I understand that a way to test IA is through tree testing.

However, a lot of articles mentioned that tree testing requires at least 30 participants (best 50 participants).

How else can we test our IA if we do not have that amount of participants - let's say we only have around 5-8 of them.

Are we able to do "qualitative tree testing"? - not sure if this is even a proper term.

Meaning that while participants are performing tree testing, we will observe their behaviour and ask questions on why they chose to navigate a certain way to understand their mental model (similar to a usability testing but using sitemap instead of wireframes). If not, how else can we go about doing it?

We would like to think that testing with skewed data results is better than testing with no results.

  • I have no experience with this myself but it might work: Before you give participants any options to choose from, you can ask them what they would look for. After that you show them the options and let them choose. Continue with the same question first. This way you have double the data you would normally have. Even when the answer given isn't one of the options, it can give some new insights.
    – jazZRo
    Jun 12, 2020 at 11:12
  • Agreed: the results from a few users, will tell you a lot more than testing with zero users.
    – PhillipW
    Jun 13, 2020 at 9:37
  • @jazZRo Hi there, thank you for your comment. Your idea sounds great, but the only issue we have is that participants may not know what they want unttil they are presented with options. We foresee that most of our participant would say "I'm not sure what I would look for/I don't know". While I understand where you're coming from, based on past experiences, I fear that it may not be the best method for my participants. Thank you once again for your suggestion! Jun 13, 2020 at 13:22

1 Answer 1


You can go about this in a couple of ways - I would suggest that rather than you giving the users a structure you think works, that you have them actually create it.

So I would create a card sort exercise using a free card sort program (there's a bunch of free trial ones out there like UserZoom or OptimalSort) Create the high level categories and the items that go in the categories (these are usually listed on the right hand side). Then users can drag and drop items into the categories you have created. This will give you a good idea of how your users group information.

If you don't have the time for that, then create a prototype using something like Maze and ask the users to do a couple of tasks that test the IA to see if they can find what they're looking for. This allows you to really test if the IA or structure works - it's really hard for users to look at a tree structure and tell you if it works or not without context. Also, the prototype allows you to test remotely or in person.

Also, I disagree with the 30 participants rule - 5 users should be enough for you to see if your IA works or not and see any big issues with it.

  • We are already planning to conduct at Card Sorting workshop with 5 participants. However, we want to ensure that the IA derived from this workshop matches the mental model of all the other users - hence the need for a quantitative tree testing data. With that said, do you suggest that all we need is just 5 users from the card sorting workshop to validate our IA, and to scrap the tree testing method? Jun 13, 2020 at 13:19
  • Yes - you should see a pattern emerge with your 5 users - just make sure they are representative of your target audience e.g. don't have all novice users but a mix of novice and experts. If you don't know who your target audience is, create user personas that represent your users and figure out who you're trying to focus on for the given task. Jun 13, 2020 at 16:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.