The tree is the URL: a tree testing tool with no backend

Tree testing is the method you reach for when you need to know whether people can find things in your structure, before you've committed to layouts, labels on mockups, or a single Figma frame. You give participants a plain hierarchy (your sitemap, your nav, your IA spreadsheet exported as an outline) and a set of tasks: "Where would you go to cancel your subscription?" They click down through the tree until they land somewhere. You record whether they succeeded, how direct their path was, and where they went wrong.

It's one of the highest-signal tests in IA work, and like card sorting, it's mostly gated behind research platforms that want a monthly fee for what is, underneath, a tree and a list of questions. So I built one into the Lab, using the same zero-backend approach as the card sort.

The study is still the URL

Same defining decision as the card sort: refuse a backend. A study (title, optional brief, the hierarchy, tasks, and which nodes count as correct answers) is serialized into the link hash. Share the URL and you've shared the study. Nothing is stored on a server. Nothing leaves the participant's browser until they export CSV or JSON and send it back to you.

Tree testing is especially well suited to this model because the input is text. You're not uploading screenshots or prototype files. You paste an indented outline (two spaces or a tab per level) and the tool turns it into a navigable hierarchy. Home, Products, Footwear. That's the whole upload step.

Text only, on purpose

Stripping visual design out is the point. A tree test measures whether your structure makes sense on its own. If people can't find "Returns" when there's no marketing chrome to guide them, they won't find it on the live site either. The austerity is a feature. It forces the IA to do the work.

One level at a time

Participants drill down one level per screen, with a breadcrumb to orient them and a back button when they take a wrong turn. Every node also has an "I'd stop here" action, because real users don't always drill to the leaf. They pick the category they think is close enough. Mobile-first by default: no drag, no hover states, no precision required.

Success, directness, and the full path

For each task you mark one or more nodes as correct. When a participant finishes, the tool records three things that matter to practitioners: did they land on a correct answer, did they get there without opening a wrong branch first (directness), and the full path they took, including detours. That detour data is often the most useful output. "Everyone went to Help before finding Returns" is a finding you can act on.

What it doesn't do

There's no cross-participant analytics dashboard, no p-value calculations, no integration with Optimal Workshop's analysis suite. You get one result set per person and merge them yourself. For a large quantitative tree test with statistical rigour, use a commercial tool. For the quick validation runs that happen before most redesigns (a link, five tasks, and a folder of CSVs), this is enough.

Both tools live in the Lab. Same philosophy: the study is the URL, the results are yours, and it costs nothing to run because it costs me nothing to host.