WebsiteDesignOutsource.com research

Evidence for Information Architecture Decisions in Website Design

A research approach to navigation labels, card sorting, and tree testing for outsourced website projects.

Navigation is often approved by taste, even though it determines whether people can find important content. Information architecture research can make the decision more observable by separating label comprehension, category placement, and end-to-end task success. This article examines evidence methods that an outsourced website design team can use while keeping sample claims modest.

Card sorting measures grouping behavior

In an open card sort, participants create groups and name them. In a closed sort, they place items into supplied categories. The two designs answer different questions. Open sorting can reveal language and mental models, while closed sorting can test a proposed structure. A result is meaningful only with the cards, instructions, participant profile, exclusions, and analysis method recorded.

Agreement is not the same as correctness. Participants may share a label that is familiar but not useful for a particular task. A design decision should therefore compare grouping evidence with the content’s purpose and the organization’s vocabulary. The external team can report patterns, while the company chooses which audience and business meanings must be preserved.

Tree testing measures findability

Tree testing removes visual design and asks participants where they would find an item in a hierarchy. It can reveal whether labels and nesting support a specified task. Success rate, directness, and selection path are useful observations, but they depend on the task wording and the tree shown. A result for one tree cannot automatically be compared with another tree when tasks or participants differ.

The useful unit is a task attempt by a defined cohort. Report the number of participants, the task, the tree version, and the outcome definition. Avoid declaring that the navigation is universally intuitive when the research examined a limited audience and vocabulary.

Labels carry content promises

A label is not only a menu caption. It predicts what a visitor will find after selection. Labels that are concise but vague may reduce confidence. Labels that use internal language may be accurate to staff but unfamiliar to prospective visitors. Reviewers should compare label meaning with the destination page, search terms, and any known audience language.

Findings

Architecture evidence is strongest when grouping, findability, and destination comprehension are considered together. A proposed menu should be tested with real tasks, then reviewed against content ownership and search behavior. The evidence can support a decision about a version of a structure at a given time, not a permanent truth about every visitor.

Limitations

Small samples can surface severe problems without estimating population rates. Recruitment affects vocabulary. Artificial trees omit visual hierarchy and page content. Tasks can lead participants. Existing brand language may constrain a label even when a test prefers another term.

Conclusion

Use card sorting to explore grouping, tree testing to examine findability, and task review to validate destinations. Preserve the method and cohort. This lets an outsourced design project explain why a navigation choice was made without overstating research precision.

Task wording can bias the result

A task that repeats the exact label being tested may make success look stronger than ordinary browsing. A task that is too vague may measure guessing rather than findability. Write the scenario with enough context to represent the visitor’s goal while avoiding the target label. Record the wording, order, time limit if any, and whether hints were allowed. A later study can then distinguish a change in tree performance from a change in the prompt.

Structure must meet content reality

Navigation research should be compared with the pages that will actually exist. A proposed category can test well with abstract card names but fail when the page titles are long, duplicated, or written in internal language. Review the destination headings, page purpose, and cross-links with the structure. Findability is an interaction between labels and content, not a property of the menu alone.

Evidence across audiences

Different cohorts may use different terms for the same service. If the audience includes customers, partners, or internal reviewers, record which cohort produced each pattern. Do not average vocabulary into a single winner when a meaningful minority has a distinct need. A final structure may use a broad label, a supporting explanation, or multiple paths to the same content.

Decision record

The final rationale should state which evidence was decisive, which evidence conflicted, and what remains to be observed after launch. This prevents a research study from being treated as a permanent authority. It also gives the outsourced team clear guidance when a new page does not fit the original categories.

Search and navigation are complementary

Navigation evidence should include the content that will populate the structure. If a proposed category has one page today but many pages later, the label and nesting may age differently. Record expected scope and revisit the structure when the content inventory changes. This is especially important for an outsourced project in which new contributors may not have attended the original research sessions. A new page should be classified against the tested structure rather than placed in a catch-all category by default.

Visitors may arrive through a search result, an internal link, or a menu. The page should make its purpose clear in each case. A navigation label cannot repair a destination with an unclear heading, and a strong heading cannot compensate for a route that is absent from the intended paths. Review the entry context and the next useful action together.

Keeping the structure current

Navigation evidence should include the content that will populate the structure. If a proposed category has one page today but many pages later, the label and nesting may age differently. Record expected scope and revisit the structure when the inventory changes. New pages should be classified against the tested structure rather than placed in a catch-all category by default. This is especially important when new contributors did not attend the original research sessions. Preserve vocabulary changes so later testing can distinguish a structure problem from a label problem.

Keeping the structure current

Navigation evidence should include the content that will populate the structure. If a proposed category has one page today but many pages later, the label and nesting may age differently. Record expected scope and revisit the structure when the inventory changes. New pages should be classified against the tested structure rather than placed in a catch-all category by default. Preserve vocabulary changes so later testing can distinguish a structure problem from a label problem.

Sources

1. Nielsen Norman Group card sorting Card sorting methods.

2. Nielsen Norman Group tree testing Tree testing methods.

3. W3C Design of Web Navigation Navigation accessibility context.

4. GOV.UK content design Plain language and user needs.

5. Google Search Central SEO starter guide Descriptive links and discoverability.

6. ISO 9241-210 overview Human-centred design context.

7. Nielsen Norman Group task analysis Task framing.

8. W3C Understanding navigation Multiple ways to locate content.

9. GOV.UK navigation Contents navigation context.

10. Nielsen Norman Group navigation Information architecture context.

Further reading

Content migration findability

Accessibility conformance evidence

Related Research

Redirect measurement

Form recovery evidence

Structured data eligibility

Frequently asked questions

How many participants are required?

That depends on the decision and study design. Report the actual cohort and avoid turning a formative study into a population estimate.

Is card sorting enough?

No. It examines grouping and labels, while tree testing and task review examine findability and destination meaning.

Should internal language always be removed?

No. Compare it with audience comprehension and the purpose of the destination before changing it.

Ready to plan your next step?

Contact WebsiteDesignOutsource.com