{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-05-21T02:07:59.459Z"},"content":[{"type":"documentation","id":"43be2478-8d9f-4e28-88ca-41b765dc37b1","slug":"card-sorting-guide","title":"Card Sorting: The Complete Guide to Information Architecture Research","url":"https://www.koji.so/docs/card-sorting-guide","summary":"This guide covers card sorting — a UX research technique where participants organize labeled cards into groups to reveal their mental models. It explains open, closed, and hybrid card sort variants, how to design a study, determine sample sizes, analyze results using similarity matrices and dendrograms, and how to combine card sorting with follow-up qualitative interviews for deeper understanding.","content":"\nCard sorting is one of the most underused research methods in product development. It takes 30–60 minutes per session, requires no special equipment, and reveals how your users mentally organize information — insights that directly improve navigation, taxonomy, and content strategy.\n\n## What Is Card Sorting?\n\nCard sorting is a participatory research technique where participants organize labeled \"cards\" (representing content, features, or topics) into groups that make sense to them. The resulting groupings reveal participants' mental models — how they think about and categorize the domain you are researching.\n\nCard sorting directly informs:\n- Website and app navigation structure\n- Feature categorization and labeling\n- Terminology and taxonomy decisions\n- Content strategy and organization\n\nThe technique has been central to UX and information architecture practice for decades. According to Nielsen Norman Group, card sorting is one of the highest-ROI research methods available because the effort is low and the structural insights are immediately actionable.\n\n## Open vs. Closed vs. Hybrid Card Sorting\n\nThe three variants of card sorting answer different research questions:\n\n**Open card sort:**\nParticipants create their own groups AND name them. Use this for discovery — when you want to understand how users naturally categorize content and what language they use. Best for: designing information architecture from scratch, understanding user mental models, improving taxonomy and labeling.\n\n**Closed card sort:**\nParticipants sort cards into predefined categories. Use this to validate an existing structure — does your current navigation match how users think? Best for: evaluating an existing site structure, confirming that a redesign matches user expectations, comparing two organizational approaches.\n\n**Hybrid card sort:**\nParticipants can use predefined categories OR create new ones. Use this when you have a starting structure but want to know where it breaks down. Best for: iterating on an existing information architecture, identifying which categories work and which create confusion.\n\n## When to Use Card Sorting\n\nCard sorting is most valuable:\n\n- **When designing a new navigation structure**: Before building, run an open sort to let users show you how they would organize content\n- **When a site has high navigation failure rates**: Users cannot find what they are looking for — a closed sort reveals where your structure diverges from user expectations\n- **When adding new features or content**: Where does a new feature belong? A hybrid sort helps decide\n- **After major content migrations**: Ensure the new structure still makes sense to users\n\nCard sorting is less useful for evaluating visual design, understanding user goals and motivations, or capturing nuanced attitudes and feelings. For those, qualitative interviews are the right tool. Many researchers combine card sorting (for structural clarity) with follow-up interviews (for contextual understanding) — a combination that platforms like Koji make easy to run in sequence.\n\n## How to Design a Card Sort Study\n\n### Step 1: Define the Research Questions\n\nWhat do you need to decide based on this study?\n- \"Should we combine these two navigation sections or keep them separate?\"\n- \"What do users call this concept — projects, workspaces, or campaigns?\"\n- \"Does our current site structure match how users think about these topics?\"\n\nSpecific questions produce actionable results. Vague questions produce interesting but unusable data.\n\n### Step 2: Select Your Cards\n\nThe ideal card set is 30–100 items. Fewer than 20 provides limited signal; more than 100 causes fatigue.\n\nCards should represent:\n- Top-level content or features in your product\n- Topics users actually encounter — no internal jargon or system terminology\n- Similar items across potential categories so you can detect how users distinguish them\n\nLabel cards clearly and consistently. Ambiguous labels introduce noise — participants will sort based on how they interpret the label, not the underlying content.