A kitchenware store has a collection for every way its owner ever thought to slice the range: by brand, by colour, by material, by "new in", by "staff picks". Most hold a handful of products that also appear in several other collections, and Google has quietly stopped indexing many of them. Meanwhile "cast iron skillets", which buyers search for by name, has no page at all. A collection earns its place when a distinct search asks for it and you have enough products to answer; everything else is a thin page that costs crawl attention and ranks for nothing.
A collection is a page that answers one category search
A collection has to be found, understood as distinct and stored in the index before it can rank, and the second step is where store collections fail. A collection holding the same products as its parent, or a couple of items with no text explaining why they belong together, looks like a version of another page. Google's guidance on consolidating duplicate URLs describes what happens to versions: it picks one to represent the set and folds the rest away. There's no penalty, but the folded page is not going to rank.
The flip side is the test for a good collection: a search names a type of product, your page holds a real range of that type with a title, heading and a few lines of text that say so, and Google has a distinct page to match, one that answers a search nothing else on your site answers.
Three tests a new collection has to pass
Run every candidate through these in order, and stop at the first failure.
- A distinct search asks for it. Someone types this phrase, it carries its own demand in your market, and it isn't a synonym of a collection you already have. "Cast iron skillets" passes; "skillets made of cast iron" is the same search in Google's eyes and fails.
- You can fill it. Enough products genuinely fit that a shopper sees a range to choose from, and that set isn't the parent collection's under a new name. Same products, different title, is a duplicate, not a collection.
- The searcher wants a range, not an answer. If the results page is product listings and collection pages from other stores, a collection is the right page. If it's guides and comparisons, the demand is real but belongs to a buying guide, and a collection would lose.
A candidate that passes all three gets a URL, a place in the navigation and a few sentences at the top saying what the range is and who it's for. One that fails the first test becomes a filter or disappears; the second, waits until the stock exists; the third, goes to the content list.
Find the candidates by clustering the keyword list
Many lines in your category list from the first lesson are the same search in different words. Clustering groups them by meaning, and each group is a page candidate: those that match an existing collection tell you which phrasing belongs in its title, and those with no match are this lesson's candidates.
The rule that makes clustering worth doing is one group, one page. Two collections targeting the same group split the signals that could have made one rank, which is called cannibalisation. Mapping keywords to pages covers the method; for a store the groups resolve into collections, product pages or guides.
Attributes that deserve a page, and ones that should stay filters
Every attribute in your catalogue could become a collection, and most shouldn't. The question is whether shoppers search by that attribute before they arrive.
Use case, audience and material often pass: "non-stick frying pans", "kids' lunch boxes" and "cast iron skillets" are searched for by name and expect a range. Price bands sometimes pass, when a "budget" or price-capped phrasing has demand of its own. Gift and occasion collections pass when the demand recurs each year; keep the URL the same every year so the page keeps what it has earned.
Size and colour usually fail. Shoppers filter by them on the page rather than searching for them, so "red frying pan" is a filter on the frying pan collection, not a page. Sort orders, "new in" and "staff picks" fail the first test outright: nobody searches for your staff's opinions. Faceted navigation and filters covers keeping those filter URLs out of the index.
What to do with collections that fail
Existing collections go through the same three tests. One with no search of its own should be merged into the collection that has the search, with a redirect so its links carry over. One holding the same products as its parent should be redirected to the parent or pointed at it with a canonical tag, which Variants and duplicate content covers. Keep any collection shoppers use from the navigation, even if it can't rank, but stop expecting search traffic from it.
The result is a shorter list of collections, each answering a search with enough products to satisfy it: a store Google can understand, and the shape the next chapter builds on.
What to take away
- A collection earns its place when a distinct search asks for it, you can fill it with a real range, and the results page shows shoppers want a range rather than a guide.
- Google folds near-duplicate collections into one and ignores the rest, without a penalty but also without a ranking.
- Cluster the category list first: one group, one page, and each group with no matching collection is a candidate.
- Use case, audience and material often deserve collections; size, colour and sort orders stay filters.
Next
With the right collections decided, arrange them so every product is close to the homepage in Collection structure and navigation.