Tree Test
A tree test shows people your navigation as plain text, gives them a task, and records where they look for the answer.
Set up your test
Study title
Tree
Up to five levels: category, subcategory, then items. Select a row to add an item under it or remove it. Blank rows are removed when you create the study.
In a name box, Tab indents the item under the one above and Shift+Tab moves it out a level. Drag the dots to move an item, or use the Up, Down, Indent and Outdent buttons.
Paste an outline instead
One item per line. Indent a child with two spaces or a tab (Tab and Shift+Tab work here; press Escape first to move on). This replaces the tree above, and you will need to re-tick each task's correct destinations. Up to five levels.
Tasks3
Write each task as something a real user would think, such as "where would you find X", not "click on X".
Options
Participant journey
Optional. Empty boxes use the wording shown.
Shown first, on the start screen.
Shown under the welcome message, before the first step.
Questions before the study
Up to 3. Asked before the task. Don't ask for names or emails unless you need them; participants are told what's collected.
Questions after the study
Up to 3. Asked after the task.
Shown when the participant has finished.
Optional. Starts with https://. Sent after 5 seconds.
People enter it before seeing the study.
Stored as a hash. Change it later in Share.
Already have a study?
Paste your study key and results (owner) key to view your results.
An example tree test task.
- Customer Service the answer
- Track My Order
- Returns & Refunds
- Warranty Information
- Contact Us
- My Account the plausible wrong turn
- Order History
- Payment Methods
- Saved Addresses
- Laptops & Computers
- Laptops
- Desktops
- Monitors
- Computer Accessories
- Phones & Tablets
- Smartphones
- Tablets
- Cases & Covers
- Chargers & Cables
- TV & Audio
- Televisions
- Soundbars
- Headphones
The wrong turn is deliberate — the example shows the tool catching a real backtrack.
This tells you what happened, not why.
An unmoderated tree test shows you where people ended up and how directly they got there. It doesn't tell you why a label confused someone. That needs a follow-up question or a moderated session. A handful of responses is enough to spot an obviously broken category; treat a small sample as indicative, not a statistically reliable measurement. This tool doesn't recruit participants for you: you bring your own, whether that's colleagues, a mailing list, or a panel you already pay for.
Frequently asked questions
What's tree testing?
Tree testing checks whether people can find things in a navigation structure using text alone. No visual design to help. See the definition above for the full explanation.
How is this different from a card sort?
A card sort is how you DESIGN a structure (participants group items into categories that make sense to them, and you use the patterns to build a hierarchy. A tree test is how you VALIDATE a structure you already have) participants are given a finished tree and try to find things in it. Do the card sort first if you don't have a structure yet; use the tree test once you do.
What's the difference between card sorting and tree testing?
Card sorting comes before you have a structure: participants see a pile of unsorted content items and group them, which tells you how people would group and name things. Tree testing comes once you have a structure to check: participants see a finished, labelled hierarchy and try to find things in it, which tells you whether people can find things in what you already built.
Can you provide an example of tree testing?
This page has a runnable one, pre-filled below: an electronics store's navigation (Laptops & Computers, Phones & Tablets, TV & Audio, Customer Service, My Account), with tasks like finding whether a recent order has shipped, buying a case for a new phone, or checking a warranty. Edit it directly to see how the tool records success, directness and where people go wrong.
How many participants do I need?
Classic usability guidance (Nielsen Norman Group) suggests five participants can surface most of the big, obvious problems with a structure, though the numbers here are indicative, not from any study of ours. More participants let you trust smaller differences between tasks and catch less common wrong turns. A handful of responses is enough to spot a badly-placed category; don't treat a five-person tree test as a statistically reliable measurement.
What does "directness" mean in the results?
Directness is the share of SUCCESSFUL attempts where the participant went straight to the right place without backtracking up a level first. A task can have a high success rate but low directness. People eventually find the right answer, but only after trying somewhere else first, which is itself a sign the label or grouping isn't quite right even though it "worked".
Who can see my results?
Only whoever holds the results (owner) link, which is a separate, much longer secret from the link you share with participants. We generate it once, show it to you once, and can't recover it if you lose it: there's no "forgot my results link" flow. Save it somewhere safe the moment you get it.
Comparing tree testing tools? See the free Treejack alternative.