Users rarely complain about “information architecture”. They say things like “I couldn’t find the fees page” or “Why is the refund policy under About Us?”. Those complaints are about structure: how content is grouped, what the groups are called, and how people move between them. That structure is the information architecture (IA) of your website or app.
Card sorting is one of the simplest research methods for designing that structure based on how users think rather than how your organisation is arranged. This guide explains the basics of IA, walks through planning and running a card sort, shows how to analyse the results, and explains how tree testing checks the structure you create. If you are new to the topic, the information architecture lesson in our UI/UX track is a good first read.
What information architecture covers
Information architecture is the practice of organising, structuring and labelling content so people can find what they need and understand where they are. It is not the same as navigation design, although the two are closely related: IA is the underlying structure, and the navigation menu is one way of showing that structure on screen.
| Part of IA | Question it answers | Example |
|---|---|---|
| Organisation | How is content grouped? | Courses grouped by subject, or by level? |
| Labelling | What do we call each group? | “Admissions” or “How to apply”? |
| Navigation | How do people move between groups? | Top menu, side menu, tabs, breadcrumbs |
| Search | How do people find things directly? | Site search with filters |
A common mistake is copying the organisation chart. A college website with top-level sections named after internal departments makes sense to staff, but a parent looking for hostel fees has no idea which department owns that page. Good IA starts from the user’s mental model – the way they expect information to be grouped.
What card sorting is
In a card sort, participants receive a set of cards, each showing one topic or piece of content, and group them in a way that makes sense to them. You can do this with paper cards on a table or with an online card-sorting tool. The way people group and name things reveals their mental model, which you then use to shape your structure and labels.
Card sorting is a generative method: it helps you create or rethink a structure. It does not prove that a finished navigation works. For that, you use tree testing, covered later in this guide.
Open, closed and hybrid card sorts
| Type | How it works | Use it when |
|---|---|---|
| Open | Participants create their own groups and name them | You are designing a new structure or want to learn users’ vocabulary |
| Closed | You provide fixed categories; participants place cards into them | You want to see whether content fits an existing set of categories |
| Hybrid | Fixed categories, but participants can also add their own | You have a partial structure and want to see where it falls short |
For beginners, an open card sort is usually the most useful, because the group names participants invent are valuable on their own. Nielsen Norman Group notes that hybrid sorts can bias participants towards your categories, and that tree testing is often a better way to evaluate an existing structure than a closed sort.
How to plan a card sort
1. Define the goal
Write one sentence: “We want to learn how prospective students group information about our courses so we can redesign the main menu.” A clear goal keeps your card list focused.
2. Choose the cards
Pick content that represents the area you are studying. Nielsen Norman Group suggests roughly 30 to 50 cards: enough to see real patterns, but not so many that participants get tired. Tips for writing cards:
- Use one clear item per card: “Hostel fees”, “Bus routes”, “Scholarships for girls”.
- Write in users’ language, not internal codes or jargon.
- Avoid repeating the same word on many cards (“Student fees”, “Student hostel”, “Student clubs”). People tend to group by the shared word instead of by meaning.
- Add a short description to a card only if the title alone could be misunderstood.
3. Choose participants
Recruit people who match your real audience. For a college site, that might be prospective students and parents, not current staff. As a guide, Nielsen Norman Group recommends at least 15 participants for a qualitative card sort and a larger sample, around 30 to 50, if you want quantitative results. See user research basics for recruiting tips and user personas for defining who your users are.
4. Moderated or unmoderated?
- Moderated (in person or on a video call): you watch participants sort and ask them to think aloud. You learn why they group things, which is very valuable. It takes more time per person.
- Unmoderated (online tool): participants sort on their own. You can reach more people quickly, and the tool calculates patterns for you, but you lose the reasoning.
A practical mix is to run a few moderated sessions first to check your cards make sense, then an unmoderated sort for numbers.
Running the session
- Explain that there are no right or wrong answers; you are testing the content, not the person.
- Ask participants to sort the cards into groups that make sense to them. Groups can be any size.
