Everything you need to know about card sorting: open sorts, closed sorts, hybrid sorts, analysis techniques, and how to turn results into better navigation. A comprehensive pillar guide for UX teams.
Card sorting is one of the most reliable methods in UX research. It answers a question that other methods cannot: How do users expect information to be organized?
Not how you think they expect it. Not how your team has always done it. How users actually, genuinely expect it — based on their mental models, not yours.
This guide covers everything: what card sorting is, when to use it, how to run a study, and how to turn raw results into better navigation and information architecture.
Card sorting is a UX research method where participants organize a set of labeled cards into groups that make sense to them, then optionally name those groups.
Each card represents a piece of content, a feature, or a page on your website or app. The sorting process reveals how your users mentally categorize information — which is often very different from how the product team categorizes it.
The core insight behind card sorting is this: your users do not know your internal labels, your product taxonomy, or your development history. They approach your product with their own vocabulary and their own expectations about where things belong.
Card sorting surfaces those expectations before you invest weeks building navigation that nobody can use.
Card sorting is the right method when you need to:
Participants sort cards AND create their own category names.
What you learn:
Best for:
Sample size: 15–30 participants for reliable patterns
Time per session: 15–25 minutes
Participants sort cards into categories you define.
What you learn:
Best for:
Sample size: 20–30 participants
Time per session: 10–20 minutes
Participants sort cards into predefined categories but can also create new ones.
What you learn:
Best for:
Sample size: 20–30 participants
Time per session: 15–25 minutes
What are you trying to organize? Pick a specific section of your site or app rather than everything at once:
Card writing is where most card sorting studies succeed or fail.
Rules for good cards:
Sample card list for an e-commerce site:
| Card | Notes |
|---|---|
| My Orders | Core navigation |
| Return an Item | Edge case — could be "My Orders" or "Help" |
| Wishlist | Should this be "My Account" or top-level? |
| Payment Methods | Ambiguous — "Settings" or "My Account"? |
| Notifications | Often under-considered in IA |
| Refer a Friend | Marketing vs. Account |
| Help Center | Top-level or under "Support"? |
| Privacy Settings | Account vs. Settings |
| Your Situation | Recommended Method |
|---|---|
| No existing navigation | Open sort |
| Testing a proposed redesign | Closed sort |
| Adding to existing navigation | Hybrid sort |
| Comparing two navigation structures | Two separate closed sorts |
| Have a hypothesis but want validation | Hybrid sort |
Who to recruit:
Screening questions:
Where to find participants:
Remote card sorting (recommended for most studies) requires a tool that:
Afkar's card sorting supports Arabic-first studies with full RTL rendering and an integrated participant panel.
Setup:
Launch and Monitor:
Add a think-aloud component to your card sort:
This adds rich qualitative data alongside the clustering patterns.
The similarity matrix is your primary analysis tool. It shows, for every pair of cards, what percentage of participants grouped them together.
Interpreting the matrix:
| Similarity | Interpretation |
|---|---|
| >70% | Strong association — these belong in the same category |
| 40–70% | Moderate association — likely the same category |
| 20–40% | Weak association — may need a bridge or subcategory |
| <20% | No consistent association — these cards belong in different places |
What to look for:
A dendrogram is a tree diagram that visualizes how items cluster together. Read it from right to left (or left to right depending on your tool):
Use dendrograms to:
In open sorts, participants create their own category names. Analyze these to:
After an open sort, standardize participant categories into a working IA:
Use the dendrogram cut at the level that produces 5-8 clusters. This typically becomes your main navigation.
For each item, assign it to the category where it has the highest similarity score. Flag items with:
For items that do not clearly belong anywhere:
Card sorting tells you how to organize. Tree testing validates whether the organization actually works.
Create your proposed navigation hierarchy from the card sort results, then run a tree test where participants find items by navigating the menu structure.
Key tree testing metrics:
Aim for >80% success rate on key tasks before finalizing your IA.
If your product serves both Arabic and English users, card sorting requires extra consideration.
Always run separate card sorts for Arabic and English users — never just translate one study. Arabic and English speakers categorize information differently because:
After running both studies, compare your category structures:
| Category Type | Approach |
|---|---|
| Direct match | Use same structure, translate labels |
| Partial match | Same top-level, different sub-categories |
| Arabic-only | Include in Arabic navigation only |
| English-only | Include in English navigation only |
When you build your navigation, run tree tests in both languages to verify each version works independently.
1. Too many cards. Over 50 cards leads to fatigue and random sorting. Be ruthless about scope.
2. Vague card labels. "Settings" means different things to different people. "Notification Preferences" is specific.
3. Treating open sort as gospel. Card sort results show patterns, not instructions. Use judgment when clusters are messy.
4. Skipping tree testing. Card sorting tells you how users think. Tree testing tells you whether your resulting IA works. Do both.
5. Running only one language. For bilingual products, running only the English sort and translating is a critical mistake.
6. Too few participants. Under 15 participants produces unreliable patterns. 20+ is the minimum for meaningful similarity matrices.
Card sorting is the foundation of information architecture design. It removes the guesswork from navigation by revealing how real users think about your content — before you build anything.
When to use it: IA design, navigation redesign, bilingual products
How many participants: 15–30 per language version
What to analyze: Similarity matrix, dendrograms, category names
What to do next: Tree testing to validate the resulting IA
Tools: Afkar card sorting for Arabic-first studies with MENA participants
The best navigation feels invisible. Users find what they need without thinking about the structure. Card sorting is how you get there.
