How to Design Faceted Filters for Ecommerce: A Step by Step Method From Research to Release

    November 19, 2026
    11 min read
    By Callan Whitmore
    Share this article

    A shopper lands on a category page with 1,240 hiking boots. She knows three things: she needs waterproof boots, her size is 39, and she does not want to spend more than €180. The filter panel on the left starts with Brand, then Material, then Collection, then Upper height. Size is the eleventh facet, below the fold. Price is a slider whose handles jump in €50 steps. She picks Waterproof, scrolls, gives up looking for size, and goes back to the search engine.

    Nothing in that panel was technically broken. Every facet came from real product data. The problem was design: the panel reflected how the catalogue is stored, not how people choose boots. This guide walks through how to design faceted filters for ecommerce from first research to release, with a procedure, a decision tree for which facets to show, a comparison of control types and a checklist for launch.

    Start from shopper questions, not the product database

    Facets work when they answer the questions a shopper already has in mind. The fastest way to find those questions is a short research loop before anyone opens a design tool.

    1. Export site search queries for the category from the last 90 days and group them by attribute: size, colour, use case, price, brand, feature.
    2. Pull filter usage from analytics, if filters already exist: which facets are opened, which values are selected, and how often a filter session ends in a product view.
    3. Read 30 to 50 customer service contacts and product reviews for the category, noting the words people use for attributes, such as "wide fit" rather than "last width".
    4. Run 5 to 8 short sessions where participants describe what they would look for before you show them the page.
    5. List every candidate facet with its data coverage: the share of products that actually have the attribute filled in.
    6. Drop or fix any facet with coverage below about 80%, because filtering on incomplete data silently hides products.
    7. Rank the remaining facets by how often each question appeared across all sources, and use that ranking as your first draft of facet order.

    The coverage step is the one teams skip and regret. If only 55% of boots have a "waterproof" value, selecting it removes many waterproof boots from the list, and the shopper never knows.

    Deciding which facets earn a place in each category

    Global filter sets, the same 12 facets on every category, are easy to build and poor to use. A frying pan does not need a size grid, and a jacket needs one urgently. Each category deserves its own short list. These questions help decide whether a candidate facet goes in.

    1. Did shoppers ask about this attribute in search queries, service contacts or interviews? If no, leave it out for now.
    2. Is the attribute filled in for at least 80% of products in the category? If no, fix the data first or leave it out.
    3. Does selecting a typical value cut the result set by at least a third? If no, it may be decoration rather than a real filter; consider a sort option or badge instead.
    4. Is it a hard requirement, such as size, compatibility or a dietary restriction? If yes, place it in the top three facets.
    5. Is it a preference, such as colour or style? If yes, show it below the hard requirements.
    6. Does the category have more than about 200 products? If no, consider showing only 3 or 4 facets, since browsing may be faster than filtering.

    For most categories this produces 5 to 8 facets. That is usually enough. When a category seems to need 15, it is often two categories that should be split.

    Choosing the right control for each facet

    The control should match the type of data and the way shoppers think about it. A slider for shoe size or a checkbox list for price usually signals that the control was chosen by default rather than by design.

    Facet type Best control Example Watch out for
    Multiple discrete values Checkbox list with counts Brand, material Show the top 6 to 8 values, then "Show all"
    Continuous numeric range Two input fields, optionally with a slider Price, weight, screen size Sliders alone are hard to set precisely on touch screens
    Visual attribute Swatches with text labels Colour, pattern Colour alone fails for colour blind users, always add the name
    Size systems Grid of buttons Shoe size 36 to 46, clothing XS to XXL Keep the standard size order, never sort sizes by popularity
    Yes or no attribute Single checkbox or toggle Waterproof, in stock, free delivery Label the positive case only
    Mutually exclusive choice Radio buttons Men, women, kids Use only when values truly cannot combine
    Rating Minimum threshold buttons 4 stars and up "3 stars" alone is ambiguous, say "3 and up"

    Two rules apply across all controls. First, labels should use shopper vocabulary from your research, not internal attribute names. Second, every value in a list should show the number of matching products, so people can see the effect of a choice before they make it.

    Order, defaults and how many values to show

    Facet order is the single cheapest improvement in most filter panels. Put hard requirements first: size, compatibility, gender or fit. Then price. Then the attributes that most often appear in search queries. Brand belongs high only in categories where shoppers really choose by brand, such as running shoes or electronics, and much lower in categories like kitchen textiles.

    On desktop, keep the first 3 or 4 facets expanded and collapse the rest, so the panel fits in roughly one screen height. Inside a facet, show 6 to 8 values by default. Order them by product count for brands and materials, by the natural scale for sizes, and alphabetically only when the list is long and shoppers search for a known name, as with car models.

    Very long value lists, such as 140 brands, need a search box inside the facet. A short "Show 32 more" link is better than a scrolling box nested inside a scrolling page, which is awkward with a mouse wheel and worse with a trackpad.

