UX glossary
Tree Testing
Tree testing is a research method for checking whether people can find content in a navigation structure. Only the outline of labels is tested, with no visual design at all.
Imagine an online store for outdoor gear that has just designed a new menu. It looks tidy, and everyone on the team thinks it makes sense. But the question that really matters is a different one: Can a first-time visitor find the returns page, the size chart for hiking boots or the gift card for the family? That’s exactly what a tree test checks, before a single screen has been designed.
Explained simply
Tree testing is a research method for checking whether people can find content in a navigation structure. The “tree” is simply the outline of your website: main categories, subcategories and the individual pages beneath them, with no colors, images or layout. Participants get tasks and click through this text-only hierarchy until they think they’ve reached the right place.
Here’s an everyday comparison: you’re standing in a hardware store where every shelf is empty and only the aisle signs are left. Can you find the aisle with the wall plugs from the signs alone? If not, the problem lies with the signs and the way things are organized, not with the products.
The method used to be run on paper with index cards. Donna Spencer described it in 2003 as “card-based classification evaluation”. It is sometimes also called “reverse card sorting”, because it works the opposite way to card sorting: a card sort generates a structure, a tree test evaluates one.
What you measure
The key metrics are the success rate (how many people end up in the right place?), directness (how many get there without detours or backtracking?) and time per task. The first click is especially revealing. According to the Nielsen Norman Group, it often predicts whether someone will complete the task. If first clicks are spread across several top-level categories, those categories probably overlap too much.
An example
Back to the outdoor store. Before, returns are tucked away under “My account” and the size chart under “Service”. In a tree test with the task “Your new hiking boots are too small. Where do you start an exchange?”, many people look under “Help” or go straight to “Footwear”. Only a small share find the right spot. The team renames “Service” to “Help and returns”, adds a link to the size chart in every footwear category and tests again. Afterwards, far more participants find the right path on the first try.
Common misconceptions
A tree test is not a full usability test. Because search, images and page context are missing, it only shows how well the structure and labels work on their own. On the finished website, results may turn out better or worse. That’s not a flaw, it’s the point: it tells you whether a problem lies in the information architecture or in the visual design. For more on how to structure menus, read Flat vs. Deep Hierarchy: Which Structure Is Best?
In practice
- Don’t give away the answer. Write tasks without the exact labels from your menu. Instead of “Find the returns page”, try “Your boots don’t fit. What do you do?”.
- Start with an easy task. A simple warm-up task helps people get going and shows you who isn’t paying attention.
- Compare versions fairly. When you test two structures against each other, use the same tasks with separate groups. For reliable differences, the Nielsen Norman Group says you typically need 50 or more participants per tree.
Related terms
Card Sorting, Information Architecture, Usability Testing, Mental Model