Key takeaways
- Categories rank for broad, high-volume searches that product pages cannot capture.
- The structure should reflect demand: each indexable category answers a group of searches with the same intent.
- Filters generate thousands of URLs; only combinations with their own demand should be indexed.
- Any structural change requires a redirect plan to avoid losing rankings.
Why does eCommerce category architecture matter so much for SEO?
eCommerce category architecture matters because category pages answer the broad, highest-volume searches, such as "women's running shoes". People searching that way want to compare a selection of products. The structure determines which searches the shop can capture and how much internal authority each page receives.
Categories a few clicks from the home page and linked from the menu are crawled more often and carry more weight than those buried in deeper levels. In large catalogues, that difference determines which pages Google visits frequently and which it takes weeks to revisit.
A common mistake in approach is copying the structure from the ERP or the supplier. Internal product families, designed for stock management, rarely match the words customers use when they search.
How to design category architecture based on demand
The process starts with searches and ends with the catalogue. Reversing the order produces categories nobody searches for and leaves groups of high-volume searches without a page.
- Gather every search in the sector covering product types, attributes, uses, brands and variants, along with their volume and seasonality.
- Group the searches by intent. If two queries return almost the same results in Google, they belong to the same page.
- Arrange the groups into categories, subcategories and attribute combinations, with no more than three levels of depth.
- Check each group against the catalogue, because a category with only two or three products adds little and is better merged into another.
- Validate margin, seasonality and commercial strategy with the business to decide which categories take priority.
A practical example: a running gear shop
A specialist shop receives very different searches around the same type of product. The table shows how each one is assigned to a page according to its intent.
| Search | Intent | Page that answers it |
|---|---|---|
| women's running shoes | Selection by audience | "Women" subcategory within Running shoes |
| waterproof trail shoes | Selection by an attribute with demand | Indexable filter page |
| asics running shoes | Selection by brand | Brand within the category |
| best shoes for a marathon | Informational and comparative | Buying guide linked to the category |
| asics gel nimbus | Specific product | Product page |
The underlying rule is one URL per intent. When two rows in the table end up on the same page, one of the searches is poorly served and the other competes against itself.
A shop gains rankings when it organises its products around how customers search, and that counts for more than having a bigger catalogue.
Juan Berges, CEO of The Baller Company
What should you do with filters and faceted navigation?
Filters should only be indexed when the combination has its own search demand and enough products. Everything else, such as sizes, colours without volume or sort orders, stays out of the index. When those combinations add up to thousands of URLs, it is also worth keeping them out of the crawl so they do not compete with the main categories.
Size, colour, price, brand or material: every filter combination generates a different URL. A shop with a few hundred products can produce hundreds of thousands of URLs, almost always without demand, which use up crawl budget and dilute the signals of the categories.
| Filter type | Example | Recommended treatment |
|---|---|---|
| Attribute with its own demand | Waterproof trail shoes | Indexable page, with content and internal links |
| Brand within a category | Asics running shoes | Indexable if there is search volume |
| Attribute without demand | Size 42, grey | Not indexable and kept out of the crawl |
| Sort order | Price, low to high | Not indexable, with a canonical to the category |
Google's documentation on faceted navigation recommends preventing the crawling of filter combinations that should not appear in search results, for example with robots.txt. For the URLs you do want indexed, it asks for a consistent parameter order.
Noindex, canonical or blocking: which to choose
| Method | What it does | When it fits |
|---|---|---|
| Noindex tag | Google crawls the URL but does not show it in search results | A small number of combinations that should be removed from the index |
| Canonical | Indicates the preferred version, although Google may ignore it | Sort orders and variants almost identical to the category |
| Blocking in robots.txt | Prevents crawling, but does not remove what is already indexed | Large-scale filter combinations without demand |
| Filters without crawlable links | The filter is applied without generating links that Google follows | New shops or redesigns that decide this from the outset |
A mistake we often come across is blocking URLs in robots.txt that carry a noindex tag. Google cannot crawl them, so it never sees the tag. URLs that were already indexed can keep appearing for months.
What content does a category page need?
Above all, a category page needs a product listing visible on the first screen. Around it sit a title aligned with the group's main search, a short introduction, links to subcategories and a block that helps with the buying decision at the bottom. The text supports the listing without pushing it aside.
- Title and main heading aligned with the group's main search.
- An introduction of two or three sentences about the products, without pushing the listing down.
