong color or an external pack listing connected to a single-unit SKU. Record the evidence once so future reviewers can repeat the same check.
Preserve identity when product presentation changes
Merchandising changes titles, images, and descriptions regularly. Those updates should not force every downstream system to rediscover the product. Use stable record relationships and retain historical references when an offer is replaced. A listing that is ended and recreated can still represent the same physical variant, while a similar-looking new model may represent a genuinely different item.
Bulk imports deserve special care. Shopify's product CSV guidance notes that changes to option values can affect variant IDs and third-party dependencies. Before a large variation cleanup, identify the integrations that use those IDs and test a sample. A cosmetic catalog edit can become an order-routing problem if the identity underneath it changes without updating the relationship records.
Handle packs and bundles before they surprise you
A listing for three bottles should not decrement one bottle unless your inventory record itself represents a stocked three-pack. Write the unit conversion explicitly. If a bundle consumes two different components, record both component requirements and define what happens when one is unavailable. Matching only the bundle's marketing name cannot answer those questions.
Test returns with the same care. A customer may return one component, a complete set, or an item in a condition that cannot be resold. The returned quantity should update the correct physical unit after inspection, rather than blindly reversing the original listing quantity. These details are easier to design when the identity model includes units and composition from the beginning.
Run a small identity audit every month
Choose a rotating sample of products and trace each one from the central catalog to a marketplace offer, an order line, a warehouse pick, and a return record if available. Include products with multiple variations, packs, renamed titles, and recently replaced listings. Look for missing relationships, unexpected duplicates, and references that now point to archived records.
Measure the number of orders held because a product could not be matched, rather than only the percentage of catalog rows with a SKU filled in. A filled field can still be wrong. Prioritize products with recent sales and those involved in repeated errors. A short, targeted identity review is usually more useful than periodically exporting the whole catalog and hoping obvious problems stand out.
Make new-product setup a handoff, not a memory test
Before a new variation goes on sale, give the next team a concise record: the exact item, approved identifiers, pack quantity, stock source, channel mappings, and any exception. Have someone other than the creator verify one representative order path. That second look often catches a copied barcode or an option that was accidentally mapped to the neighboring size.
Keep the process light enough that it happens for ordinary products, not only major launches. The payoff is practical: fewer ambiguous orders, cleaner inventory adjustments, and a faster answer when support asks which item the customer bought. Product identity is rarely the most glamorous part of commerce, but it is the part that lets the rest of the workflow behave predictably.
Sources & further reading
Official documentation for the concepts discussed. Platform requirements can change; check the linked source for the current details.