How Many Participants Do You Need for Card Sorting?

The short answer: 20 participants works for most projects. That's enough to see stable patterns in how people group your content, without blowing your budget or timeline.

If you need higher confidence — say, you're making a big navigation overhaul that affects millions of users — go to 30. If you're just checking whether your card labels make sense, 5 is fine for a pilot. And if you're comparing different user groups, you'll need 15–20 per group.

The rest of this guide walks through when each number makes sense, what the research says, and what it'll actually cost you.

Participant Requirements by Research Goal

For the typical UX project — redesigning a nav, validating an IA, cleaning up a help center — 20 participants gets you there.

Why 30 Is the Ceiling for Most Studies

Card sorting patterns tend to stabilize somewhere around 15–20 participants when your user group is fairly uniform. Tullis and Wood (2004) found that similarity matrices stop changing meaningfully after about 30 participants. Past that point, you're paying for data that tells you the same thing.

You can run a study with 100 people. The dataset will be large and impressive. But you'll draw the same conclusions you would have drawn at 30.

The exception: if you're comparing distinct user segments — say, doctors versus patients, or power users versus newcomers — you need enough people in each group. That means 15–20 per segment, so a two-group comparison needs at least 30–40 total.

Why Fewer Than 15 Gets Risky

With fewer than 15 participants, one or two odd sorters can throw off your whole similarity matrix. Patterns that look clear with 8 people sometimes vanish completely at 20. That's not a real pattern — it's noise.

Small samples give you false confidence. You see clean-looking clusters and think you've found something, but you've really just found the quirks of a handful of people. If you're going to make design decisions based on the results, 15 is the floor.

Start With a 5-Person Pilot

Before you recruit your full sample, run a quick pilot with 5 participants. You're not collecting usable data here — you're checking for problems.

A pilot will catch things like cards that confuse people, labels that mean different things to different participants, instructions that lead people astray, and platform bugs. It's cheap insurance. Fix the issues, then run the real study. You can also use AI-generated responses to test your analysis workflow before spending money on recruitment.

When You Actually Need More Than 30

Sometimes 30 isn't enough. Here's when to go bigger:

Comparing user segments. If you're looking at how different groups sort differently, you need 20+ per segment to see real differences. Think: different regions, expertise levels, or job roles.

Fuzzy content domains. When your cards cover abstract concepts, overlapping categories, or technical jargon, sorting variance goes up. You might need 40–50 participants to separate signal from noise in those cases.

What It Costs on Prolific

If you're recruiting through Prolific, expect to pay roughly $1.50 per completed response. Here's what that looks like:

A solid, reliable card sort study for under $60. That's within reach for freelancers and small teams, not just big research budgets.

Putting It All Together

For most projects, here's the playbook:

• 5 participants for your pilot. Fix whatever's broken.
• 15–20 participants for a standard UX project. This covers most nav redesigns and IA work.
• 30 participants when the stakes are high and you need confidence that'll hold up in a stakeholder presentation.
• 50+ participants when you're comparing multiple user segments or publishing academic research.

Start at 20 unless you have a specific reason to go higher. You can always add more participants later if the data looks noisy — but most of the time, 20 gets you clear, stable patterns.

Further Reading

• What is Card Sorting? Complete Guide
• Card Sorting (UX Glossary)
• Information Architecture (UX Glossary)
• How To Run Your First Card Sort Study

Frequently Asked Questions

How many participants do I need for card sorting with a limited budget? Try not to go below 15. With fewer than that, one weird participant can skew your entire similarity matrix, and you'll end up with groupings you can't trust. Fifteen is tight but workable — and on Prolific, that's about $30.

Can I reduce participant numbers if my content cards are exceptionally clear? Not really. Clear cards help reduce variance in how people sort, but you still need enough participants to tell real patterns apart from random noise. Fifteen remains the minimum regardless of how well-written your cards are.

When should I recruit separate participant groups for different user types? Whenever your users have fundamentally different relationships with your content. Experts vs. beginners, different markets, different professional roles — these groups will sort differently, and you need 15–20 people in each group to see those differences clearly.

What's the maximum useful number of participants for card sorting studies? For a single user group, around 50 is where you hit hard diminishing returns. After that, new participants just confirm what you already know. Your money is better spent on a follow-up study or tree test than on participant 51.

Should I add more participants if initial card sorting results appear unclear? Maybe, but first figure out why the results are unclear. Are your cards ambiguous? Do your users actually have conflicting mental models? Is the content just genuinely complex? More participants only help if the problem was an underpowered sample. If the study design has issues, more data won't fix that.