Information Architecture and Card Sorting

Information architecture, often shortened to IA, is the practice of organizing, structuring, and labeling content so people can find what they need and understand where they are. It's the invisible foundation under navigation menus, search, page layouts, and sitemaps.

When IA is good, people move through a product confidently. When it's poor, they get lost, use search as a last resort, or contact support for things that were there all along. This guide covers the core concepts and the research methods that make IA decisions evidence-based.

The core questions of IA#

Every IA project answers a few questions:

What content and features exist?
How should they be grouped?
What should each group be called?
How will people move between them?

Visual design can make navigation look polished, but it can't fix content that's grouped illogically or labeled in ways users don't understand.

Step 1: Take a content inventory#

Before reorganizing anything, list what you have. A content inventory is a spreadsheet of every page, section, feature, or content type, with details like:

Title and location.
Purpose and audience.
Owner.
Traffic or usage, where available.
Notes on quality or duplication.

For existing products, an inventory almost always reveals duplicate pages, outdated content, and features buried in unexpected places. It's also the input for card sorting.

Step 2: Understand user mental models#

People have expectations about how information should be organized, shaped by their experience, their work, and other products they use. Good IA aligns with those mental models rather than with the company's internal structure.

A classic mistake is organizing a website by department — because that's how the company thinks — when customers think in terms of tasks or needs. Research methods like card sorting reveal how users actually group things.

Card sorting#

In a card sort, participants organize a set of items — written on cards, physical or digital — into groups that make sense to them.

Open card sorting#

Participants create and name their own groups. This is useful early on, when you want to discover how people naturally categorize content and what words they use for the categories.

Closed card sorting#

Participants sort items into predefined categories. This is useful for testing whether a proposed structure works — do people consistently place items where you expect?

Hybrid card sorting#

Participants sort into predefined categories but can create new ones if nothing fits. This is a practical middle ground.

Running a card sort#

Choose 30 to 60 items that represent the content. Too few won't reveal structure; too many fatigue participants.
Write clear item labels without using words that hint at a category.
Recruit representative users — often 15 or more for online card sorts to see reliable patterns.
Ask participants to think aloud if moderated, to understand their reasoning.

Analyzing results#

Look for:

Agreement: items most participants grouped together.
Disagreement: items placed in many different groups, which may need clearer labels or multiple access paths.
Category names: the words participants used, which are strong candidates for navigation labels.

Online card sorting tools often provide similarity matrices and dendrograms to visualize which items are frequently grouped together.

Step 3: Design the structure#

Using inventory and card sort findings, create a proposed hierarchy. Common structures include:

Hierarchical: broad categories with progressively specific subcategories. The most common structure for websites and apps.
Sequential: step-by-step paths, such as checkout or onboarding.
Matrix: content accessible through multiple dimensions, like products filterable by category, price, and brand.
Database-driven: content organized by metadata and surfaced through search and filters.

Most products combine several. A store might use a hierarchy for categories, filters for browsing, a sequence for checkout, and search across everything.

Breadth vs. depth#

A broad, shallow structure has many top-level categories and few levels. A narrow, deep one has few top-level categories and many levels. Generally, moderately broad structures perform better than deep ones, because each extra level is another chance for people to choose wrong. But very broad menus become overwhelming. Aim for clear, distinct categories at each level.

Step 4: Choose clear labels#

Labels are where IA succeeds or fails. Good labels are:

Familiar — the words users use, not internal jargon.
Specific — "Billing and invoices" rather than "Resources."
Distinct — categories that don't overlap in meaning.
Consistent — similar in grammatical form across a menu.

Vague labels like "Solutions," "Resources," or "More" are common because they're easy to agree on internally. They're hard for users because they don't predict what's inside.

Tree testing#

Tree testing evaluates a structure without any visual design. Participants see a text-only version of the hierarchy and are asked to find where they'd go to complete specific tasks: "Where would you find your last invoice?"

Tree testing reveals:

Success rate: whether people find the right location.
Directness: whether they go straight there or backtrack.
Where they go wrong: which categories attract incorrect clicks.

Because it removes visual design and search, tree testing isolates the structure and labels. It's ideal for validating a proposed IA before investing in design and development. Run it again after changes to measure improvement.

Once the structure is validated, design how people will move through it:

Global navigation for top-level sections.
Local navigation within a section.
Breadcrumbs to show location in deep hierarchies.
Contextual links connecting related content.
Search for direct access, especially in large content sets.

Navigation should make the current location clear and the next step obvious.

Step 6: Plan for growth#

IA needs to accommodate future content. Consider:

Where will new features or content types go?
Will categories still make sense with twice as much content?
Who owns decisions about adding new sections?

Documenting the IA and governance rules helps prevent gradual decay, where every new feature gets its own top-level menu item.

Common IA mistakes#

Organizing by internal departments instead of user needs.
Using vague or clever labels.
Creating overlapping categories.
Building very deep hierarchies.
Hiding important content in "More" or "Other."
Relying on search to compensate for poor structure.
Skipping validation with real users.

Structure is a design decision#

Information architecture is invisible when it works, which makes it easy to overlook. But it determines whether people can find what they came for, and that determines whether the rest of the design matters at all.

Inventory your content, learn users' mental models through card sorting, design a clear structure with familiar labels, validate it with tree testing, and plan for growth. Good IA turns a collection of pages into a product people can navigate with confidence.

Information Architecture and Card Sorting for Designers