Once responses come in, a tree test gives you a lot of numbers. Here's what each one means, what good looks like, and how to turn them into decisions.
Success
The share of people who chose a correct answer. This is the headline number for each task.
There's no universal pass mark, but as a rough guide:
- 75% or more: working well.
- 50% to 75%: worth a look. Something is pulling people away.
- Below 50%: a real problem. Most people can't find it.
Always read success alongside its likely range. With 20 participants, 60% could plausibly be anywhere from about 39% to 78%. With 50, the range is much tighter.
Directness
The share of people who went straight there, without going back up the tree. People who backtracked and still succeeded are an indirect success.
High success with low directness means people find it eventually, but only after hunting. That's a problem on a real site, where every wrong turn is a page load and a chance to give up.
First click
Which top-level section people opened first. It's one of the most telling numbers: people who start in the right section are far more likely to succeed. If many people start somewhere wrong, the top-level labels are the problem, not the deeper ones.
Time
How long the task took. Long times on a task that people eventually got right suggest hesitation. Compare each task's time with the others rather than reading it on its own.
Where people ended up
For each task, look at the wrong answers that attracted the most people. These are usually your quickest wins:
- One wrong answer with lots of votes suggests a label that overlaps with the correct one, or content that people genuinely expect in two places. Consider renaming, merging or cross-linking.
- Wrong answers spread thinly across the tree suggest people had no idea where to start. The structure may not match how people think about the topic.
Pietrees and paths
A pietree shows the routes people took through the tree for one task. Each circle is a place people visited, and its slices show what they did next: went deeper, went back, or chose it as their answer. Thicker lines mean more people.
Look for:
- Busy wrong branches, where many people explored before backing out.
- The point where paths split, often a label that two branches both seem to own.
- Near misses, where people reached the right section but picked the wrong item inside it.
Turning results into changes
- Rank tasks by success, lowest first. Start with anything below 50%.
- For each problem task, find the cause: a wrong first click, a strong competing answer, or a near miss.
- Make the smallest change that addresses it: a clearer label, moving one item, or adding a cross-link.
- Test again with the same tasks and new people, then compare the two rounds.
ClickTree does the first two steps for you in "What to look at", which lists the tasks to fix first and why, using fixed rules applied to your results.
Common traps
- Blaming the tree for a bad task. If almost everyone skipped or failed, reread the task. Is it ambiguous, or does it use words that point elsewhere?
- Chasing small differences. A task at 72% isn't really worse than one at 76% unless you have a lot of participants.
- Forgetting the baseline. A redesign that moves a task from 40% to 65% is a big win, even though 65% isn't perfect.
New to tree testing? Start with how to run a tree test.