Information Architecture and Navigation Research
If visitors cannot find things, the problem may be the structure of the site rather than any single page. We use card sorting and tree testing to learn how your audience groups and names your content, then propose a structure that matches.
Short answer
Information architecture research tests how your audience expects content to be grouped and labeled. We use card sorting to see how people group topics and tree testing to check whether they can find items in a proposed menu. You receive the results, a proposed site structure and a label list.
On this page
Signs your structure needs work
Structure problems show up in familiar ways. Site search is used heavily for things that should be easy to find in the menu. Support receives questions answered on pages nobody can locate. Teams argue about where a new page belongs. The menu uses internal names, such as product codes or department titles, that visitors do not recognize. Each new section was added wherever there was space, and the menu has grown without a plan.
The methods we use
Content inventory
We list the pages and content types that exist today, with their current place in the structure and, where available, their traffic. The inventory can expose duplicate pages, orphaned content and sections nobody owns.
Card sorting
Participants sort cards, each representing a piece of content, into groups that make sense to them, and in an open sort they name the groups too. The results show which items people see as related and which words they use. Card sorting is a quantitative-leaning method, and Jakob Nielsen's guidance suggests about 15 participants for it, compared with five for a qualitative usability test 1.
Tree testing
Participants see a text-only version of a proposed menu and try to find specific items. Because the visual design is removed, the test checks the structure and labels alone. Results show where people look first, where they backtrack and which items they fail to find.
Label testing
Where specific labels are disputed, we test alternatives with participants to see which ones people understand without explanation.
What you receive
| Deliverable | Contents |
|---|---|
| Content inventory | Current pages, content types and structure, with notes on duplicates and gaps |
| Card sort results | Groupings, agreement between participants and the labels they used |
| Tree test results | Success and directness for each task, with the paths people took |
| Proposed structure | A revised site map with reasoning for each top-level group |
| Label list | Recommended menu labels and the evidence behind each one |
How many participants
Card sorting and tree testing produce numbers, so they need more participants than a qualitative usability test. Raluca Budiu of the Nielsen Norman Group explains that the five-user guideline applies only to qualitative studies, and that quantitative studies need larger samples, in most cases more than 30 participants 2. We set the number in the plan based on your audience and the decisions the results need to support.
Timeline
We plan for three to five weeks for an inventory, one card sort, a proposed structure and one round of tree testing. A second round of tree testing after revisions adds about a week.
Common pitfalls we plan around
Navigation research goes wrong in predictable ways, so the plan addresses each one in advance. Card sorts that include too many cards tire participants and produce noisy results, so we keep the card set focused on the content that matters most and split very large sets. Tree tests that reuse the exact words from the menu in their tasks become word-matching exercises, so tasks describe what the visitor wants rather than echoing the labels. And results from one audience are not assumed to hold for another: if your site serves very different groups, the plan says whether each group is tested separately.
Navigation research questions
Should we redesign the menu before or after testing?
Test first. A card sort shows how your audience thinks about the content, and a tree test of the current structure gives a baseline. Both make the redesign easier to justify.
Can you test a mega menu?
Yes. Tree testing works for any hierarchy, including large menus. We may split a very large structure into sections for testing.
Do card sorts tell us exactly what the menu should be?
No. They show patterns in how people group content. A good structure also reflects business priorities and content that does not exist yet, so we combine card sort results with those inputs.
Sources
- Why You Only Need to Test with 5 Users, Nielsen Norman Group (Jakob Nielsen), March 18, 2000.Secondary source, research firm article
- Why 5 Participants Are Okay in a Qualitative Study, but Not in a Quantitative One, Nielsen Norman Group (Raluca Budiu), July 11, 2021.Secondary source, research firm article
Planning a redesign or a big change?
Find out what visitors struggle with first. Tell us the deadline and we will scope research that fits before launch.
- No obligation
- Written scope
- Plain-English findings