Clothing is the hardest category to get right in a product feed, and the reason is variants. A shopper does not search for a shirt. They search for a navy linen shirt in a large, and the difference between showing up for that search and not showing up at all is whether your feed sends the colour, the size and the relationship between the two.
Most apparel feeds we open are exporting one row per style. The store thinks of the product as a shirt with options. The shopper thinks of it as a specific object. Google needs the second version, because it has to decide which of your items is a candidate for a query that contains a colour and a size.
So the work in apparel is not clever copywriting. It is making the feed describe the wardrobe rather than the catalogue, then keeping it that way through every seasonal drop, because an apparel catalogue turns over faster than any other kind.
Send every variant, and tell Google they are family
Each buyable variant should be its own row, with its own identifier, its own image and its own availability. Then all the rows for one style are tied together by a shared grouping value, so the system understands it is looking at one product in several forms rather than at fourteen unrelated shirts.
Get the grouping wrong and two things break. Variants compete against each other in the same auction, spending your budget to outbid yourself. And when one size sells out, the whole style can behave as though it is gone.
The common mistake at the other extreme is grouping too broadly, tying together products that are genuinely different styles because they share a name or a fabric. That confuses the matching in the opposite direction.
The test is simple. One grouping value covers exactly the set of variants a customer would consider the same product with a different size or colour chosen. Nothing more, nothing less. Set this up once, correctly, in the export, and it stops being a recurring problem.
A shopper does not search for a shirt. They search for a navy linen shirt in a large.
Size, colour and fit are searches, not decoration
Apparel attributes are not paperwork. They are the words people put in the query, so every one you leave empty is a set of searches you are not eligible for.
Size needs the value people use, and it needs a size system attached where relevant, because a 10 means different things in different markets. If you sell internationally with one feed, this is worth an hour of thought rather than a guess.
Colour needs the plain word alongside your poetic name. A shopper searching for a green jacket will not type your name for that green. Send the recognisable colour, and keep the brand name for the page. If a garment is two colours, order matters, with the dominant one first.
Size type and fit matter for anything outside the standard cut. Petite, tall, plus, maternity, relaxed. These are the terms shoppers use when they have been let down before, and they carry high intent.
Gender and age group are required in this category and stores routinely leave them at a guessed default. An adult item marked as unisex when it is cut for women loses the searches that specify women.
Titles written for a rail, not for a search box
Apparel titles are usually the shortest in the feed and they should often be the longest. A title of two words was written for a page where the image and the filters carry the meaning. In a results grid, the title is doing the work alone.
The order we use is brand, then the plain product noun, then the attributes that people search on: colour, material, fit, size where it belongs in the title, and the detail that distinguishes this style. A stranger should be able to picture the item from the title alone.
The thing to resist is the internal name. Season codes, style numbers and invented product names mean nothing to a shopper and take space from words that would earn a match. Keep them in your own systems.
One more habit worth building. Write the title for the variant, not for the style, so the size and colour that variant actually is appear in its own title. That is what makes a specific search find a specific item, and it is what stops your grid showing a black version to someone who typed cream.
Images that lose the click before the page loads
In apparel the image is the advert. It is chosen at a glance, against a row of competitors, in a small square on a phone.
The first rule is that the main image should show the item that row is for. Sending the same model shot for every colour is the most common failure in the category, and it produces exactly the disappointment you would expect when the shopper arrives to find the colour they clicked is not the one shown.
The second is that the main image must be free of overlays. Sale badges, discount flashes and text burned into the picture are not allowed on the main image, and apparel stores add them more than anyone because promotional imagery is part of how the category sells.
The third is composition. A garment shot at a distance in a busy scene reads as a blur in a grid. Fill the frame with the product, keep the background plain, and put the atmospheric photography in the additional images where it still gets used.
And keep availability honest per variant. Nothing wastes budget faster than paying for a click on the one size that sold out this morning.
The seasonal habit apparel needs and others do not
Fashion catalogues change faster than the systems around them, and the feed problems in this category are mostly maintenance problems rather than setup problems.
New season arrivals are the moment things break. Products are created quickly, attributes are inherited from a template, and a whole drop enters the feed with the wrong gender, a missing size system or an internal style code as its title. The fix is a checklist applied at creation rather than an audit applied later, because by the time you audit, you have already run the launch.
End of season is the mirror image. Sizes sell out unevenly, and a style that is technically still in stock in one size keeps spending budget while converting badly. Watch conversion by variant, not by style, and let the thin ones drop out.
Sales periods create price mismatches, since promotional pricing in apparel is frequent and rarely applies evenly. Automatic item updates and a shorter feed schedule matter more here than in a stable catalogue.
And returns feed back into all of it. Where a style returns heavily for fit, the fix is in the size guidance on the page and in the feed's fit attributes, not in the ad account.