Faceted search best practices for ecommerce

A shopper lands on your jackets page and sees 412 products. She knows she wants something waterproof, knee length, in a medium, and under $250. If your filters can get her from 412 to six in three taps, you've got a sale in progress. If they can't, she's going back to the search results on Google, and the next store gets its turn.

Faceted search is the set of filters that does that narrowing, and most stores still get it wrong. Baymard Institute's 2025 product list benchmark found 58% of desktop sites and 78% of mobile sites have mediocre or worse product list and filtering UX. That gap costs real money: in Baymard's usability testing, sites with mediocre product list usability saw 67–90% of shoppers abandon the task, while sites with a slightly better toolset saw 17–33%.

This guide covers how to choose, order, and maintain facets, how to keep them working on mobile, what to do when a filter combination comes up empty, how to keep filtered pages from hurting SEO, how to measure whether facets are working, and what to do for the shoppers facets can't help.

What faceted search is and how it differs from filters

Faceted search lets shoppers narrow a product list by combining attributes like size, color, price, brand, and material, and the available options update after every selection. Pick "waterproof" and the color list shrinks to the colors that come in waterproof styles. That last part is what separates facets from plain filters. A plain filter removes products that don't match. A facet also tells the shopper what's still possible.

You'll see the same pattern called faceted navigation, faceted filtering, or guided navigation. Category navigation is a different thing: it moves shoppers down a fixed tree like Women, then Outerwear, then Jackets. Faceted search starts where the category tree stops and lets shoppers cut the list along whatever dimension they care about.

Site search and faceted search work together but solve different problems. The search bar has to understand what a shopper typed. Facets have to help her narrow what came back, and that second job is what this post covers.

Choosing which facets actually help shoppers narrow down

The fastest way to build bad facets is to expose every attribute in your product feed. What shoppers need is the five or six filters that decide whether a product is right for them.

Start with the attributes that decide the purchase

For every category, ask what makes a shopper rule a product in or out. For apparel, that's usually size, fit, color, and price. For a sofa, it's width, depth, fabric, and whether it'll fit through the door. For headphones, it's battery life, noise canceling, and wired or wireless. Those decision attributes go first, and they should never be missing.

A useful test: if an attribute shows up on the product card, shoppers will want to filter by it. Baymard found 38% of sites don't offer filters for every attribute shown in the product listing, which leaves shoppers scanning cards one by one for a detail the site already knows.

Make facets category specific

A single sitewide filter template is the most common shortcut and the most expensive one. Outerwear needs length, warmth rating, waterproofing, and hood style, boots need shaft height and width, and skin care needs skin type and concern. If the boots page shows the same facets as the scarves page, the shopper does the filtering in her head and usually gives up.

Let your own data decide what stays

Look at two sources before adding or removing a facet. Your site search queries show the words shoppers use, and if "wide calf" keeps showing up in boot searches, that's a facet you're missing. Your facet usage data shows which filters people touch. A facet that almost nobody uses in a category can move below the fold or go away.

Ordering and grouping facets for scannability

Shoppers scan the filter panel the way they scan a menu. Put the most used facets at the top, in the order shoppers reach for them. Alphabetical order and database order both feel tidy to the people who built the site and random to everyone else.

Order values the way shoppers think

Inside each facet, order matters too. Sizes should run small to large, and numeric values should sort as numbers so 2-pack comes before 10-pack. Brands can go alphabetical. Colors usually work best grouped into primary color families, so "navy," "midnight," and "ink" all show up under blue and don't split the results three ways.

Write labels shoppers recognize

Labels deserve the same attention as order. Baymard's research shows 25% of desktop sites and 40% of mobile sites use unclear filter labels that shoppers skip. Short and literal wins, especially on a phone: "Waterproof" beats "Weather protection rating."

Get the multi-select logic right

Shoppers expect to pick more than one value in the same facet, like a medium and a large, and see both. Baymard reports 14% of sites still don't allow multiple selections within a filter, and shoppers abandon products because of it. The standard logic is OR inside a facet and AND across facets. Medium OR large, AND black, AND under $100.

Handle long lists without hiding them

A brand facet with 80 entries needs help. Show the top handful, add a "show more" link, and give long lists a small search box of their own. Keep counts next to every value so shoppers know what they'll get before they tap.

Promote the filters shoppers use first

On a long product list, the first thing most shoppers look for is a way to narrow it. Promoting your most used filters as tappable chips at the top of the list gives them a head start, yet according to Baymard, 61% of sites don't promote filters in the product list. On a sofa page, chips for sectional, loveseat, and sleeper get shoppers moving before they ever open the filter panel.

Keeping counts and availability honest

Shoppers trust facet counts to be right. If "size medium" says 42 results and the shopper finds three, she stops trusting the filters and goes back to scrolling.

Build counts from what's in stock

Two things break that trust most often. The first is counts that don't update after a selection, so they describe the whole category when the shopper is looking at a narrowed list. The second is stock. A jacket that comes in medium will match a "size medium" filter even when every medium is sold out. Build facet counts from in-stock variants so a filter only shows what someone can buy today. That one fix matters most in fashion and footwear, where size runs out first and shoppers get the most annoyed when they find out on the product page.

