Variants and attributes: where a catalogue actually breaks

A catalogue with a thousand products and no variants is simple. The same catalogue with three sizes and four colours per product is twelve thousand combinations, and nearly every serious ecommerce problem starts there: in rejected feeds, in search that finds nothing, in near-duplicate pages, in returns.

What a variant is, to whoever has to read it

To you a variant is the same t-shirt in red. To a search engine, an ad feed and a marketplace they are three different things, and each wants them declared differently.

The practical rule: a variant is something that changes the barcode and does not change the buying decision. A t-shirt's colour is a variant. A t-shirt and a sweatshirt with the same print are not, even if the back office keeps them together. Getting this distinction wrong produces pages nobody can find, because they describe two different products at once.

The four ways it breaks

The shared barcode. Every variant with the same code, or with none. From there on no channel can tell them apart, and in Shopping they are treated as one product or rejected as a block.

The attribute in the title instead of the field. "Red cotton t-shirt XL" with colour and size written only in the name. To a human it is clear; to a filter and a feed it does not exist: the colour field is empty, so the product appears in no colour selection.

The out-of-stock variant that sinks the whole product. If the default variant is out of stock, several channels treat the entire product as unavailable even when ten sizes out of twelve are in the warehouse. It is a loss you cannot see, because the product is there, it is simply not shown.

One page per variant. The opposite error: twelve nearly identical pages instead of one with twelve variants. Signals get split across twelve addresses, near-duplicate content is created, and none of the twelve ranks as one would have.

The attributes everyone forgets

Beyond size and colour, four are missing almost everywhere and needed constantly.

Material, written as data. Not "premium fabric": 95% cotton, 5% elastane. It is filterable, it is quotable, and it is among the most searched things there is.

Product dimensions, not packaging dimensions. Confusing the two is a classic, and it produces expectation returns.

Compatibility. On anything that attaches, fits into or replaces something, it is the attribute that decides the sale and that almost nobody puts in a field.

Unit and quantity. A pack of six is a different product from a single, and without that field price comparisons become meaningless, including for engines.

Why it never gets fixed

Because the work is dull, belongs to nobody, and has no deadline. Nobody is assessed on having filled the material field across three thousand items, while everybody is assessed on the campaign launched this week.

Which is also why it is an available advantage: whoever does it ends up with a catalogue the channels can read, while competitors keep buying traffic to pages the channels do not understand.

Where to start

With a brutal count: for each important attribute, how many products have it filled. No tool needed, an export and a spreadsheet will do. You will almost always find two or three fields empty across eighty per cent of the catalogue, and they are always the same ones everything else depends on.

Frequently asked questions

One page with twelve variants, or twelve pages?

One page with twelve variants, in the great majority of cases. The exception is when variants are searched as different products and have demand of their own: then a dedicated page is worth it, but that should be a deliberate choice, not whatever the back office produces.

What if a product has no barcode?

It happens with handmade goods, made-to-order items and own brands, and it is provided for: there are correct ways to declare that none exists. The error is not lacking one, it is leaving the field empty without saying so, because then the channel cannot tell whether it is missing or absent.

How many attributes should I have?

Few and complete across the catalogue, rather than many and half filled. An attribute populated on thirty per cent of products is worse than not having it: filters built on it hide products that would in fact have suited.

Do I need a PIM for this?

You need one place where the data is right, which is sometimes a PIM and sometimes Shopify used with discipline. The real problem is not the tool, it is the same attributes being maintained in two places that drift apart.

Do variants slow the site down?

Catalogues with very many variants can weigh on collection pages, but that is a theme implementation problem, not a reason to cut variants. Cutting them for speed is solving a technical problem by removing sellable products.