Search facets are a merchandising surface: clean them up and shoppers find the right stock

Audit the facet list the way a shopper meets it

Run your top fifty searches, open the facet panel on each, and write down every value that is a duplicate, a typo, a format variant, or a dead end with one product behind it. On the last store where we did this, size appeared as S, Small, and a numeric range from a legacy feed, and material carried both cotton and 100 percent cotton as separate filters. Each split quietly hid stock from the shoppers most ready to buy, the ones filtering with intent.

Count products per facet value while you are there. Values with a handful of products are usually feed errors, not range decisions. Values with hundreds deserve the top positions. Facet order should follow shopper use and stock depth, not alphabetical accident or the order an import created them in 2021.

Normalise at the feed, not in the theme

Fixing facet labels in the theme is paint over rust. The next import restores the mess. Build a mapping table in the feed or the PIM: supplier colour names to your colour family, size strings to your size scale, material descriptions to your controlled list, and enforce it on every import. New values that are not on the list should queue for review instead of publishing themselves. That one gate stops the problem regrowing.

Decide the granularity shoppers actually think in. Colour family for filtering, exact shade name on the product page. One store split denim washes into eleven filter values and wondered why filtering felt like homework. Shoppers filter to narrow a set they then browse. If the filter does the browsing too, it has gone too far.

Pin, hide, and sequence by collection

The facet that matters changes by collection. Size and fit lead in apparel, compatibility leads in parts and accessories, price leads in a sale collection where the whole point is the discount band. Set facet order and visibility per collection template where your platform allows it, and review the top collections quarterly. A global facet order is a compromise nobody chose.

Hide facets with no discriminating power on that collection. If everything in the set is the same material, the material filter is furniture. Removing it shortens the panel and raises use of the filters that do work. Measure filter use before and after. On most stores, cleaning the panel lifts filter engagement more than adding a smarter search box does, at a fraction of the cost.

Every facet combination that returns zero is a shopper told the store has nothing, when it usually means the combination was allowed carelessly. Prevent impossible combinations from being selectable where the tooling allows, and where a zero still happens, answer with the nearest stocked alternatives and say what was relaxed. Last quarter we shipped thirty one changes on one store. The facet cleanup was the one the owner could see in the numbers by the following month, in search conversion, not in applause.