Card Sorting vs Tree Testing: Which Should You Run First?
Card sorting and tree testing are complementary IA research methods, not alternatives. Card sorting reveals how users naturally group content; tree testing validates whether they can find specific items in your navigation. Run card sorting first to discover mental models, then tree testing to check if your design works. Here's the difference and the sequence that makes sense.
Card sorting happens during discovery to inform structure. Tree testing happens after design to measure findability. Always run card sorting before tree testing.
Side by Side
| Card Sorting | Tree Testing | |
|---|---|---|
| Question answered | How do users naturally group this content? | Can users find specific items in this structure? |
| When in the project | Discovery phase, before design | Validation phase, after design |
| Output | Category groupings, cluster analysis | Success rate, time on task, directness score |
| Sample size | Varies by project goals | Varies by project goals |
Free plan: 3 studies. Unlock the full analysis for a single study for $15, or choose Starter at $19/mo.
1,800+ users
The Decision Flow
Start with an open card sort
Discover how users naturally group your content without any predefined structure. This reveals mental models and category names that make sense to your audience.
Optional: Run a closed card sort
If you want to validate specific category proposals from your open sort, run a closed sort where participants place cards into your predefined categories. Measures agreement with your structure.
Design your navigation
Turn the card sort clusters into a navigation structure. Use clear labels, logical hierarchy, and consistent terminology informed by your participants' language.
Validate with a tree test
Test whether users can actually find things in your new structure. Tree testing measures success rate, time, and path directness. Recruit different participants who haven't seen the card sort.
Running Both in Sequence
Card sorting informs design; tree testing validates it. Here's why running both matters:
- Card sorting alone tells you how users think about content, but not whether your interpretation of that data produces a navigation structure that actually works for findability tasks.
- Tree testing alone measures performance against a structure that may be poorly aligned with user mental models in the first place, leading to low success rates for the wrong reasons.
- Both together give you discovery (card sorting) and validation (tree testing). You build an IA informed by how users think, then measure whether it delivers the findability you need.
Tree Testing: The Next Step
After you've designed your navigation from card sort results, a tree test checks whether users can complete real tasks:
- Task-based: "Where would you find the return policy?" rather than "Sort these cards."
- Text-only navigation: No visual design, just the hierarchy and labels.
- Quantitative output: Success rate (did they find it?), time on task, and directness (did they take the optimal path?).
Tree testing is part of the ValidateThat platform. Learn more about tree testing or build a tree test.
ValidateThat tree testing: validate your card sort results with task-based findability tests.
Frequently Asked Questions
Can I skip card sorting and just run a tree test?
You can, but you'll be testing a structure that may not match how users think about the content. Card sorting informs better navigation design; tree testing measures whether that design works. Skipping card sorting means you're validating a guess rather than a user-informed structure.
How long should I wait between card sorting and tree testing?
Wait until you've designed the navigation structure from your card sort results. The tree test validates your interpretation of the card sort clusters, so the structure needs to exist before you can test it.
Do I need different participants for each method?
Yes. Participants who did the card sort have seen all the content and labels, which biases their tree test performance. Recruit separate groups to avoid contamination and get realistic findability data from people who haven't studied your IA already.
What if my tree test results are poor?
Go back to your card sort data and check whether your navigation design actually reflects the clusters users created. Poor tree test results often mean you misinterpreted the card sort, used unclear labels, or made too many structural compromises. Adjust the design and test again with new participants.