Journal

How to design filters people can actually use

A practical guide to filter design: choosing controls, showing active filters, handling zero results and testing with real tasks.

How to design filters people can actually use

Filters look like a solved problem. Put some checkboxes in a sidebar, add a "Clear all" link, ship it. Then you watch a session recording and see someone tick three boxes, get zero results, and leave without knowing which of the three caused it.

Good filtering is less about the controls and more about the conversation between the person and the list. They ask a question, the interface answers, and they adjust. This guide walks through the decisions that make that conversation work, whether you're designing a product catalogue, a job board, a dashboard table or a template marketplace.

Start with the questions people actually ask

Before you draw a single checkbox, list the questions your users bring to the page. "Which of these fit my budget?" "What's available this week?" "Which ones work on mobile?" Each real question becomes a filter. Each filter that answers nobody's question becomes noise.

The fastest sources for these questions are search logs, support tickets and sales calls. If people keep typing "under 50" into search, price deserves a prominent filter. If nobody has ever asked about a field, it doesn't need a control just because it exists in the database.

  • Keep: filters tied to a decision people make before buying, booking or clicking.
  • Demote: filters a minority uses, behind a "More filters" disclosure.
  • Cut: filters that mirror internal data structure rather than user language.

Pick the right control for each filter

The control should match the shape of the data. Mismatched controls are the most common reason filters feel clumsy.

  • Checkboxes for categories where people often want several values at once, like colours or sizes.
  • Radio buttons or a segmented control when only one value makes sense, like "Sort by" or a single plan type.
  • A range slider with number inputs for price, duration or size. The slider alone is fiddly; pair it with editable fields.
  • A toggle for a single yes/no condition, like "In stock only" or "Free".
  • A searchable list once a category has more than about ten options, such as brands or locations.

Show counts next to options when you can ("Linen (12)"). Counts tell people what's behind each choice before they commit, and they prevent the dead-end click.

Decide between instant and batch filtering

There are two models. Instant filtering updates the results the moment someone changes a control. Batch filtering waits until they press "Apply".

On desktop, instant filtering usually wins. People see cause and effect immediately and can undo a choice in one click. Just make sure each update is fast and doesn't shift the layout under the cursor.

On mobile, filters typically live in a full-screen sheet that covers the results, so instant updates happen out of sight. Here a batch model makes more sense, with one improvement: put the live result count on the button itself, such as "Show 24 results". People get the feedback without leaving the sheet.

Always show what's active

Once someone has applied filters, the state should be visible at the top of the results, not buried in a collapsed sidebar. The usual pattern is a row of removable chips above the list.

  1. One chip per active value, written in plain words: "Under $50", not "price_max=50".
  2. A clear "×" on each chip to remove just that value.
  3. A single "Clear all" at the end of the row, only when two or more filters are active.
  4. The total result count next to the chips, so every change has a visible effect.

This row does a lot of quiet work. It explains why the list looks the way it does, and it gives people a fast way to back out of a narrow search.

Design the zero-results state on purpose

Zero results is where most filter designs fail. A blank area with "No results found" leaves people guessing which choice to undo.

A better empty state names the active filters and suggests the smallest change that would bring results back. "No templates match Free + Dark mode + E-commerce. Remove E-commerce to see 8 templates." If you can, prevent the problem earlier by disabling or greying out options that would return nothing, while still showing them so people know the category exists.

Treat an empty result as a question the interface failed to answer, not as the user's mistake.

Keep filters in the URL

Filtered views get shared, bookmarked and reopened. If the filter state lives only in memory, a refresh or a back-button press wipes it and people start again. Write active filters to the URL query string so the view is reproducible.

This is mostly a developer task, but it's a design decision too. Agree on readable parameter names with your developer, decide what the back button should do (undo the last filter, or leave the page), and make sure a shared link opens with the chips already visible.

Order and group with intent

Put the filters people use most at the top. That sounds obvious, yet many sidebars follow database order or alphabetical order. Group related filters under short headings and keep the first two or three groups expanded by default. Collapse the rest.

Within a group, order options by how often they're chosen or by their natural sequence (sizes small to large, dates nearest first). Alphabetical order only makes sense for long named lists like brands or cities.

Test it with real tasks

Filters are hard to judge from a static mockup. Build a quick prototype with realistic data, at least a few hundred items, and give people tasks instead of tours: "Find a jacket under $100 that ships this week in your size." Watch for three things:

  • How often they hit zero results, and whether they recover.
  • Whether they notice the active filter chips.
  • Which filters nobody touches. Those are candidates for the "More filters" drawer.

Analytics after launch tell the same story at scale. Track which filters get used, which combinations lead to zero results and which lead to a click on an item.

Wrapping up

Good filters answer the questions people already have, use controls that match the data, show their state clearly and never leave someone stuck on an empty page. Start from user language, design the active and empty states as carefully as the default one, and keep the whole state in the URL. Do that, and filters stop being a sidebar of checkboxes and become the fastest way through your content.

Oybek Ruziev

Designer & Framer developer · AidanCreates

Keep reading

All posts