\n\nCommon mistake: including too many cards. When participants are overwhelmed, they sort more quickly and with less care. Quality over quantity is the rule.\n\n### Step 3: Choose Your Moderation Style\n\n**Unmoderated card sorting** is the most common approach. Participants complete the study independently via tools like OptimalSort, Lyssna, or Maze. This allows for larger sample sizes (30–100 participants) and lower cost per participant.\n\n**Moderated card sorting** adds a researcher who observes and asks think-aloud questions. This produces richer data about WHY participants organized items the way they did, but is more time-intensive.\n\nFor most information architecture decisions: start with unmoderated sorts for pattern detection, then run 3–5 moderated sessions to understand the reasoning behind the patterns. The moderated sessions are where you ask follow-up questions that reveal the mental models driving sorting behavior.\n\n### Step 4: Determine Your Sample Size\n\n- **Unmoderated open sorts**: 15–30 participants typically produce stable results\n- **Closed sorts**: 20–50 participants gives statistical reliability\n- **Per segment**: If you have meaningfully different user types, recruit 15–20 per segment — their mental models may differ significantly\n\n### Step 5: Analyze the Results\n\nCard sort analysis uses several complementary techniques:\n\n**Similarity matrix**: A grid showing how often pairs of cards were grouped together by participants. Pairs with high co-occurrence belong together in your structure.\n\n**Dendrogram**: A hierarchical tree diagram showing how cards cluster. The closer two cards appear in the dendrogram, the more often participants grouped them together. Most card sort tools generate this automatically.\n\n**Category agreement**: In closed sorts, how often did participants put each card in its expected category? Low agreement signals a structural mismatch between your navigation and user mental models.\n\n**Participant-generated category names**: In open sorts, the labels participants create reveal the language your users actually use — invaluable for navigation labels, menu item names, and microcopy.\n\n### Step 6: Follow Up with Qualitative Interviews\n\nNumbers tell you what happened; interviews tell you why. After analyzing your card sort data, you will likely have questions:\n\n- Why did participants split this topic between two categories?\n- What does \"settings\" mean to users — do they expect both billing and profile info there?\n- Why did 40% of participants create a catch-all category?\n\nRunning a follow-up interview study with Koji lets you probe these questions at scale. Distribute an AI-moderated interview to your card sort participants or a new representative sample, and Koji will conduct structured conversations exploring the patterns you observed. The AI automatically synthesizes themes across responses, so you get qualitative context without manually analyzing dozens of interviews.\n\n## Card Sorting vs. Tree Testing\n\nCard sorting and tree testing are often confused — they are actually two complementary techniques that work best in sequence:\n\n| Method | Purpose | When to Use |\n|--------|---------|------------|\n| Card sorting | Reveals how users organize information | Before you build your structure |\n| Tree testing | Tests whether users can navigate a structure | After you build your structure |\n\nUse card sorting to design your information architecture, then tree testing to validate it. The two together give you both the generative insight (how should it be organized?) and the evaluative confirmation (can users actually navigate it?).\n\n## Key Things to Know\n\n- **Remote card sorting is the norm**: Most card sorts are conducted remotely with tools like OptimalSort, Lyssna, or Maze\n- **Terminology matters enormously**: The label on a card significantly affects how it is sorted — test your labels with 2–3 participants before running the full study\n- **Demographic segments sort differently**: Power users and casual users often have different mental models; analyze by segment when relevant\n- **Randomize card order**: The sequence cards appear can bias sorting patterns; always randomize the presentation\n- **Expect inconsistency**: Some participants will sort idiosyncratically — look for statistical patterns, not unanimous agreement\n\n## Tips & Best Practices\n\n- **Write a clear intro screen**: Participants need to understand what a card represents and what you are asking them to do — poor instructions produce noisy data\n- **Invite participants to leave comments**: Many tools support freetext notes per group; these are often the most insightful data points in the entire study\n- **Do not sort by completion time**: Fast completion does not mean high quality — look for outlier categories and unusual groupings as signals worth exploring\n- **Combine with tree testing**: Card sorting tells you how to structure; tree testing confirms it works — run both before committing to a major navigation redesign\n- **Involve developers in analysis**: Sharing card sort findings with engineering helps explain why an IA decision was made, reducing pushback later\n\n## Related Articles\n\n- [UX Research Process: A Complete Framework](/docs/ux-research-process)\n- [Usability Testing Guide](/docs/usability-testing-guide)\n- [The Definitive Guide to User Interviews](/docs/user-interview-guide)\n- [Generative vs. Evaluative Research](/docs/generative-vs-evaluative-research)\n- [How to Analyze Qualitative Data](/docs/how-to-analyze-qualitative-data)\n\n## Frequently Asked Questions\n\n**Q: How many participants do I need for a card sort?