How do you research information architecture?

ProcrustesPR

To research information architecture, first define what content needs to be organized, who will use it, and what users need to find. Then choose methods based on your research questions. Use card sorting to explore how users understand and group resources, and tree testing to check whether proposed categories help people find important items. Usability testing can provide additional findings. This research is not just about naming menus; it examines whether the way information is organized meets users’ specific needs when looking for information.

ProcrustesPRProcrustesPRDoes my website need a speed score of 100?

Define the research scope and tasks first

The research should focus on a clearly defined set of content and its categories, rather than the broad question of whether an entire website is easy to use. When planning, start by listing:

  • Content scope: Which resources, categories, and levels will this study examine?
  • Target users: Which users do you want to study to understand how they categorize these resources?
  • Tasks: Which important items do users need to find?

For example, when studying a food website, you could set “find tomatoes” as a task and observe which category users choose. This is only an example task; you cannot assume in advance that everyone will follow the same path.

Use card sorting to explore how to organize content, but watch for selection bias

If you want to know which pieces of content users would group together, card sorting can help you explore how they naturally categorize resources. This provides clues about how to organize content; tree testing can then check whether the proposed structure helps people find items.

The content selected for the cards also affects the results. NN/g’s 2022 research guide notes that selecting card items in proportion to the types of content actually available helps avoid research bias. You should therefore avoid selecting content from just one category and generalizing the sorting results to all resources.

Use tree testing to find where users get lost

Tree testing can check whether users can find important items within a proposed information architecture. Look not only at success rates but also at users’ first clicks and final selections. A 2024 NN/g article identifies two patterns worth investigating:

  • First clicks are spread across multiple categories for many tasks: There may be too much overlap between categories. Consider card sorting or alternative ways of organizing the content.
  • First clicks are correct, but success rates are low: Examine users’ final selections; lower-level categories may overlap.

These are only diagnostic clues; scattered clicks alone do not establish the cause. Tree testing results do not necessarily reflect how users will perform in the actual interface. The article notes that menus displaying multiple category levels at once may improve performance, provided users first enter the correct menu.

Use the same tasks when comparing candidate structures

If you have two or more options, comparisons typically use a “between-subjects design”: different participants test each version using the same tasks. You can then compare success rates, time taken, how direct the paths are, and metrics for specific paths.

Higher or lower metric values alone do not establish that one version is better. Quantitative comparisons must also account for the sample size needed and whether the differences are statistically significant. The 2024 NN/g article cautions that most tree testing software does not automatically perform significance tests. General usability testing can also provide additional findings about information architecture.

ProcrustesPRProcrustesPREight SEO essentials for beginners: Website speed

Information architecture research: perspectives from across the web

Does the same content need to appear under multiple categories?

Not necessarily. If several categories count as correct answers for a resource, you can observe how often users actually choose each category. If one correct category is rarely or never selected, it may not be necessary to keep that resource listed under that additional category. This judgment should apply to the resources and categories tested, rather than imply that all content needs only one location.

How many participants do you need to compare two structures?

A 2024 NN/g article states that quantitative comparisons typically require at least 50 participants per structure to obtain reasonably narrow confidence intervals and identify statistically significant differences, so this type of research can be relatively expensive. This is the article’s guidance on a typical study size, not a guarantee that reaching that number will reveal a significant difference.

ProcrustesPRWhat Ten Years of SEO Taught Me About Its Real Purpose

References

  1. Tree Testing Part 2: Interpreting the Resultswww.nngroup.com · 2024-01-19T17:00:00+0000 — Discusses: Does the same content need to appear under multiple categories?; How many participants do you need to compare two structures?
  2. Information Architecture: Study Guidewww.nngroup.com · 2022-04-10T16:00:00+0000 — Source used in the article above.

Related articles

Does my website need a speed score of 100?Eight SEO essentials for beginners: Website speedWhat Ten Years of SEO Taught Me About Its Real PurposeHow Do RWD and AWD Work, and Which Should You Choose for Your Website?What are the benefits and development challenges of SPAs?How do UTF-8 and Big5 differ in their effects on search rankings and user experience?What are single-page websites best suited for, and what are their pros and cons?Keep different topics on one site or split them up? Start with audience overlap and maintenance capacity
ProcrustesPR
ProcrustesPR