# Tree Testing

A task-based evaluation of a text-only navigation hierarchy that measures whether people can find where information should live.

- Type: method
- Domain: Research & behavior
- Level: Intermediate
- Last reviewed: 2026-09-21
- Original: [Soroush’s Product Wiki](<https://wiki.soroush.uk/topic/tree-testing/>)

## What it means

Tree testing presents participants with a simplified, text-only hierarchy and asks where they would go to complete specific tasks. By removing page layout, search, and visual cues, it isolates the findability of category labels and structure. Results commonly include success, directness, time, and the paths people choose, supported by qualitative comments about confusing branches.

## Why it matters

A navigation can look elegant while sending users down the wrong branch. Tree testing checks a proposed or existing hierarchy before substantial interface work and identifies labels that compete, categories that overlap, and tasks with no obvious home. It is a useful counterpart to card sorting because it evaluates a structure rather than merely generating ideas for one.

## When to use it

Use tree testing when revising menus, documentation, settings, catalogs, or other hierarchical information spaces. Run it after drafting an architecture and before visual design, and repeat it to compare meaningful structural changes rather than cosmetic label tweaks.

## How it works

1. Build a realistic text hierarchy that includes the branches and leaf destinations needed for the study.
2. Write task prompts around user goals without repeating category labels or revealing the answer.
3. Recruit representative participants and randomize task order where order effects are plausible.
4. Measure destination success, first choice, path directness, backtracking, and completion time.
5. Inspect wrong paths and comments, revise labels or structure, and retest the most important tasks.

## A practical example

A software company tests a help-center tree with the task “Stop teammates from inviting new members.” Most participants choose “Security” instead of “Workspace roles,” and several bounce between the two. The team moves invitation permissions into a clearer access-control branch and retests before rebuilding the help center and settings navigation.

## Where it breaks

A text tree is intentionally stripped of search, page content, visual hierarchy, and interaction cues, so it cannot predict the whole product experience. Task wording can prime a branch, while easy tasks may inflate overall success. A successful destination does not prove that the content there is understandable. Results need interpretation, not only a single score.

## Common mistakes

- Writing tasks with the same words used in the correct branch label.
- Reporting overall success while ignoring important tasks and misleading first choices.
- Using tree testing to judge visual design, page content, or interactive behavior it does not display.

## Sources

1. [Tree Testing: Fast, Iterative Evaluation of Menu Labels and Categories](<https://www.nngroup.com/articles/tree-testing>) — Nielsen Norman Group · Professional guidance
2. [Card Sorting vs. Tree Testing](<https://www.nngroup.com/articles/card-sorting-tree-testing-differences>) — Nielsen Norman Group · Professional guidance

## Related topics

- [Card Sorting](<https://wiki.soroush.uk/topic/card-sorting/>)
- [First-Click Testing](<https://wiki.soroush.uk/topic/first-click-testing/>)
- [Usability Testing](<https://wiki.soroush.uk/topic/usability-testing/>)
