Card Sorting for Better Information Architecture

Card sorting is a research method where participants organize topics or content into categories that make sense to them, revealing the mental models people use to understand how information relates. Researchers use the results to build information architecture around how users actually expect content to be structured, rather than around internal organizational assumptions that may not match user expectations.

Navigation built around internal team structures instead of user expectations creates confusion that shows up as high bounce rates, repeated support questions, and content people cannot find even when it exists. Card sorting addresses this by grounding structure in evidence rather than assumption. The two-minute video above walks through the core ideas.

What is card sorting and what does it reveal?

Card sorting is an activity where participants group a set of topics into categories that feel natural to them, often also assigning labels to those categories. The activity reveals participants' mental models, meaning the way they understand relationships between different concepts and where they expect to find specific information.

This is valuable because the categories that make sense internally to a team that built a product are not always the categories that make sense to the people using it. Card sorting surfaces that gap directly, using real behavior instead of assumptions about how users think.

What are the different approaches to card sorting?

There are two main approaches to card sorting, and they answer different questions. In an open card sort, participants create their own categories and labels from scratch, which is useful for discovering how users would organize information with no constraints. In a closed card sort, researchers provide existing categories and evaluate whether participants can place content into them correctly.

Open sorts work well early in a project, when a team wants to understand user mental models without bias from an existing structure. Closed sorts work well later, when a team already has a proposed structure and wants to validate whether it makes sense to real users.

How should a card sorting study be designed?

The quality of card sorting results depends on thoughtful study design, including clear instructions, relevant participants, and carefully selected content items that represent real information needs. Weak design at any of these points undermines what the results can reliably show.

Key design considerations include:

• Selecting participants who represent the actual audience the information architecture is meant to serve.
• Choosing content items that reflect real information people need to find, not an arbitrary sample.
• Writing instructions that are clear enough for participants to complete the task without confusion.
• Deciding between an open or closed approach based on whether the goal is discovery or validation.

How should researchers analyze and act on card sorting results?

Analysis involves looking for repeated grouping patterns, common labels, and areas where participants disagree, since both agreement and disagreement are informative. Repeated patterns suggest a structure most users would recognize, while disagreement points to areas where navigation and terminology may need extra clarity or testing.

Card sorting does not produce a perfect structure automatically; researchers still need to interpret findings alongside other evidence, since users may organize information differently depending on their specific goals in the moment. Validating a proposed structure afterward is often done through a related method, described in tree testing explained.

Key takeaways

• Card sorting reveals users' mental models by having them group topics into categories that make sense to them.
• Open card sorts support discovery of new structures, while closed card sorts validate an existing proposed structure.
• Result quality depends on relevant participants, realistic content items, and clear instructions.
• Analysis should examine both agreement and disagreement between participants, since disagreement often points to confusing terminology.
• Card sorting results still need interpretation alongside other evidence rather than being treated as a final answer.

How PulseLake helps

PulseLake's traditional research mode supports advanced methods like card sorting alongside surveys and qualitative research, keeping objectives, participant data, and results in the same persistent study context. AI agents for research design and analysis can help structure a card sort and surface grouping patterns, while researchers retain judgment over how findings translate into navigation decisions. To plan a card sorting study, talk to our team.

Frequently asked questions

How many participants are needed for a card sorting study?

There is no single required number, but many teams find that a moderate sample, often in the range of fifteen to thirty participants, is enough to identify clear, recurring grouping patterns. The right number depends on how much variation exists in the content and audience, and researchers should watch for the point where additional participants stop revealing new patterns.

What is the difference between card sorting and tree testing?

Card sorting asks participants to create or evaluate categories by grouping content, which is typically used to design or refine an information architecture. Tree testing asks participants to find specific items within an existing hierarchical structure, without any visual design, to validate whether that structure works. The two methods are often used together, with card sorting informing structure and tree testing validating it.

Can card sorting be done remotely without in-person sessions?

Yes, card sorting is commonly conducted remotely using digital tools that let participants complete the sorting task on their own devices. Remote card sorting can reach a broader and more representative set of participants than in-person sessions alone, though researchers should still provide clear instructions since there is no facilitator present to clarify confusion in real time.

Does card sorting work for redesigning an existing website or app?

Yes, card sorting is commonly used both to design new information architecture and to evaluate an existing one. When applied to an existing product, researchers can compare how users naturally group content against the current navigation structure, which often reveals specific areas where the existing organization does not match user expectations.