    Defaults matter too. If a store sells products that are frequently out of stock, an "In stock" filter enabled by default can reduce frustration, but it must be visible as an applied filter, or shoppers will think the catalogue is smaller than it is.

    Applied filters, counts and zero results

    Once a shopper has selected a few values, she needs to see what is applied and undo it in one step. A row of removable chips above the product grid does this well: each chip shows a value such as "Size 39" with a close icon, followed by a "Clear all" link. This row should appear on desktop and mobile alike.

    Counts next to values should update after every selection. When a value would lead to zero products, there are two reasonable options: show it greyed out with a count of 0, or hide it. Greying out preserves the layout and tells shoppers the option exists in the catalogue. Hiding keeps lists short. For size grids, greying out is almost always better, because a missing size 39 is information, while a disappearing button looks like a bug.

    Zero results should still be possible to reach, for example through a search query combined with filters, so design the empty state deliberately. Name the applied filters, offer to remove the most restrictive one with its resulting count, such as "Remove Waterproof to see 46 boots", and never show a blank grid.

    On desktop, results can update instantly after each click, provided the update arrives quickly, ideally within about 300 milliseconds, and the page does not jump. If it takes longer, a loading state on the grid is essential, otherwise shoppers click twice and toggle a value off again.

    Mobile filter patterns that keep context

    On a phone, filters usually open in a full screen sheet or a bottom drawer. The risk is that the shopper loses all context: she cannot see products, cannot see how many results remain, and does not know what changed when she closes the sheet.

    Four patterns address this. Keep a sticky button at the bottom of the sheet that shows a live count, such as "Show 128 boots", and update it after each selection. Show the applied filter chips at the top of the sheet as well as on the list page.

    Return the shopper to the same scroll position in the list after closing filters, or to the top of the new results if the set changed substantially. And put the 2 or 3 most used facets as quick filter chips above the grid, so the most common choices need no sheet at all.

    Touch targets need room. WCAG 2.2 sets a minimum target size of 24 by 24 CSS pixels at level AA, and many design systems use 44 or 48 pixels for primary touch controls. Size grid buttons and checkbox rows that look fine on desktop are often far smaller on a phone.

    Finally, test with one hand on a 6 inch screen. If the Apply button sits behind the browser toolbar, or the drawer closes when the shopper scrolls a long brand list, the filters are effectively unusable no matter how elegant the panel looks in a mockup.

    Accessibility and the release checklist

    Filters are form controls, and they should behave like a well built form. Group each facet's options in a fieldset with a legend that names the facet, so screen reader users hear "Size, group" before the values. Announce result count changes with a polite live region rather than moving focus, which would disorient keyboard and screen reader users. Make sure a facet can be expanded, navigated and changed with the keyboard alone, and that focus returns to a sensible place after the mobile sheet closes.

    Before launch, run through this list for each category template.

    • Every facet in the panel came out of research data, and its data coverage is at least 80%.
    • Hard requirement facets, such as size or compatibility, sit in the top three positions.
    • Each value shows a product count that updates after every selection.
    • Size facets use a grid in standard size order, with unavailable sizes greyed out.
    • Colour swatches carry text labels.
    • Applied filters appear as removable chips with a Clear all option on desktop and mobile.
    • The zero result state names the filters and offers a specific way out.
    • The mobile sheet has a sticky button with a live count and keeps scroll position.
    • Each facet group uses a fieldset and legend, and result updates are announced politely.
    • Filter selections survive a visit to a product page and the back button.
    • Filter interactions are tracked, so the next round of design can use real usage data.

    Questions teams ask about faceted filters

    Should values inside one facet combine with OR or AND? Within one facet, such as selecting both Blue and Green, use OR so the result set grows; across different facets, such as Blue plus size 42, use AND so it narrows. Mixing these up is one of the most common causes of unexpected zero results.

    Should filter combinations be indexable by search engines? Usually only a few high demand combinations deserve their own indexable landing page with unique copy; the thousands of other combinations should point a canonical tag at the parent category or be kept out of the crawl, otherwise they waste crawl budget on near duplicate pages.

    How many facets is too many for one category? There is no fixed limit, but if a category shows more than about 10 facets, analytics usually reveal that the last few are used in well under 1% of sessions; collapse those by default or move them into a single More filters group.

    Should a filter panel remember selections when the shopper returns to the category? Keep selections while the shopper moves between a product page and the list in the same session, since losing them on the back button is a major frustration, but start fresh on a new visit unless the shopper is signed in and has saved preferences such as shoe size.

    The trade-off at the heart of faceted filter design is breadth against clarity. Every facet you add helps a few shoppers find a niche product and makes the panel slightly harder for everyone else.

    Category specific facets, ordered by research and backed by complete data, cost more to build and maintain than a single global filter set, and they pay that cost back on every category with more than a couple of hundred products. For small catalogues the verdict flips: a few well chosen facets and a good default sort will beat an elaborate panel that nobody needs.

    Related Articles