To diagnose information architecture problems through click paths, look beyond popular pages: examine which category users start from, whether they take detours, and where they end up. Success rate and directness—whether users take detours—can indicate how easy information is to find. Detailed paths help determine whether to adjust top-level categories, lower-level labels, or item placement. However, behavioral records from a live website and tree tests answer different questions. Repeated clicks are only a clue that users may be struggling, not proof that the categorization is wrong.
First, distinguish actual browsing paths from tree tests
CXL explains that path analysis can use a tree diagram to show a sequence of actions on a website or app: which pages users commonly visit after arriving, whether they loop back to the same place, and what they do after a particular event. This helps identify paths worth investigating, but loops alone do not establish information architecture as the cause.
NN/g focuses instead on tree testing: asking users to find items for specified tasks to check whether the categories are easy to understand. In this type of test, subcategories appear only after a parent category is selected, so the first click must be on a top-level category. This restriction does not apply to every website interface.
Wrong first click: check where users expect to find the information
When task success is low or users often take detours, examine their first clicks. Beyond whether users eventually find the answer, first clicks reveal where they initially expected the information to be.
- If users often look for an item in another category, consider moving it or listing it under multiple categories.
- If first clicks are spread across several categories for many tasks, the categories may overlap too much. Consider card sorting or reviewing the overall organizational structure.
In a task cited by NN/g that involved finding a teacher, only 14% of users first selected the correct category, Directory, and the overall success rate was 20%. Many continued wandering after entering other sections. This example illustrates an association between an incorrect starting point and failure to find information; it does not mean all websites have the same rates.
Right first click, but no answer: investigate lower-level categories
If most users enter the correct section but still cannot find the answer, examine where they ultimately choose to stop. In NN/g’s example involving arts festival information, everyone correctly selected Recreation first, but their final selections were spread across different subcategories. Only 30% chose the correct location. Based on this result, NN/g identifies excessive overlap among subcategories at the same level.
Results like these suggest making subcategory labels easier to distinguish, rather than merely changing the top-level entry point. Distinguish between users who start in the wrong place and those who start correctly but get lost later, rather than treating every problem as unclear navigation.
Account for interface differences when comparing redesign options
To compare information architectures, use the same tasks to test multiple category trees. Tests usually use a between-subjects design, meaning different groups of participants test different trees. Then compare success rate, time taken, directness, and path metrics. NN/g notes that quantitative comparisons typically require at least 50 participants per tree to obtain reasonably narrow confidence intervals and detect statistically significant differences. This is not a fixed threshold: reaching that number does not guarantee a difference will be detected.
When interpreting the results, also consider how the final interface will display the categories. A mega menu that shows multiple category levels at once may produce a higher success rate than a tree test, provided users enter the correct menu first. Incorrect first clicks therefore still need attention.
Click paths and information architecture: perspectives from across the web
Should the same information appear under multiple categories?
Not necessarily. It depends on whether users look for the same information in different categories, which tree testing can help determine. If several locations count as correct answers but almost no one selects one of them, listing the item there may be unnecessary. If users do look in different categories, consider retaining multiple entry points so they can find the information where they expect it.
When comparing two information architectures, do differences in metrics establish that one is better?
You cannot draw a conclusion from higher or lower numbers alone. In a quantitative comparison, examine success rate, time taken, directness, and path metrics, but also check whether the differences between the alternatives are statistically significant. NN/g cautions that most tree-testing software does not perform this comparison automatically, so researchers need to calculate it themselves. Sample size requirements can also increase research costs.
References
- All You Need to Know About Google Analytics 4 (A Complete Guide to GA4)cxl.com · 2023-07-05T10:54:41+00:00 — Source used in the article above.
- Tree Testing Part 2: Interpreting the Resultswww.nngroup.com · 2024-01-19T17:00:00+0000 — Discusses: Should the same information appear under multiple categories?; When comparing two information architectures, do differences in metrics establish that one is better?


