Fix navigation people cannot find their way through.
Tree Testing shows you whether people can find things in your menu structure, with the design stripped away. Card Sorting shows you the groups and labels people expect before you name anything. User Sage runs both, with the AI Panel for a first look and real participants for the answer.
How fixing navigation works
Test the structure, not the screens
Paste your menu as a tree of labels, the way it sits in your product today. Then write the tasks as things people come to do: find where to change a password, find the price of the team plan. No design and no styling, so a failure can only mean the structure or the words are wrong.
Ask people how they would group it
A Card Sort gives people your topics as cards and lets them make their own groups and name them. You see the categories people carry in their heads, which are often different from the ones your team chose.
Rename and regroup with the results in front of you
For every task you see how many people found the right place, how many took a wrong branch, and which label sent them there. Change the tree, run the test again, and see whether more people find the right place this time, before you change the navigation in your product.
What you get
For every task, how many people found the item, and how many ended up somewhere else.
The paths people took instead, and the label where they turned.
The groups and group names your participants made for themselves in the card sort.
A before and after: run the same tasks on the new tree and compare the scores.
Most navigation problems are label problems.
A word your team uses every day can mean something else to everyone outside the building, and no amount of visual polish fixes a label people walk past. A tree test shows you, task by task, which labels people miss. A card sort shows you the names they reach for instead. The AI Panel run points at the weak labels first, and real participants decide which names stay.
Questions about tree testing.
What is the difference between a tree test and a card sort?
A tree test works backward from your current structure: here is the menu, can you find this thing in it. A card sort works forward from people's expectations: here are the topics, how would you group them and what would you call the groups. Run the card sort when you are designing the structure and the tree test when you want to check it.
Do I need the design finished?
No, and that is the point of both methods. They test words and structure only, so you can fix the navigation before a designer has drawn a single screen, and again after the redesign if the scores did not improve.
How many participants does a tree test need?
More than a prototype test, because you are comparing percentages rather than watching a handful of people. Thirty or more real participants per tree test gives you numbers you can compare before and after. Start with the AI Panel to catch the labels that fail for everyone.