Operations
Ecommerce category page SEO: filters, faceted navigation, and crawl budget
Ecommerce category page SEO under a real limit: one person, no new tools, and faceted navigation that eats crawl budget. What to drop, what to keep.
What to take away
- One operator at six to eight hours a week can run ecommerce category page SEO, but only by indexing a chosen set of facet values.
- Full indexation of every filter combination is ruled out. It needs engineering time this setup does not have.
- Server-rendered links for facets with demand, plus blocking for the rest, is the workable middle.
- The workaround stops paying when parameter URLs hold a quarter or more of indexed pages for two straight months.
- Past that point, fund an engineering fix instead of adding hours to one person's week.
The limit in measurable terms
One operator. Six to eight hours a week, no new software spend, no developer sprint and no agency retainer.
The catalog holds an illustrative 4,000 to 12,000 SKUs across 30 to 90 categories, with filters for size, color, material, price band and availability. That grid resolves into hundreds of thousands of URLs, and almost none of them carry search demand of their own.
A search engine crawler spends a limited number of requests on any one host. Faceted navigation multiplies the addressable set faster than one person can curate it.
What this rules out
A dedicated, indexable landing page for every filter combination is out. So is real-time index control through the Indexing API, which Google limits to job posting and livestream markup.
You will not outrank a large marketplace for a broad head term such as running shoes. That is a loss, and it points to the same answer: index fewer URLs, and make the ones you keep earn their place.
What still works on one desk
The category page itself still works, and it should carry the head term plus two or three close variants. Hand-written copy above and below the product grid stays one of the few on-page levers you control. The routine is set out in on-page SEO.
Single-value facets with demand are the second piece. Blue running shoes, or size 10 boots: one page each, server rendered, in the sitemap, with a short paragraph of original text. Ten to thirty such pages per large category is realistic.
Blocking is the third. A short robots.txt pattern set removes sort orders, price bands and view modes from the crawl path in an afternoon.
| Facet type | Handling here | Indexed |
|---|---|---|
| Single value with demand (color, size) | Server-rendered link, own copy | Yes |
| Two values combined | Canonical to the leading value | No |
| Sort, price band, view mode | robots.txt disallow | No |
The monthly pass
Set aside about two hours. The crawl-side work that matters most is covered in technical SEO.
- Open Crawl Stats and group requests by directory and by parameter.
- Sort each parameter into keep, canonical or block.
- Ship the robots.txt and canonical edits.
- Re-export in 30 days and compare parameter requests as a share of the total.
Then run the checks:
- Filter links resolve to crawlable href values, not script handlers.
- Every kept facet page carries at least 80 words of unique copy.
- Blocked parameter URLs are not linked from category pages.
- Parameter crawl requests fell month over month.
Compromises worth making
Part of the catalog gets a curated filter page and the rest does not: 10 to 30 single-value pages per large category beats thin copy spread across 600 URLs.
Prefer a canonical over a noindex tag on combination pages. A web crawler has to visit a URL before a noindex takes effect, and that visit is the one you are trying to avoid.
Keep the sitemap to canonical URLs only, and use crawl data you already pay for rather than buying another tool.
Compromises that are not worth making
Do not block a filter that converts. If a size or color filter is the last click before checkout on a large share of orders, blocking it trades revenue for a cleaner number. Check the funnel in Analytics for SEO before you touch robots.txt.
Do not hide filter links behind JavaScript with no href. The URLs still exist, and shoppers on slow connections lose the filter entirely.
A full rebuild is not the answer at this stage either. It suits a large catalog, and it is not a six-hour-a-week project.
When the workaround stops paying
Two triggers end the manual approach. Parameter URLs hold 25 percent or more of indexed pages for two straight monthly exports. Crawl requests to parameter URLs exceed 30 percent of all crawl requests across the same two months.
Either one means the pass is falling behind. At that scale the fix is permanent: server-side facet link generation, a controlled index set, and checks in the deployment pipeline. Enterprise SEO covers what to keep once you fund it.
Below those numbers, extra hours will not buy much.
Common questions
How many facet combinations should I let Google index? Only the single values that match demand. For most catalogs that is 10 to 30 pages per large category, each with its own copy.
Can robots.txt alone fix crawl waste? No. It stops crawling, but it also keeps canonical signals from reaching Google. Use it for sort and price parameters, and canonicals for combinations.
When does the one-person approach stop working? When parameter URLs pass a quarter of indexed pages, or a third of crawl requests, for two months running. That is the signal to fund engineering, not to add hours.
Should I block filters that drive conversions? No. Check which filters appear on the path to checkout first. Blocking a converting filter costs orders, a worse trade than some wasted crawl.
