Summary
Information Architecture (IA) determines how content is organized, labeled, and connected. Card sorting (open and closed) reveals how users naturally group content, informing structure decisions. Tree testing validates whether a proposed IA is navigable before any visual design begins. These techniques catch structural problems when they are cheap to fix.
While this book focuses on broader methods, certain specialized techniques deserve attention. Information Architecture (IA), the structure and organization of content, is a foundational layer that affects every user interaction.
Poor IA is a root cause of many usability problems. Users may struggle not because individual screens are confusing, but because the overall structure does not match their mental model of how information should be organized.
The Core Methods: Sorting & Testing
IA research relies on two complementary methods: one to discover how users think, and one to validate that your structure works.
| Method | Type | Purpose | When to Use |
|---|---|---|---|
| Card Sorting | Generative | Discover how users categorize information | Early design, building from scratch |
| Tree Testing | Evaluative | Validate if your structure is navigable | Before development, after IA is drafted |
Card Sorting (Generative)
Used to discover how users categorize information. There are two variants:
| Variant | User Task | Best For |
|---|---|---|
| Open Sort | Users organize topics and name the groups themselves | Starting from scratch; understanding the user's mental model |
| Closed Sort | Users organize topics into pre-defined groups you provide | Validating if your category labels make sense |
Tree Testing (Evaluative)
Often called "Reverse Card Sorting." Where card sorting asks "how would you organize this?", tree testing asks "can you find this?"
The Method: Remove the design and show only the navigation tree (text-only). Give the user a task ("Find the shipping costs") and observe where they click.
The Goal: Test if the labels and hierarchy work in isolation, before you distract users with visual design. If users fail on a text-only tree, no amount of beautiful icons will save your navigation.
Card Sorting: Deep Dive
Card sorting is a method that falls under our "asking" building block. You ask participants to organize topics into groups that make sense to them.
Purpose
The goal is to understand your users' mental models to create an intuitive navigation structure. In essence, you are having users perform the act of tagging for you.
Open Card Sort
Participants group topics (on cards) and then create their own names for the groups. They are effectively creating their own taxonomy from the bottom up.
Best for: Early-stage IA design when you do not have a predefined structure.
Output: Natural groupings and user-generated labels that reveal how your audience thinks about the content.
Closed Card Sort
Participants are given pre-defined group names and must sort the cards into those categories. This is like asking them to apply an existing set of tags.
Best for: Validating whether an existing or proposed structure matches user expectations.
Output: Agreement rates showing how well your categories align with user mental models.
Conducting Card Sorts
Card sorting can be conducted:
- Using specialized online tools (functioning like a survey)
- As a hands-on activity at the beginning of an interview or UX test
Tree Testing: Deep Dive
Tree testing is the evaluative counterpart to card sorting. It validates whether your proposed IA is easy to navigate before you have designed any screens.
How It Works
- Present users with a text-only version of your site's structure (the "tree")
- Give them a task (e.g., "Find information on shipping costs")
- Observe whether they can successfully navigate the tree to find the correct location
- Measure success rates and identify where users go wrong
Why Text-Only?
By removing visual design elements, colors, images, icons, tree testing isolates the structure itself. This reveals whether problems are caused by the organization of content rather than visual presentation.
Key Metrics
| Metric | What It Measures |
|---|---|
| Task success rate | Percentage of users finding the correct location |
| Directness | Whether users went straight to the answer or backtracked |
| Time to complete | How long users took to navigate |
| First click | Where users started their journey (often the most predictive metric) |
When to Use Each
| Technique | Stage | Question Answered |
|---|---|---|
| Open card sort | Early design | "How do users naturally organize this content?" |
| Closed card sort | Mid-design | "Does our proposed structure match user expectations?" |
| Tree testing | Pre-development | "Can users find what they need in this structure?" |
The Connection to Qualitative Analysis
Card sorting directly overlaps with the principles of qualitative analysis. In an open card sort, participants are effectively performing coding for you:
- They identify themes (create groups)
- They label those themes (name the groups)
- They reveal their reasoning (through think-aloud during the activity)
This makes card sorting results particularly valuable for building taxonomies that genuinely reflect user mental models.
Practical Considerations
Sample size: For card sorts, 15-20 participants typically reveal stable patterns. For tree testing, 30-50 participants provide more reliable quantitative metrics.
Number of cards: For card sorting, 30-60 items is typical. More than 100 becomes cognitively demanding.
Task design: For tree testing, tasks should be realistic scenarios, not just "find X", give context about why the user would need that information.
What This Means for Practice
IA problems are among the most expensive to fix after development. A poorly structured navigation affects every interaction a user has with your product.
Card sorting and tree testing catch these problems when they are cheap to fix, before visual design and development have begun. They provide evidence-based foundations for structural decisions, rather than relying on internal assumptions about how content "should" be organized.
Start with card sorting to understand how users think. Validate with tree testing before committing to development.