Card sorting is a user research method in which participants organize labeled cards, each representing a piece of content or a function, into groups that make sense to them. In UX design, card sorting is used to shape information architecture: it reveals how users mentally group topics and what they expect categories to be called, so that navigation matches users’ mental models rather than the organization’s internal structure.
The method is generative rather than evaluative. It does not test whether an existing navigation works; it supplies evidence for designing one. Its usual partner is tree testing, which later checks whether people can actually find items in the proposed structure.
How a card sorting study works
A study starts with a set of cards, each carrying a short, plain label for a content item, such as Track my order or Change password. Participants sort the cards into groups and, depending on the format, name those groups. The researcher then aggregates all sorts to find patterns that recur across participants.
Card sorting can run in person with physical cards or remotely with online tools. In-person sessions let the researcher hear participants think aloud and ask why a card belongs where it does. Remote unmoderated studies trade that insight for larger samples and lower cost.
| Parameter | Common range | What it affects |
|---|---|---|
| Number of cards | 30 to 60, rarely above 100 | Participant fatigue and data quality |
| Participants, moderated | About 15 to 20 | Depth of qualitative insight |
| Participants, unmoderated | About 20 to 40 or more | Stability of statistical patterns |
| Session length | 15 to 30 minutes | Completion rate and attention |
| Label length | Two to five words | Speed and clarity of sorting |
Open, closed and hybrid card sorting
The three formats differ in who defines the categories. Choosing the right one depends on how much of the structure already exists and which question the team needs answered.
| Format | Categories | Answers the question | Best stage |
|---|---|---|---|
| Open sort | Participants create and name them | How do users group and label this content | New site or major redesign |
| Closed sort | Predefined by the researcher | Do users place items where we expect | Refining an existing structure |
| Hybrid sort | Predefined, plus new ones allowed | Which categories are missing or unclear | Partially defined structure |
Open sorting produces the richest insight into vocabulary, because the group names show the words users naturally reach for. Closed sorting is easier to analyze but cannot reveal categories the team never thought of, which is why it is usually run after an open sort rather than instead of one.
Analyzing card sorting results
Analysis turns many individual sorts into a shared picture of agreement. Most tools produce several standard views, and experienced researchers read them together rather than relying on any single chart.
| Output | What it shows | How to use it |
|---|---|---|
| Similarity matrix | Share of participants who grouped each pair of cards together | Find strong and weak pairings |
| Dendrogram | Hierarchical clustering of cards by agreement | Suggest candidate groups at different levels |
| Category agreement | How consistently cards land in a closed category | Spot ambiguous items and weak labels |
| Participant labels | Names users gave their groups | Choose navigation wording |
Results should be treated as evidence, not a blueprint. A pair grouped together by about 70 percent of participants is a strong signal, while cards that scatter across many groups usually point to an unclear label or content that genuinely belongs in several places, which may call for cross-linking rather than a single home.
Card sorting among other UX research methods
Card sorting answers a narrow question about grouping and naming. It complements other methods in UX design rather than replacing them, and the table below shows how it relates to its closest neighbors.
| Method | Main question | Output | Relation to card sorting |
|---|---|---|---|
| Card sorting | How do users group content | Candidate categories and labels | Generates structure |
| Tree testing | Can users find items in a hierarchy | Success rates and paths | Validates the structure afterward |
| Usability testing | Can users complete tasks in the interface | Observed problems | Tests the full design later |
| User interviews | What do users need and why | Goals and context | Informs which cards to include |
A frequent sequence is interviews, then an open sort, then a tree test. Interviews and analytics decide which content deserves a card, the open sort proposes groups and labels, and the tree test measures whether the resulting hierarchy actually works. Card sorting also feeds persona and user journey work, because different user groups often sort the same cards differently, which can justify audience-specific entry points or a search-first approach for some sections.
Preparing cards and avoiding bias
The quality of a card sort depends largely on the quality of the cards. Most failed studies trace back to labels that lead participants or content sets that do not reflect real use.
- Avoid shared keywords: cards such as Account settings and Account history get grouped by the word, not by meaning.
- Avoid internal jargon: department names and product codes measure familiarity with the organization, not user expectations.
- Keep items at a similar level: mixing broad topics with single pages confuses participants.
- Recruit real users: colleagues and designers sort by knowledge of the organization and distort the results.
- Leave room for doubt: an Unsure group or the option to leave cards unsorted prevents forced, meaningless placements.
- Combine with tree testing: a structure built from card sorting should be validated before implementation.
Explained without jargon
Card sorting is like asking people to organize a messy kitchen drawer. Each person gets a pile of cards, one for each thing a website offers. They put the cards into piles that feel natural to them and give each pile a name.
When many people do this, clear patterns appear. The website can then be organized the way its visitors think. That makes things easier to find, because people look where they expect them to be.
A typical use case
A city library is redesigning its website, whose menu mirrors internal departments such as Collection Services and Outreach. Visitors complain they cannot find how to renew a book or reserve a meeting room. The UX team writes 48 cards in plain language, including Renew a loan, Book a study room and Children’s story time.
An open card sort with 30 library users runs online for one week. The analysis shows strong agreement on four groups that participants call roughly Borrowing, Events, Spaces and My account. Cards about printing scatter across groups, so the team decides to link them from several places.
A follow-up tree test with a new group of participants confirms the structure: most of them find the room booking task on the first attempt. The new navigation replaces department names with the categories users created themselves.