Faceted search: filters that help customers choose

What is faceted search?

Faceted search allows visitors to refine a result set on multiple properties, or facets. For shoes, that can be size, brand, type, color and background. For laptops, screen size, memory, processor and operating system make more sense. The available filters ideally change with the product context.

A facet is more than a field from the product feed. It has a customer-centric name, understandable values, a logical order and numbers that show how many results remain. The search layer must apply the selection reliably; the interface must clarify which choices are active.

Faceted search is different from a fixed category tree. Categories offer a primary structure, while facets allow combinations within the current set. Both must work together. A filter that excludes all results or a category that shows illogical attributes makes the route more difficult.

Choose only facets that support a real choice

Not every product attribute is a usable filter. Internal codes, technical source fields and features with virtually one value add little. Select facets based on customer questions, product differences and decision behavior. Search data, customer service questions, comparison pages and category expertise help with this.

Look at the distinctiveness as well. A filter is not very useful when almost all products have the same value. Conversely, a facet with hundreds of unique values can be unusable without search field, grouping, or range selection. The design depends on the type of data.

Limit the default view to the most important choices and make less used filters accessible without hiding them behind an unclear interaction. The right balance varies by category and device. Test with real tasks instead of just assessing whether the interface looks tidy.

• Does this characteristic help customers compare products in terms of content?
• Do the values differ sufficiently within the current result set?
• Is the data complete and consistent enough to provide reliable selection?
• Does the target group understand the name and the values shown?
• Can it be used on mobile without losing context?

Create filters contextually by category and search query

Same general filter list for all results causes noise. Size and fit are important for clothing, but not for office supplies. Power and connection help with electronics, while material and dimensions often weigh more heavily on furniture. Let category and result set determine which facets are given priority.

Search questions provide additional context. In ‘red dress size 40’, color and size may already be derived from the query. The interface can make that selection visible and show additional relevant filters. Prevent an automatically recognized value from being applied invisibly; the visitor must understand why the result set is smaller and be able to remove the choice.

Consider hybrid results. A wide query can contain multiple product groups with different characteristics. First, show facets that are meaningful across the groups, or help the visitor choose a product group before very specific filters appear.

Normalize labels, values and units

Filters make data quality visible. When the same color as ‘navy’, ‘navy’ and ‘dark blue’ occurs, the customer sees three choices that may mean the same thing. Determine whether values are truly equivalent and present a customer-centric group without unnecessarily losing the original product information.

Units require extra attention. Do not mix centimeters, millimeters and free text in one numerical filter. Save a standardized calculation value and a consistent view. For achievements, such as price, weight or screen size, relevant steps or popular intervals can work better than random boundaries.

Use understandable labels. A technical attribute may be correct in terms of content, but may require explanation. Short help text or an explanatory link can reduce doubt. Avoid marketing labels that hide multiple technical values when customers want to select on those differences.

• One preferred label for equivalent values.
• Fixed notation for capital letters, spaces and punctuation marks.
• Standardized unit for calculation and filtering.
• Natural sorting order, for example, size order rather than alphabetical.
• Clear treatment of unknown or missing values.

Design the filter interaction for desktop and mobile

On desktop, facets can be next to results, but screen space is limited there too. Clearly display active filters above or at the results and offer one recognizable way to remove a choice. Maintain scroll and result position where it supports the task.

On mobile, filters are often in a panel. Before opening, show how many filters are active and show in the panel how many results the current combination yields. Make the difference between ‘apply’, ‘erasing’ and closing without change clear. Prevent a visitor from jumping up unintentionally after each small choice.

Accessibility is part of the base. Use real form controls, clear labels, visible focus and logical keyboard sequence. Communicate updated result numbers in a way that can be understood even without visual animation. Test with longer translations and magnification.

Measure whether filters really improve the choice

A lot of use does not automatically prove that a filter is good. Visitors can often use a facet because the original results are too broad. Therefore, combine use with results: does the set get smaller, follow relevant product clicks and do not many combinations without results arise?

Analyze filters by category and device. A little-used facet can be essential for a small but valuable target group. A popular filter can be a symptom of a poorly set standard ranking. Use data as a starting point for research, not as an automatic removal rule.

Make changes checked. When you merge values, change order, or add default selections, compare the same result sets and important query tasks. Also check index coverage: An attractive filter design does not help when relevant products disappear due to missing data.

• Use per facet and per value.
• Number of results before and after filtering.
• Combinations that lead to zero-result searches.
• Product click and return after a filter action.
• Differences between mobile, desktop, categories and stores.