A product team, deeply familiar with their own product's internal architecture, designs a navigation menu that groups features and content in a way that feels perfectly logical from inside the team's own mental model — and user testing later reveals that customers consistently look for the same content in an entirely different place, organized around a different underlying logic than the one the team used. Card sorting exists specifically to surface this kind of gap before launch, by asking representative users to organize content themselves, rather than asking the internal team to guess how users think.
What the method actually involves
Participants are given a set of individual items — feature names, content topics, page titles — typically represented as physical or digital cards, and asked to sort them into groups that make sense to them, based on whatever criteria feel natural, without being told the categories in advance. This directly surfaces the participant's own mental model for organizing the content, which is precisely the information a navigation designer needs and can't reliably generate purely through internal team discussion, however thoughtful that discussion might be.
The distinction between open and closed card sorting
An open card sort lets participants create their own category labels and structure from scratch, which is most useful early in a design process when the goal is discovering what natural groupings and category language users bring to the content, independent of any structure the team has already proposed. A closed card sort instead asks participants to sort the same items into a set of predefined categories the team has already developed, which is more useful later in the process, to validate or stress-test a specific proposed navigation structure rather than generate a new one from scratch.
Why internal team logic and user logic diverge so often
A product team's internal organization of content is shaped by how the product was actually built — technical architecture, feature development history, internal terminology — none of which a typical user has any exposure to or reason to organize their own thinking around; a user instead brings their own goals, prior experience with comparable products, and everyday language to the task of finding something, which frequently produces a genuinely different, and not obviously wrong, way of grouping the same underlying content.
What makes the analysis of card sort results actually rigorous
A single participant's card sort, however interesting, doesn't establish a reliable pattern on its own — the analytical value comes from running the exercise with a reasonable number of participants and looking for groupings and category labels that recur consistently across many of them, using formal similarity-clustering analysis where the sample size allows for it, rather than treating any one particularly articulate or confident participant's specific sort as the definitive answer.
What this means for using card sorting effectively in information architecture work
- Use open card sorting early to discover natural groupings and category language, and closed card sorting later to validate a specific proposed structure
- Run the exercise with enough participants to identify genuinely recurring patterns, rather than drawing conclusions from a small handful of individual sorts
- Be specifically skeptical of a navigation structure designed purely from internal team logic without any direct user input on how they naturally organize the same content
- Treat card sorting as one input alongside other UX research methods, particularly tree testing, which directly tests whether a proposed structure actually helps users find things successfully
Card sorting works precisely because it sidesteps the trap of a team designing navigation around their own internal understanding of the product — it asks the people who'll actually use the structure to reveal their own mental model directly, rather than asking the team to guess at it from the inside.