What is tree testing?

How does tree testing work?

Participants see a text-only hierarchy of the site's navigation. They are given a task, such as "find out whether the product integrates with Salesforce", and they click through the tree until they arrive where they believe the answer lives. No page content, no images, no search.

The stripped-back format is the point. A visually appealing page can carry a confusing structure, and a clean structure can be hidden by poor design. Testing the tree alone answers whether the organization and the labels work.

It is typically unmoderated and run remotely, which is what makes the sample sizes it needs affordable.

What does it measure?

Three metrics, and the third is the one that names a fix.

Success rate. The share of participants who arrived at the correct location.

Directness. The share who got there without backtracking. High success with low directness means the structure is navigable and not obvious, which shows up on a live site as hesitation and extra clicks.

First click. Where participants went first, which points at a cause rather than a symptom. A task where most first clicks land in the wrong top-level branch has a labeling problem in that branch, and that is something you can rewrite.

Failures point at specific causes: a label using internal vocabulary, two categories whose scope overlaps, or content filed somewhere logical to the business and not to the visitor. This is the structural version of an information scent problem.

How does it fit with card sorting?

They are complementary halves of the same work, run in sequence.

Card sorting is generative. Participants group content items themselves and name the groups, which produces candidate structures reflecting how the audience thinks.

Tree testing is evaluative. It takes a proposed structure and measures whether people can navigate it.

The usual sequence: card sort to generate a structure, tree test to validate it, revise, tree test again. Both rounds happen before any design work, which is where the saving is. A labeling problem found in a text tree costs an afternoon. The same problem found after the build costs a rebuild.

For CRO specifically, tree testing earns its place in three situations: a site migration where the structure is being rebuilt, a navigation with a high exit rate on category pages, and any site where internal search queries use words absent from the navigation.

How does tree testing apply to a Webflow site?

Webflow constrains what a tree can be, so test a structure the platform can actually produce.

Collection pages sit one level deep. A CMS collection page renders at /collection-slug/item-slug and does not nest further natively. A tested structure that files items three levels down needs static pages or rewriting, which is a decision worth making before the test rather than after it.

The collection slug is the whole branch. Renaming it changes the URL of every item inside it, so map the redirects as part of the structural change, not as cleanup afterward.

Static page folders carry the rest of the tree, and the navigation is built separately from them. Reorganizing the page tree does not update the nav, and updating the nav does not move the pages. Tree test the structure you intend to ship, then reconcile both.

Related terms

Information scent · Usability testing · Five second test · Hick's law · Exit rate

Service: Conversion Rate Optimization.

FAQ

How many participants does tree testing need?

30 to 50 per audience group. Higher than qualitative usability testing because the output is quantitative, meaning success and directness rates that need enough responses to be stable.

What counts as a good tree testing success rate?

There is no universal benchmark worth quoting, and a rate borrowed from someone else's study says nothing about your labels. Judge within your own study: the task scoring well below the rest is the one to fix, and the same tasks re-run on a revised tree tell you whether the revision worked.

Can you tree test an existing site?

Yes, and it is a useful baseline. Testing the current structure establishes where it fails, which gives you something to measure a proposed replacement against.