Discussion starter: an ecommerce SEO question, posted with an answer from Niraj Raut to open the thread. If you have dealt with this on a site, reply with what you saw, especially where it differs.
- eCommerce SEO
Deep taxonomy with narrow categories, or broad categories with filters: which served long-tail demand better in your store?
Disclosure: Niraj Raut, who posted this answer, runs the SEO consultancy linked at the end of it.
1 reply
-
adminAdmin
Neither structure wins on its own. What serves long-tail demand is a deliberate decision about which combinations deserve an indexable landing page, and both structures can produce those pages. A deep taxonomy creates them as real categories; a broad structure with filters creates them by promoting selected filter combinations to indexable pages. The deep taxonomy tends to do better when long-tail concepts are stable and have enough products to stand as categories. Broad categories with curated, indexable filter pages tend to do better when demand is spread across many attribute combinations and inventory shifts. The failure modes are symmetrical: too many thin categories on one side, uncontrolled filter URLs on the other.
What each structure is good at
Deep taxonomy, narrow categories Broad categories with filters Long-tail landing pages Each node is a proper page with its own title, heading, copy and internal links Only the filter combinations you choose to make indexable User navigation Can become a maze of near-identical choices Faster for shoppers who know what they want Crawl control Predictable URL set Needs active control, or filters generate huge numbers of crawlable URLs Thin page risk High for nodes with few products High only if you index combinations without checking inventory Maintenance Every category needs merchandising and copy Landing pages can be generated from attributes, with light curation Cannibalisation Sibling categories often overlap Filter pages can overlap with categories unless you choose one owner per concept The test I use for each long-tail concept
For every candidate, whether it would become a category or an indexable filter page, I ask four questions:
- Is there demand for it as a concept? “Waterproof hiking boots for women” is something people search. “Hiking boots, size 7, brown, under £120” is a filter state, not a concept.
- Is there enough inventory to make the page useful, most of the time? A page that shows three products half the year is thin; one that shows forty is a real destination.
- Is it distinct from its parent and siblings? If the product set is almost identical to another page, it will compete rather than add coverage.
- Can we give it something beyond a product grid? A useful heading, a short buying note, relevant links. Not long copy, but enough to show the page exists for a reason.
Concepts that pass all four deserve an indexable page. Whether it lives in the taxonomy or as a promoted filter page is then a question of how stable the concept is and how the shop wants to merchandise it.
Making filter pages indexable without creating a crawl trap
Google’s documentation on faceted navigation is direct about the risk: filter combinations can create a near-infinite URL space that wastes crawling. It suggests keeping faceted URLs you do not want indexed out of crawling, for example with robots.txt rules or by not exposing them as crawlable links, and, for facets you do want indexed, using a consistent URL structure, a stable parameter order and a 404 when a combination returns no results. In practice that means:
- An explicit allowlist of indexable combinations, generated from demand and inventory data and reviewed periodically.
- Clean, stable URLs for those pages, a self-referencing canonical, a unique title and heading, and a place in internal linking, such as “popular” links on the parent category or a related-searches block.
- Everything else kept out of the crawlable link graph or blocked, and never listed in sitemaps.
- Automatic handling when inventory drops: noindex or remove a promoted combination when it falls below the product threshold for a sustained period, rather than letting it become thin.
Watch for this:
Using canonical tags on every filter URL to point back to the parent, while leaving them all crawlable, does not solve the crawl problem. Google still has to fetch the URLs to see the canonical, and canonicals are hints that can be ignored when the content differs. Control crawling for combinations you never want indexed, and reserve canonicals for genuine duplicates.
When a deep taxonomy is the better choice
- The long-tail concepts are how customers think about the range, and they change slowly: specialist retailers with established product types, parts catalogues with fitment categories, B2B catalogues organised by application.
- Merchandising teams want to curate each concept with its own content and promotions.
- The platform handles filter URLs poorly and cannot produce clean indexable filter pages.
When broad categories with filters are the better choice
- Demand is spread across many attribute combinations: colour, material, size, brand, occasion.
- Inventory changes often, so fixed categories would constantly be empty or overflowing.
- The shop has good attribute data and can generate landing pages from it with controlled quality.
Hypothetical example:
A homewares store has 12,000 products and a six-level taxonomy with 1,400 categories. A quarter of the categories hold fewer than five products, and many siblings overlap, such as “Linen Cushions” and “Natural Cushions”. The team flattens the taxonomy to about 200 categories, keeps filters for attributes, and promotes 350 filter combinations with proven demand and at least twenty products to indexable pages linked from their parents. Long-tail impressions to category-type pages are the metric to watch over the next few months, alongside conversion from those pages and the number of indexed URLs, which should fall.
How to measure which setup serves the long tail better
Track the number of distinct non-brand queries that generate impressions for category and landing pages, and the clicks and revenue from them, grouped by page type. Compare the share of indexable landing pages that receive any impressions over a quarter; a large share with none suggests you have created pages without demand. Watch Crawl stats and logs for crawling of non-indexable filter URLs after any change. If long-tail coverage grows while crawl waste falls, the structure is working, whichever label you give it.
My default for most mid-size and large stores: broad, shopper-friendly categories, a curated set of indexable filter pages chosen by demand and inventory, and tight control over everything else. Deep taxonomies are right where the long tail is made of stable concepts that merchandisers genuinely curate.
Need help with this on your own site?
Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.
Add a reply Cancel reply
Replies are open to approved members. Log in or apply to join.