Card sorting is one of the most powerful — and underused — UX research methods. It reveals how real users think about your content, navigation, and information architecture.

Whether you're redesigning a website, building a new app, or trying to fix confusing navigation, card sorting gives you data-driven answers instead of guesswork.

In this complete guide, you'll learn what card sorting is, when to use it, the different types, how to run a session, and how to analyze results — with practical examples from real projects.

What Is Card Sorting?

Card sorting is a UX research method where participants organize topics, labels, or content into groups that make sense to them. Each "card" represents a piece of content, feature, or navigation item.

The goal is simple: understand how your users mentally organize information so you can build navigation and structures that match their expectations.

Why Card Sorting Matters

| Problem | How Card Sorting Helps | |---------|----------------------| | Users can't find what they need | Reveals natural mental models | | Navigation feels confusing | Shows how users expect items grouped | | Team disagrees on structure | Provides objective data for decisions | | Redesigning information architecture | Creates user-centered categories | | Labeling is inconsistent | Tests which labels users understand |

Card sorting works because it removes assumptions. Instead of your team deciding where "Account Settings" should live, you let 20+ users show you where they expect to find it.

Three Types of Card Sorting

1. Open Card Sorting

Participants sort cards into groups they create themselves and name the groups.

Best for:

Early-stage research when you have no existing structure
Discovering how users naturally categorize information
Generating ideas for navigation labels

Example: You give participants 40 cards with feature names from your SaaS product. They create groups like "Account Stuff," "Reports," and "Team Management" — revealing categories you hadn't considered.

2. Closed Card Sorting

You provide predefined categories and participants sort cards into them.

Best for:

Validating an existing navigation structure
Testing whether your categories make sense
Comparing two proposed structures

Example: You have 5 navigation categories already designed. Participants sort 30 content items into those categories. If "Billing History" ends up in 3 different categories, your labels need work.

3. Hybrid Card Sorting

Participants sort cards into predefined categories but can also create new ones.

Best for:

Testing a proposed structure while staying open to improvements
Mid-stage design when you have a rough idea but want validation

Example: You provide 4 categories but let participants add more. If 70% of users create a "Help & Support" category you didn't include, that's a clear signal.

When to Use Card Sorting (and When Not To)

Use Card Sorting When:

Building or redesigning navigation — Get structure right before development
Creating a content strategy — Understand how users think about your topics
Merging two products — Find a unified structure that works for both user bases
Localizing a product — Cultural differences affect mental models (Arabic-speaking users may categorize differently than English-speaking users)
Onboarding is confusing — Find out why users can't find features

Don't Use Card Sorting When:

You need to test task completion (use tree testing or usability testing instead)
Your content has fewer than 10 items (too simple for card sorting)
You need to test visual design (card sorting is about structure, not aesthetics)

How to Run a Card Sort: Step-by-Step

Step 1: Define Your Research Question

Start with a clear question:

"How do users expect our help center articles to be organized?"
"Which navigation structure works better for our dashboard?"
"Do users understand our feature categories?"

Step 2: Prepare Your Cards

Guidelines for creating cards:

Use 30–60 cards for most studies (fewer than 20 gives weak data; more than 80 causes fatigue)
Write cards in plain language — avoid jargon
Each card should represent one concept (not "Billing & Invoicing" — split into "View Invoices" and "Update Payment Method")
Include cards that are intentionally tricky — these reveal the most about mental models

Step 3: Choose Your Participants

Recruit 15–30 participants for statistically meaningful results. Fewer than 15 and you'll miss patterns; more than 50 gives diminishing returns.

For the best results:

Recruit from your actual target audience
Include a mix of new and experienced users if relevant
For bilingual products, run separate sessions per language — mental models differ across cultures

Step 4: Run the Session

For remote card sorting (recommended):

Send participants a link to your card sorting tool
Include brief instructions: "Sort these items into groups that make sense to you"
For open sorts, ask them to name their groups
Set a reasonable time limit (15–25 minutes)
Add a follow-up question: "Was anything confusing or hard to place?"

For in-person sessions:

Print cards on index cards or sticky notes
Provide a large table or wall space
Observe silently — don't guide or suggest
Take photos of the final arrangement
Ask participants to explain their reasoning

Pro tip: Run a pilot test with 2–3 people first. You'll catch confusing cards, missing items, or unclear instructions before your real study.