- A buying guide or FAQs at the bottom, with information that helps people decide.
- Links to subcategories and related categories.
- Breadcrumbs with their structured data markup.
- Price, availability and ratings visible in the listing.
The text at the bottom of the page works well when it answers specific buyer questions: how to choose the right size, what sets two types of product apart or what maintenance they need. Generic blocks written to repeat the keyword add nothing. To write them with editorial judgement and in the brand's voice, the work draws on content and brand.
Breadcrumbs serve two purposes. They help Google understand the hierarchy, and they appear in search results when they carry the right markup; the details are in the guide to structured data for AI search engines.
The speed of the listing matters too. Image-heavy categories are often the heaviest pages in the shop, and their effect on sales is covered in the relationship between Core Web Vitals and conversion.
Well-crafted buying guides bring an added benefit. They are the kind of content AI assistants cite when someone asks for advice on choosing a product, a topic covered in SEO for AI search engines.
How do you detect and fix cannibalisation?
Cannibalisation is detected in Search Console by filtering for a query and checking how many pages receive impressions for it. If two or more URLs take turns in the results, they are competing with each other. It is fixed by merging the pages, redirecting the secondary one or clearly differentiating the intent of each.
It mainly occurs in three situations: a category and an indexable filter with the same focus, two categories with similar names, or a category and a blog article. Google alternates between them and neither consolidates its ranking.
The choice of solution depends on the data. If one of the pages attracts almost all the clicks and links, the other is redirected to it. If both receive traffic for different queries, titles, headings and internal links are adjusted to separate the intent. The blog article is refocused on the informational search and links to the category.
How to change the architecture without losing rankings
Reorganising categories in a shop that already ranks amounts to a partial migration. URLs change, internal links break and Google has to learn the structure again. With an orderly plan, the temporary dip is usually small and short-lived.
- An inventory of all current URLs with their traffic, links and rankings.
- A 301 redirect map from each old URL to its closest equivalent.
- Updated internal links, sitemap and breadcrumbs.
- Testing in a staging environment before launch.
- Daily monitoring of crawl errors and rankings during the first few weeks.
Timing matters as much as technique. Launching a new structure just before the Christmas campaign or the sales leaves no room to fix errors. On custom-built platforms, the redirects and the category template are prepared with the web development team weeks before going live.
Common mistakes in category architecture
Some failures recur in shops across very different sectors. None of them is complicated to fix, but all of them cost traffic for as long as they remain in place.
Seasonal categories that disappear
A Black Friday category is created with a new URL every year and deleted once the campaign ends. The links it earned are lost and the page starts from scratch. A permanent URL, updated each season, builds up authority year after year.
The same product with several URLs
When the product URL includes the category path, an item listed in two categories generates two product pages. The solution is a single product URL, independent of the path used to reach it.
Pagination with a canonical to the first page
If every page of a listing points its canonical to the first page, Google may stop crawling the products on subsequent pages. Each paginated page should carry its own canonical.
Empty categories
An out-of-stock category that displays an empty listing behaves like a page with no content. If the gap is temporary, keep it live with alternative products; if it is permanent, redirect it to the parent category.
Where to start
The most cost-effective first step is to compare the sector's searches with the current categories before changing anything. Within a few days, that comparison shows which demand groups have no page, which pages compete with each other and which filters consume crawl budget without bringing in traffic. With that list prioritised by volume and margin, the change can be tackled in phases. It is the starting point of our work on SEO for online shops.
Frequently asked questions
How many levels deep should an online shop be?
As a guideline, no more than three levels between the home page and the product listing. Greater depth reduces how often the final pages are crawled and makes navigation harder. If the catalogue demands it, indexable filters let you cover specific searches without adding levels.
Is it better to use parameters or clean URLs for filters?
For filters you want to rank, a clean, stable URL is easier to link to and to control. Filters without demand can work with parameters, provided they follow a fixed order and are kept out of the crawl when they generate too many combinations.
How much text does a category page need?
There is no right figure. A short introduction to the listing and a block at the bottom that answers genuine buying questions are enough. Long, generic text does not improve rankings and can make the mobile experience worse.
How long does it take for architecture changes to show results?
It depends on the size of the shop and how often Google crawls it. The first movements are usually visible within a few weeks, while rankings can take several months to settle. It is worth measuring traffic and conversion by category through analytics and CRO.
