Facet Order in Filtered Search: The Filters You Show First Decide What Sells
A shopper lands on a search results page, glances at the filter rail, and sees the facets in the order the database happened to store them: colour, then a technical specification most buyers do not understand, then price near the bottom. They do what shoppers do. They use what is visible, ignore what is not, and conclude the store does not stock the version they wanted. Facet order is merchandising, and most stores leave it to an accident of implementation.
Filters are a conversation in order
Facets answer the shopper's narrowing questions in the sequence the shopper asks them. For clothing that sequence is usually category, size, colour, then price and brand. For parts and tools it is compatibility first, before anything else is worth discussing. The order differs by category, which is precisely why a single global facet order, set once by a developer, is wrong somewhere in the catalogue at all times.
Work the order out from the click data per category: which facet gets used first, which combination of two facets precedes most purchases, where shoppers use a facet and immediately remove it, the signature of a facet whose values confuse or disappoint. That last signal is gold. A facet people apply and instantly undo is labelled wrong, valued wrong, or placed where a different question was expected.
Counts, dead ends, and honest numbers
Every facet value should carry a live count, and values that would return zero results in the current filter state should be greyed or hidden, never left clickable into a dead end. A shopper who narrows twice and hits an empty page blames the store's stock, not the interface, and they are right to: the interface offered a door it knew opened onto a wall.
Counts also discipline the facet list itself. A facet where one value holds ninety eight percent of products is not a filter, it is a fact, and it wastes a slot a working facet could use. Retire low variance facets from the rail and let the page carry fewer, better questions. Five facets that each split the set meaningfully beat twelve that mostly decorate it.
Selected filters must be visible and removable in place
The applied filter state belongs at the top of the results, as removable chips, in the order applied, with a clear all that is genuinely one click. Burying the state in the rail, forcing the shopper to hunt for which checkbox they ticked, is how sessions end in a full reset and a lost thread. The shopper should always be able to answer, at a glance, what the store thinks they asked for.
That visibility protects ranking too. Many relevance complaints are actually filter complaints: a forgotten size filter from three pages ago silently narrowing every later search. Surfacing the active filters on every results page turns a mysterious relevance failure into a one click self correction.
Reorder by evidence, then leave it alone
Facet order changes are cheap to ship and easy to overdo. Test a new order on one category, measure filter use, dead end rate, and conversion to product view, then roll it out or roll it back on that evidence. Once an order is working, resist weekly tinkering. Shoppers build muscle memory for a store's rails the way they do for a supermarket's aisles, and a store that rearranges its filters constantly feels unreliable even when every individual change tested well.
Assign the rail an owner per category, reviewed when the range changes materially. New product lines import new deciding attributes, and the facet that matters most for the new range will not promote itself.
Keep reading
Related guides on this site: Ranking When Stock Runs Out: Search Results for Out-of-Stock Items Voice Search on Storefronts: How Spoken Queries Change Ranking Practical answers about personalization.
Frequently asked questions
What is a facet in site search?
A filter dimension offered on a results page, such as size, colour, price, or compatibility, with values a shopper can select to narrow the set. Facets are the main way shoppers refine a search without typing a new query.
In what order should facets appear?
In the order shoppers in that category actually narrow their choice, derived from facet use data: which is used first, which pairs precede purchases, and which get applied and immediately removed. The right order differs by category, so one global order is wrong somewhere.
Should facet values with zero results be shown?
No. They should be hidden or visibly disabled in the current filter state. A clickable value that returns an empty page reads to the shopper as missing stock and breaks trust in the filters that remain.
How many facets should a results page show?
Fewer, better ones. A facet only earns its slot if its values split the current results meaningfully. Five working facets outperform twelve decorative ones, and low variance facets should be retired from the rail.