**\nA: For unmoderated open sorts, 15–30 participants typically reach stable results. For closed sorts, 20–50 participants gives reliable pattern data. If you have multiple user segments, recruit 15–20 per segment.\n\n**Q: What is the difference between card sorting and tree testing?**\nA: Card sorting is generative — it reveals how users organize information. Tree testing is evaluative — it tests whether users can navigate a specific structure. Use card sorting to design your IA, then tree testing to validate it.\n\n**Q: Should I run open or closed card sorting first?**\nA: If you are designing a new structure from scratch, start with open sorting to let users show you their mental model. If you are evaluating an existing structure, use closed sorting. If you have a draft structure to test, hybrid sorting gives the best of both approaches.\n\n**Q: How do I analyze an open card sort?**\nA: Focus on three outputs: a similarity matrix showing which items are consistently grouped together, a dendrogram visualizing clustering patterns, and the category names participants created. The labels reveal user vocabulary and mental models that are directly actionable for your navigation design.\n\n**Q: Can I combine card sorting with qualitative interviews?**\nA: Yes — this combination is highly effective. Run unmoderated card sorts for statistical patterns, then use a platform like Koji to conduct follow-up AI interviews that probe the reasoning behind the patterns. The combination gives you both quantitative structure and qualitative understanding.\n\n**Q: How is card sorting different from user interviews?**\nA: Card sorting is a structured task-based method that reveals how users categorize information. User interviews are conversational and reveal goals, motivations, and mental models. They are complementary: card sorting gives you structural data, interviews give you contextual understanding.\n\n\n## Further reading on the blog\n\n- [Best AI Market Research Tools in 2026: The Complete Buyer's Guide](/blog/ai-market-research-tools-2026) — AI has fundamentally changed market research. This guide compares the leading AI market research platforms—from AI-native interview tools li\n- [B2B Customer Research: The Complete Guide for Product Teams (2026)](/blog/b2b-customer-research-guide-2026) — B2B customer research is harder than B2C — you are navigating buying groups of 10+ stakeholders, gatekeepers, and enterprise procurement cyc\n- [Best AI Tools for UX Research in 2026: The Complete Buyer's Guide](/blog/best-ai-tools-ux-research-2026) — The UX research AI tool landscape has exploded. This guide maps the best AI tools for every phase of the research workflow in 2026 — from pl\n\n<!-- further-reading:blog -->\n","category":"Research Methods","lastModified":"2026-05-13T00:26:36.807295+00:00","metaTitle":"Card Sorting Guide — Koji Docs","metaDescription":"The complete guide to card sorting for UX and IA research. Covers open, closed, and hybrid card sorts, sample sizes, analysis techniques, and combining card sorting with qualitative interviews.","keywords":["card sorting","card sorting UX","card sorting research","information architecture research","open card sort","closed card sort","card sorting methodology"],"aiSummary":"This guide covers card sorting — a UX research technique where participants organize labeled cards into groups to reveal their mental models. It explains open, closed, and hybrid card sort variants, how to design a study, determine sample sizes, analyze results using similarity matrices and dendrograms, and how to combine card sorting with follow-up qualitative interviews for deeper understanding.","aiPrerequisites":["ux-research-process","qualitative-vs-quantitative-research"],"aiLearningOutcomes":["Choose between open, closed, and hybrid card sorting based on research goals","Design a card sort study with the right card set and prompts","Analyze card sort results using similarity matrices and dendrograms","Combine card sorting with qualitative follow-up interviews for richer insight"],"aiDifficulty":"intermediate","aiEstimatedTime":"8 min read"}],"pagination":{"total":1,"returned":1,"offset":0}}