By Mehdi HassanWhat are card sorting and tree testing?
Card sorting is a research method in which participants sort cards, each naming a page or topic from your website, into groups that make sense to them, and often name those groups. Tree testing is the reverse: participants see your proposed menu structure as plain text and try to find where they'd go to complete a task. Card sorting shapes a structure; tree testing checks it.
Both methods test the structure without any visual design, which is exactly why they're useful. Design can hide a confusing structure for a while, but buyers still get lost. Testing on text alone shows whether the labels and groupings work on their own. For the structure you're testing, start with our website information architecture guide.
Card sorting | Tree testing | |
|---|---|---|
Question it answers | How do people group and name our content? | Can people find things in our structure? |
When to run it | Before drafting the sitemap | After drafting it, before design |
What participants see | Cards with page or topic names | A text-only menu hierarchy |
What you get | Groupings and label ideas | Success rates and wrong turns per task |
Best for | New sites or big restructures | Checking a proposed or existing menu |
How do you use card sorting and tree testing to structure a website?
Use card sorting first to learn how buyers think about your content, draft the sitemap from what you learn, then use tree testing to check that buyers can find key pages in that sitemap. Fix the labels and groupings that fail, and test again if the changes are large. All of this happens before a designer opens Figma.
The lightweight version below is built for small teams: a founder or marketer, five participants and an afternoon. It won't give you statistically solid numbers, but it reliably surfaces the labels nobody understands and the pages nobody can find, which are the problems that matter most on a marketing site. Nielsen Norman Group at nngroup.com publishes deeper guidance if you want to run larger studies.
How do you run a card sort with five users?
Run a card sort by writing your topics on cards, asking each participant to group them and name the groups, and noting where they hesitate. Five participants from your real audience is enough for a small site, because the biggest disagreements with your current structure show up quickly.
List 20 to 40 topics
Write one card per page or topic you expect on the site: features, segments, pricing, security, customer stories, guides. Use plain descriptions, not your current menu labels.
Recruit five target buyers
Pick people who match your buyers: customers, prospects from sales calls or people in the right role. Colleagues and friends know too much about your product to be useful.
Choose open or closed sorting
In an open sort, participants create and name their own groups, which is best for new sites. In a closed sort, they place cards into categories you provide, which is best for checking an existing menu.
Run the sessions
Use an online card-sorting tool, a shared whiteboard or a spreadsheet on a video call. Ask participants to think aloud, and resist explaining what any card means.
Record groupings and names
For each participant, note which cards went together, what they named each group and which cards they couldn't place.
Look for patterns
Cards most people grouped together belong together. Group names people chose are label candidates. Cards that moved around a lot need clearer names or a different home.
How do you run a tree test on your navigation?
Run a tree test by turning your proposed sitemap into a clickable text hierarchy, giving participants realistic tasks, and recording where they look first and whether they reach the right page. The first click matters most: when buyers start in the wrong section, they rarely recover quickly.
- Write the tree. Turn your sitemap into a plain-text hierarchy with your proposed labels, two or three levels deep. Our SaaS website sitemap template is a good starting point.
- Write five to eight tasks. Base them on what buyers really try to do: "You want to know whether the product works for a team of fifty. Where would you look?" Don't use the words in your labels.
- Mark the correct answers. For each task, note the page or pages that count as success.
- Test with five participants. Use a tree-testing tool or a simple shared document, and ask each person to say where they would click first and why.
- Score each task. Record whether they succeeded, where they clicked first and whether they backtracked.
- Fix and retest. Rename labels or move pages for tasks that failed, then retest those tasks if the changes are significant.
For example, suppose buyers can't find your security information
For example, suppose a tree test includes the task: "Your IT team needs to know how this product stores customer data. Where would you go?" Three of five participants open Product first, one opens Resources, and one finds Security under Company.
That's a clear signal. The security page is filed where the company thinks about it, as a company policy, not where buyers look for it. The fix could be a Security link in the footer and a link from the pricing page's enterprise tier, plus a Trust section in the Product menu.
Rerun just that task with a few new participants after the change. Our checklist of B2B website trust pages covers where those pages should live.
How do you turn the results into a website structure?
Turn the results into structure by treating them as evidence for specific decisions: which pages to group, what to call each group and which pages need a second route. Don't average the results into a compromise menu; look for the clear signals and the specific points of confusion.
- Strong groupings become sections of the sitemap.
- Repeated group names become navigation labels, especially when they're simpler than yours.
- Cards that wander need a clearer name, or a link from more than one place.
- Failed tree-test tasks point to labels to rename or pages to move.
- Successful tasks tell you what to leave alone in a redesign.
From there, decide which groups deserve header placement and whether any need a dropdown or larger panel, using our website navigation best practices and the rules on when a mega menu helps.
When is card sorting not worth doing?
Card sorting isn't worth it for very small sites where the structure is obvious, such as a five-page seed-stage site with home, product, pricing, about and demo. There's nothing to sort. A quick tree test of your labels may still catch a confusing name, but a full card sort would be overkill.
It's most valuable when you have many pages to organize, several products or segments, or a site where analytics and sales calls suggest buyers can't find things. In those cases, an afternoon of testing is far cheaper than redesigning the navigation after launch. If you also want to test pages once they're designed, our guide to usability testing on a budget covers the next step, and our guide to testing website accessibility covers checks for keyboard and screen reader users.
