Faceted Navigation and Search Relevance: Filters That Help Instead of Hiding Stock
Facets work when they reflect how shoppers decide, show honest counts, and never hide in stock products behind contradictory defaults. Order facets by decision importance, keep counts live as filters combine, and make sure search ranking still applies inside filtered views instead of reverting to arbitrary order.
Order facets by the decision, not the database
Shoppers filter by the attribute that decides the purchase first: size for clothing, compatibility for parts, capacity for storage. Systems that order facets by database field order or alphabetically force shoppers to hunt for the one control they need. Derive facet order from click data and purchase paths, per category, because the deciding attribute changes by category.
For more on this, see this related guide and our earlier explainer on the same theme.
Limit visible facets to the handful that matter and fold the rest under a more filters control. A wall of twenty filters reads as work. The categories with the best conversion show restraint up front and depth on request.
Counts must tell the truth live
Facet counts that do not update as filters combine teach shoppers the interface lies. Every count should reflect the current selection, including the zero cases, which should be shown disabled rather than silently omitted, so shoppers understand why a combination is empty instead of suspecting a bug.
Zero result combinations deserve a recovery path: show which filter to remove to restore results, and keep the removed option one click away. Shoppers forgive empty results when the interface explains them. They abandon when the page simply goes blank.
Ranking still applies inside filters
Filtering narrows the set. It should not flatten the order. Relevance, popularity, and margin aware ranking still need to sort the filtered results, or your best products sit wherever the catalogue order leaves them. Test filtered views explicitly, because many platforms quietly fall back to name or SKU order there.
Readers also found this companion piece from our archive useful.
Watch the interaction with search terms. A shopper who searched and then filtered expects the search intent to keep influencing order. Dropping the query context when a facet is clicked is a common platform bug that measurably hurts conversion and is rarely in anyone's test plan.
Facets and SEO need one policy
Every facet combination can generate a URL, and uncontrolled, that is an index bloat machine. Decide which combinations are indexable, usually the few with real search demand, canonicalise or noindex the rest, and keep the policy consistent between the storefront and the sitemap.
Do not let the SEO policy break the shopping experience. Shoppers should be able to combine filters freely even where the resulting URL is not indexable. Indexing is a publishing decision. Filtering is a customer tool. Conflating them serves neither.
Frequently asked questions
How many facets should a category page show?
Typically five to eight primary facets visible, with the rest available under an expansion control. The right number comes from your shoppers' decision order in that category. If a facet is rarely used and rarely decides a purchase, fold it away.
Why do some filters show products that are out of stock?
Usually because availability is not part of the count query or the default sort. In stock items should rank first and counts should reflect availability settings honestly. Showing appealing counts that resolve to unavailable products is a trust error.
Should filtered pages be indexed by search engines?
Only combinations with genuine external search demand, such as a popular brand plus category. Everything else should be canonicalised to the parent category or set to noindex, to avoid diluting the site with thin near duplicate pages.
What is the biggest facet mistake you see?
Counts that do not update as filters combine. Shoppers build a mental model from counts, act on it, and get contradictory results. Once the interface is caught lying once, shoppers stop trusting every number on the page.