Quick definition
Card sorting is a research activity in which people organize labeled items into groups that make sense to them. Marketers use it to understand how an audience expects products, services, resources, or support topics to be organized. Each card represents one item, such as delivery information, pricing, customer stories, or a particular product family. The arrangement exposes relationships that a company's internal department structure may hide. A visitor might associate installation advice with buying a product, while the business stores it under technical support. That mismatch is useful evidence for designing categories and labels. A card sort produces input for an information architecture; it does not automatically produce a finished website menu or prove that visitors can complete a real task.
Choose the question before the format
An open sort lets participants create groups and name them. A closed sort supplies the categories and asks people to place items within them. A hybrid sort supplies some categories while allowing additional groups. Choose according to the decision: discovering unfamiliar audience language calls for more freedom, while investigating whether existing labels make sense calls for a more constrained exercise. Consider a training company deciding whether its courses should appear under job roles or skills. An open exercise can reveal the organizing principle people naturally use. A closed exercise can then examine a proposed structure. Avoid treating either exercise as a vote on your preferred answer. If participants cannot place an item comfortably, preserve that difficulty as a finding instead of forcing agreement.
Prepare cards people can understand
Start with a content inventory and select items that represent the decisions visitors actually face. Use comparable levels of detail: mixing a broad subject such as marketing with a single article about writing an email subject line makes the grouping task confusing. Write neutral, concise labels and add brief explanations only where necessary. Avoid including the intended category in every card name, because that gives away the structure you are trying to investigate. Recruit people resembling the intended audience, including meaningful differences in experience or buying responsibility. Pilot the instructions and card set before launching the full study. Check whether unfamiliar wording, missing items, or excessive volume causes fatigue. Make the activity accessible and provide an alternative way to participate when the chosen interface creates barriers.
Run sessions without teaching your structure
Explain that there are no correct groups and that the purpose is to understand the participant's expectations. In a moderated session, ask people to explain uncertain placements and describe what a group label means to them. Neutral prompts such as “What made those items belong together?” reveal more than asking whether a proposed label sounds good. Record changes as well as the final arrangement: moving an item repeatedly can expose ambiguity. Remote unmoderated exercises can collect more arrangements efficiently, but usually provide less explanation of unusual choices. Use a short follow-up question to capture difficult items. Keep employee opinions separate from customer evidence. Employees can help identify operational constraints, but familiarity with internal naming often makes their grouping behavior different from that of a first-time visitor.
Interpret patterns with a practical example
Imagine a fictional software company sorting twelve resource topics with eighteen prospective buyers. Fourteen participants group implementation timelines with migration guidance, while only six group implementation timelines with pricing. Those observations suggest a relationship worth exploring; they do not mean that fourteen votes establish the only correct location. Read participants' explanations and compare experienced buyers with newcomers. The company might create a “Getting started” resource group while also linking implementation guidance from pricing pages. Look for consistently paired items, competing category names, and topics that fit several purposes. Do not average away an important audience difference merely to create one tidy diagram. Document the proposed grouping, supporting observations, ambiguous items, and assumptions that need another test. This makes the design decision reviewable rather than dependent on a visually impressive clustering chart.
Turn findings into a tested structure
Translate the findings into a draft hierarchy with clear labels, then test whether people can locate information for realistic tasks. Tree testing can examine a text-based structure without the distraction of visual design. Usability testing adds page layouts, navigation controls, search, and content to the picture. A card sort may show that returns and delivery belong together, but a task test reveals whether the label “Order help” actually guides someone seeking a refund. Prioritize consequential journeys such as finding eligibility requirements or comparing services. After release, examine failed searches, repeated navigation loops, support questions, and completion of relevant tasks. Keep the evidence proportional: a small exploratory study can generate useful design hypotheses, but it cannot establish precise population preferences or guarantee an increase in conversion rate.
Frequently asked questions
Is card sorting the same as tree testing?
No. Card sorting explores how people group items or interpret proposed categories. Tree testing asks people to find information within a hierarchy while completing specific tasks. The methods answer related but different questions. A useful sequence is to use sorting findings to shape a draft structure and then examine whether that structure supports successful navigation.
How many cards and participants should a study include?
There is no single number that fits every study. The scope of the content, diversity of audiences, complexity of labels, and intended analysis all matter. Pilot the exercise to identify fatigue and confusing instructions. For exploratory work, prioritize useful explanations and audience coverage; stronger quantitative claims require an appropriately designed sample and careful interpretation.
Can we use employees instead of customers?
Employees can help inventory content, expose missing topics, and identify constraints, but their arrangements should not be treated as equivalent to customer research. They often know internal product names and department responsibilities that visitors do not. If an employee exercise is the only available starting point, label it as an internal hypothesis and test consequential decisions with intended users.
What if participants disagree about where an item belongs?
Disagreement can reveal vague wording, different audience needs, or content that naturally supports several journeys. Examine the explanations before changing the item. A clearer label, contextual links, multiple navigation paths, or audience-specific guidance may solve the problem. The aim is useful findability, not forcing every topic into one category that receives unanimous agreement.
Related terms
- Information architecture: The organizing structure that card-sorting evidence helps shape, including relationships between content, categories, and labels.
- Navigation design: The visible controls and paths that help visitors move through an information architecture and reach useful destinations.
- Tree testing: A task-based method for checking whether people can find information within a proposed hierarchy after grouping decisions have been made.
- User research: The broader practice of investigating audience needs and behavior, of which card sorting is one focused method.
- Usability testing: Observation of people attempting realistic tasks, helping reveal problems that a grouping exercise alone cannot uncover.
- Customer journey mapping: A way to understand the broader tasks and stages that should inform which content items and participants a sorting study includes.