Fix the data underneath

Most facet problems start in the catalog data, not the interface. Standardize attribute values before they reach the catalog, normalize units so 75 cm and 0.75 m land in the same range, and set price ranges that fit what you sell. A $0 to $1,000 slider on a catalog where everything costs less than $60 doesn't help anyone.

Mobile faceted search patterns

Mobile is where filtering falls apart. The screen is small, the filter panel hides behind a button, and the moment a shopper applies a filter, the panel closes and takes her choices out of view.

A good mobile facet experience has a few consistent pieces. The filter button stays visible as the shopper scrolls. Tapping it opens a full-screen panel with room for every facet. The apply button shows a live result count, like "Show 34 items," so shoppers know what they'll get before they leave the panel. And applied filters stay on the results page as removable chips.

That last one matters on every device. Baymard found 20% of sites fail to keep applied filters visible while shoppers browse. On mobile it hurts even more, because a shopper who can't see her choices has no idea why the list looks the way it does.

Handling facet combinations that return zero results

Zero results from a filter combination is the worst dead end on your site, because it hits shoppers who've already told you exactly what they want. Someone who picked waterproof, size 10, and under $150 is close to buying. Sending them to an empty page right then is how you lose a sale you nearly had.

Start with prevention. Disable or hide values that would return nothing given the current selections, and keep the counts honest so shoppers can see a dead end coming. If waterproof boots never come in red, red shouldn't be clickable once waterproof is on.

When a dead end can't be avoided, the empty page still has a job. Tell the shopper which filter is doing the damage, offer a one-tap way to remove it, and show the closest matches, like the same boots in size 10.5 or the waterproof pair that's $20 over budget. A plain "no products found" message gives her nothing to do next. An empty page that comes from a search query is a different problem. That's search relevance, and the fixes, like synonyms and typo tolerance, live in the search engine itself.

Track your filter-driven zero-result rate as its own number. If it's climbing, the cause is usually data: a new attribute that's only half populated, or a color value somebody typed three different ways.

Faceted navigation and SEO

Every facet combination can create its own URL, and a catalog with a handful of facets can generate millions of them. Most of those pages are near duplicates with thin content, and crawlers will waste time on them that should go to your real category and product pages.

Google's guidance on faceted navigation gives you two paths. If you don't need filtered URLs in search results, stop them from being crawled with robots.txt rules or URL fragments. If some filtered pages deserve to rank, follow a few rules for those pages: use standard parameter formatting, keep filter order consistent so the same combination always produces the same URL, and return a 404 for combinations with no results.

Pick the combinations worth indexing based on real search demand. "Women's waterproof hiking boots" might have enough volume to justify its own indexable page with a unique title and intro copy. "Women's waterproof hiking boots, size 7.5, under $150, sorted by newest" doesn't. Canonical tags pointing back to the parent category help with the rest, though Google says canonicals are less effective over the long run than blocking the crawl.

Measuring facet usage and its impact on conversion

Once facets are live, treat them like any other part of your conversion rate optimization program. A handful of metrics will tell you whether they're working.

Facet engagement rate is the share of product list sessions where a shopper applies at least one filter. Low engagement on a big category usually means the facets are hidden, irrelevant, or badly labeled. Usage by facet tells you which filters earn their spot and which ones can go.

Conversion for filtered versus unfiltered sessions shows how much facets contribute to revenue, and filter-to-cart rate tells you whether narrowing leads to adding. Facet abandonment, meaning sessions where a shopper applies filters and then leaves without opening a single product, points to results that looked wrong after narrowing.

Watch one more number that most teams skip: how often shoppers apply a filter, remove it, and try another. A lot of that back-and-forth means your facets don't match how people think about the category, and it's a good signal for which categories need new attributes or a conversational option.

Review the setup every quarter. Catalogs change, new attributes get added half-filled, and seasonal categories need different facets in December than in June.

When facets run out, give shoppers a way to ask

Facets work when the shopper knows the vocabulary. They assume she already knows she needs a 600-fill down jacket, a 30-inch inseam, or a lens tint for low light. Plenty of shoppers don't. They know they're going to a rainy wedding in Portland, or they need goggles for flat light on spring snow, and no checkbox on your site says that.

Those shoppers tend to do one of two things. They over-filter, stacking guesses until the list hits zero, or they bounce between product pages comparing spec sheets they don't fully understand. Both are high-intent sessions, and both usually end without a purchase.

This is where conversational help belongs, right next to your facets. A shopper describes what she needs in her own words, and customer experience AI connected to your product catalog translates that into the attributes your facets use: rain-ready, dressy, knee length, in stock in her size. It can compare two options, explain why one runs warmer, and recommend something she wouldn't have found by filtering. When a question needs a person, a team member picks it up with the full conversation in front of them, so the shopper doesn't start over.

Gladly does this inside the same conversation brands already use for service, so product questions and post-purchase questions share one customer record. For one global luxury apparel brand, Gladly leads 80% of product conversations and resolves 48% of conversations outright, while 32% move to team members with full context.

The shopper who knows exactly what she wants will filter her way there in three taps, and the one who's unsure can ask, so the strongest stores offer both. The AI shopping assistant guide covers how to choose one that fits your catalog.