Best practices for creating a tree test

6 min read Optimal Workshop

“I’m building a Treejack study and want to know if it’s a good idea to include a top level option for ‘I don’t think this fits anywhere’, in your experience is this a good idea? Why/why not?”


Tree testing is one of my favorite types of user research. It allows you to test the findability of content in your website information architecture (IA) and pinpoint exactly where people are getting lost. To answer your question Keegan, no, I don’t recommend including what is essentially a ‘too-hard basket’ branch in your tree. Don’t let that get you down though! I’m going to explain why and I’ll also give you all the tips and tricks you’ll need to create an effective tree test!

Building a strong tree from the ground up

So you’ve decided to do a tree test and your first job is to create that tree. The best way to do this is to build it out in a spreadsheet. You may already have your IA in a spreadsheet ready to go, but if not, it’s super easy to create. Your IA tree looks a lot like a sitemap – it shows all the hierarchical structure of all the content on your website. All you have to do is start with your highest level link labels and work your way down recording them as you go. From left to right, each column represents a level in your IA and each pathway should be staggered like steps (see below).

Tree labels inside of an Excel spreadsheet

It can be tricky to know what to include and what not to include in your tree test, which is why I’ve pulled together a quick guide to help you on your way.

What to include in your tree test

The parts of your information architecture that you want to test

Seems obvious, but it’s always best to double check that the IA or the part of the IA you’re testing aligns to your testing objectives. Also, don’t feel like you have to tree test your entire IA in one go! You might want to focus on a specific area of your IA instead. Information architecture tools like Treejack are designed to be flexible and scalable enabling you to do what best suits your research and business needs.

A starting point

Every tree test must have a single starting point. More often than not, it’s the ‘Home’ label but if you’re testing a smaller, deeper chunk of your IA, it’s probably something else. It could be a lower level landing page or any other logical starting point for a particular task or process.

Individual branches for sections of content on long scrolling pages

A key characteristic of an IA is its hierarchical structure of parent and child sections of content. Sometimes we’ll have long scrolling pages that contain multiple smaller chunks of content that aren’t links to new pages, but are still children for the purposes of a tree test because we need to know if users can reach that content. For pages like these, include child nodes in your tree for each section of content on long pages even if they’re not links.

Well designed task questions

It doesn’t matter how well you’ve designed your tree — if your task questions aren’t up to scratch, the whole thing will come crashing down. Your task questions are linked to your testing objectives and you want to write them in a way that naturally reflects how a user might interact with your website. You want to use language that makes sense to your users but also doesn’t give the game away by being an exact match to the labels on your tree! Write your tasks as hypothetical scenarios that tell a story and never include more than 10 tasks per tree test to ensure the time commitment required of your participants is fair and reasonable.

What not to include in your tree test

Labels that aren’t links and don’t have content associated with them

There’s plenty of websites out there that use subheading-like labels that aren’t actually links and don’t have any content associated with them. They’re sometimes used to provide additional context like in the below example from ASOS US.

A screenshot of the ASOS 'Women' menu

Whether that is or isn’t a good idea is a whole other story but I can tell you that anything that doesn’t have content associated with it, does not belong in a tree test.


Filters are a navigational element that help sort or refine content on a page. They work in parallel to the IA but they’re not part of it, so they’re out.

‘Too hard baskets’ and anything that would never appear in a live website IA

Here’s that explanation I promised you Keegan! There are two main reasons why I recommend you avoid including a label like this. The first one being that in a live website experience, there are no too-hard baskets. Instead, people leave and get what they need elsewhere or they contact you. Including that won’t help you understand where people are getting lost in your content. For really robust, usable data, you want your tree to resemble a real IA that could actually exist out in the world so you can learn where the real problems are. The second reason is, it’s an easy escape hatch for participants — let’s not make it too easy for them to give up!

I completely understand where you are coming from but it’s best to simply allow participants to skip questions in Treejack because the tool will show you exactly where they skipped the task and where they went before they skipped out. It also helps to include a post-study question providing participants with the opportunity to give you more information about their tree test experience. After they’ve completed the activity, you might ask them: “How confident were you in your responses during this activity? Why/Why not?”

Instead (or in addition to) you might also ask: “Is there anything else you would like to share about the activity you’ve just completed?”

This gives them a chance to explain and also allows you to gain even more insights into their thinking without compromising your tree test!

Labels like Search, Contact us and anything to do with a chatbot

These labels don’t add value to a tree test for a number of reasons. They won’t behave the same way in the real world compared to during a tree test and much like that too-hard basket branch, they give participants an easy out that won’t give you the data you need. It’s really important to remember that you want to see if they can find their way around your IA before turning to search or your chatbot. Same goes for Help and FAQ content.

Happy tree testing Keegan!