Performance Max on a big catalog: send traffic to category pages or product pages?

Take a fashion store with around 400 SKUs where Performance Max keeps pouring most of the spend into a handful of bestsellers through product pages. One fix is pointing the final URLs at category pages, which gives the algorithm more room to match something like "black linen shirt men" to the right item instead of defaulting to whatever SKU already converts.

I think category pages tend to work better on large catalogs, but only when each category has enough products and filters to convert on its own. A thin category with four items and no sorting just wastes the click. Product pages make more sense when the feed is small or margins vary a lot, because you know exactly what each click is buying. Curious how others have set final URLs once the catalog goes past a few hundred items.

I'm not convinced the URL is what drives the bestseller concentration. The final URL mostly decides where people land after the click. Performance Max picks which products to show from the feed, and listing groups control how the budget gets split between them. If all 400 SKUs sit in one listing group, the bestsellers will keep winning on raw conversion volume whether the final URL is a category page or a product page.

Splitting the feed with custom labels, by margin tier or price band for example, gives the campaign room to spend on the long tail. Category pages still help with the thin-category problem you mentioned, but I'd sort out the listing structure first and test URLs after. Otherwise you can't tell which change did the work.

How is your listing group tree set up right now, one group for everything or split by something?

Fair point on listing groups, but I'm not sure it holds for Performance Max, because the final URL still shapes what the algorithm learns from. A category page that matches the query gives people a relevant landing experience, and that tends to show up in conversion rate, not just in where they land. It depends on whether the category pages convert on their own, so that's the first thing to check before testing anything else.

Agreed that landing relevance shows up in conversion rate. One catch for the test is final URL expansion. If it's on, Google can send clicks to pages other than the ones you set, so a category test may partly be a product page test. Check the landing page report first.

Kai's right about expansion: turn it off for the test, otherwise you're comparing a mix of pages. For picking the URL in the first place, a simple rule works. Send traffic to a category page only if it would stand on its own as a landing page for the query. That means enough products to browse, working sort and filters, and a clear match to what people searched. If it doesn't pass, use the product page, even for a bestseller. A weak landing page loses the click before the algorithm learns anything useful, so you end up training on bounce data instead of on what actually sells.

Fair rule, but the default sort matters. If category pages load bestsellers first, the landing page just repeats the concentration you're trying to fix. What sort order do yours load with?