A SKU is supposed to be boring. One code, one sellable thing. On a lot of catalogs I inherit, a SKU is a suggestion. The same knit cap exists as 44821, 44821-BLK, and “black cap — warehouse.” Shopify will let you sell all three. A wholesale buyer will order the wrong one and then call somebody.
I learned this the hard way on apparel and general merchandise catalogs, not from a course. United Textile had pack quantities. Universal Wholesale had thousands of dollar-store SKUs that looked similar until you read the case pack. If attributes are sloppy, search is sloppy, filters are sloppy, and the order is sloppy.
SKU, variant, attribute — I keep these separate on purpose
People mix these words. I do not, because the import file will punish you if you do.
- The SKU is the unique ID for the thing that ships. If two rows share a SKU, that is a fight waiting to happen.
- A variant is a sellable difference on the same parent: size, color, voltage, pack count.
- An attribute is how you describe and filter that difference — and how you keep it consistent across the catalog.
If “12-pack” is a variant on one product and a title prefix on another, your collections stop meaning anything. Buyers stop trusting the grid. They email for a spreadsheet, which means your store did not do its job.
How I clean a variant structure before it goes live
I write the option names down first. Color. Size. Pack. That’s it, unless the line honestly needs a fourth. Then I normalize the values: Black not BLK not “blk.” Large not L not “Lg.” Case of 12 not “12ct” on one row and “12 / cs” on the next.
On WooCommerce this becomes attributes and variations. On Shopify it becomes options and variants, with a SKU on each variant, not only on the parent. Image mapping follows the variant, not the parent, when the photo actually changes.
What messy data costs in real orders
It is not theoretical. Wrong pack quantity is a margin problem. Duplicate SKUs break inventory counts. Missing attributes mean a B2B buyer cannot filter 1,000 caps down to the ones they restock. They bounce, or they place a tiny test order because they do not trust the catalog.
I would rather argue with a spreadsheet for a day than argue with a customer about a case pack they thought was a unit. Product data management is sales support. It just does not look like sales support while you are doing it.
A simple test before I call the catalog done
I pick one parent product and try to order every variant in my head. If I cannot tell which SKU ships, a buyer cannot either. Then I search the store for a color word and a pack word. If the filters lie, the attributes still need work. This takes twenty minutes and saves a week of “can you check this SKU” emails.