- Allow an “I don’t know” or “doesn’t fit” pile. Cards that end up there often reveal unclear labels.
- Ask them to name each group in their own words (open sort).
- In moderated sessions, ask short follow-up questions: “What made you put these together?”, “Was any card hard to place?”.
- Note cards they hesitated over, re-read or moved several times.
Analysing the results
You do not need statistics software to learn a lot from a card sort. Look for these patterns:
| What to look at | What it tells you |
|---|---|
| Cards often grouped together | Strong candidates for the same section |
| Cards placed in many different groups | Unclear content or a card that belongs in more than one place – consider cross-links |
| Common group names | Labels that match users’ vocabulary |
| Cards in the “doesn’t fit” pile | Confusing wording or content users do not expect on the site |
| Differences between user types | Students and parents may need different entry points |
Online tools usually provide a similarity matrix (how often each pair of cards was grouped together) and a dendrogram (a tree diagram of clusters). These are helpful summaries, but combine them with what you heard in moderated sessions. The numbers tell you what happened; participants’ comments tell you why.
Watch out: Card sorting gives you evidence, not a ready-made menu. Use it to inform your structure, then make design decisions, such as how many top-level items a mobile menu can hold, that participants could not make for you.
From results to a sitemap
- Draft top-level groups from the strongest clusters.
- Name them with the most common, clearest participant labels.
- Place each piece of content, noting items that need to appear in two places through links.
- Draw the structure as a sitemap or tree diagram (FigJam or a simple spreadsheet works).
- Turn the sitemap into navigation and page layouts through wireframing.
Tree testing: checking your new structure
Tree testing evaluates a structure without any visual design. Participants see only a text version of your hierarchy (the “tree”) and are given tasks such as “Where would you find out whether the college has a hostel for girls?”. They click through the tree to the place they think is right.
Because there is no visual design, tree testing isolates the IA itself: if people fail, the problem is your grouping or labels, not your colours or layout. Card sorting and tree testing work well as a pair – card sort to generate a structure, tree test to check it, then adjust and test again. Usability testing of the final design, described in our usability testing lesson, comes after that.
Common beginner mistakes
- Testing with colleagues instead of real users, who think in internal terms.
- Using too many cards, so participants rush the second half.
- Card labels that hint at the answer by sharing keywords.
- Treating the most popular grouping as the final answer without considering the minority patterns.
- Skipping validation – a structure that came from a card sort can still fail a tree test.
Information architecture fits naturally into the “define” and “ideate” stages of design thinking. It also supports the heuristic “match between the system and the real world” from Nielsen’s usability heuristics: labels and groupings should follow users’ language and expectations.
Card sorting checklist
- Clear research goal written down.
- About 30–50 cards in plain user language, without repeated keywords.
- Participants who match the real audience.
- A pilot session to catch confusing cards.
- Moderated sessions for reasons, unmoderated for scale (if needed).
- Analysis of strong clusters, scattered cards and group names.
- Draft sitemap based on findings, then a tree test before visual design.
Frequently asked questions
What is the difference between information architecture and navigation?
Information architecture is the underlying structure and labelling of content. Navigation is the set of interface elements, such as menus, tabs and breadcrumbs, that lets people move through that structure. Good navigation depends on good IA.
How many participants do I need for a card sort?
Nielsen Norman Group recommends at least 15 participants for a qualitative card sort, and a larger group, around 30 to 50, when you want reliable quantitative patterns.
Can I do card sorting without paid tools?
Yes. Paper cards on a table work well for moderated sessions, and you can record groupings in a spreadsheet. Online tools mainly save time when you have many participants and want automatic analysis.
Should I use an open or closed card sort?
Use an open sort when you are designing a new structure or want to learn users’ words. Use a closed sort when you already have fixed categories and want to see where content fits. To evaluate a finished structure, tree testing is usually better.
Is card sorting useful for mobile apps?
Yes. Apps also need content grouped in ways users understand, such as which features go in the bottom navigation and which belong in settings. The method is the same; you just design the result for a smaller screen.