Step 5: Analyze Results

This is where card sorting gets powerful. Look for:

Agreement patterns:

Which cards consistently end up in the same group? These belong together.
Which cards are placed in different groups by different people? These need clearer labels or might belong in multiple places.

Category patterns (open sort):

What names do participants give their groups?
Are certain groups created by 70%+ of participants? Those should be in your navigation.
Do any groups have only 1–2 cards? These might not need their own category.

Similarity matrix: Create a matrix showing how often each pair of cards was grouped together:

Pairs grouped together by 80%+ participants → definitely belong together
Pairs grouped together by 40–60% → consider placing together
Pairs grouped together by less than 20% → keep separate

Step 6: Create a Dendrogram

A dendrogram (tree diagram) visualizes the clustering of your cards. It shows which items are most closely related and where natural groupings form.

Most card sorting tools generate dendrograms automatically. Look for:

Tight clusters — items that almost always appear together
Loose clusters — items that sometimes group together (candidates for cross-linking)
Orphans — items that don't consistently fit anywhere (need rethinking)

Best Practices for Better Results

Writing Better Card Labels

| ❌ Avoid | ✅ Better | Why | |----------|----------|-----| | "Manage subscriptions and billing" | "View subscription" + "Update payment" | One concept per card | | "CRM Integration" | "Connect your CRM" | Plain language | | "Settings" | "Account settings" | More specific | | "Analytics" | "View website traffic" | Action-oriented |

Avoiding Common Mistakes

Too many cards — 80+ cards cause fatigue and random sorting
Leading category names — In closed sorts, don't use names that make the answer obvious
Mixing levels of specificity — Don't combine "Email" with "Communication Tools" as separate cards
Ignoring outliers — A user who sorts differently might represent an important persona
Running only one type — Use open sorting first to explore, then closed sorting to validate

Card Sorting for Arabic and Bilingual Products

When building products for MENA markets, card sorting requires special consideration:

Cultural mental models differ:

Arabic-speaking users may group "Help" and "Contact" differently than English speakers
Financial terms often have different organizational expectations across cultures
Navigation expectations can differ between RTL and LTR interfaces

Recommendations for bilingual card sorting:

Run separate sessions for each language — don't translate cards mid-session
Use culturally appropriate examples — not direct translations
Compare results across languages to find universal patterns and cultural differences
When patterns conflict, prioritize the primary language audience

With Afkar, you can run card sorting studies with native Arabic-speaking participants who understand local context — not just translated tasks.

Analyzing and Presenting Results

For Stakeholders

Present card sorting results as actionable recommendations:

"Here's what users expect" — Show the top 5 groupings with agreement percentages
"Here's where users struggle" — Highlight cards with no clear home
"Here's our recommendation" — Propose a structure based on the data
"Here's how we'll validate" — Follow up with tree testing to confirm

Turning Results into Navigation

Use clusters with 70%+ agreement as primary categories
Cards sorted into multiple categories should appear in navigation shortcuts or cross-links
Name categories using the most common labels participants created (open sort)
Test the new structure with a tree test before building

Tools for Running Card Sorts

You can run card sorts using:

Dedicated research platforms like Afkar that support Arabic and English
Physical cards for in-person sessions
Spreadsheet-based analysis for small studies

The best tool is the one your participants can actually use. For remote studies with Arabic-speaking users, use a platform that supports RTL interfaces natively.

What Comes After Card Sorting?

Card sorting is most powerful as part of a research sequence:

Card sorting → Discover natural groupings
Tree testing → Validate the proposed structure
Usability testing → Confirm users can complete tasks in the new structure

Don't stop at card sorting alone. The groupings tell you how users think — tree testing tells you whether your implementation of those groupings actually works.

Key Takeaways

Card sorting reveals how users mentally organize your content
Use open sorting to explore, closed sorting to validate
Recruit 15–30 participants for reliable patterns
Look for agreement percentages — 70%+ means strong signal
Always follow up with tree testing to validate your new structure
For bilingual products, run separate sessions per language
Present results as actionable recommendations, not raw data

Ready to run your first card sort? Create a free card sorting study on Afkar and get real participant feedback in hours, not weeks.