A focus group is a qualitative research session where 6-10 participants discuss a topic together, guided by a moderator. The format generates rich conversational data fast, but it comes with a well-known liability: group dynamics distort individual opinions. In UX research, focus groups work best as an input-gathering tool, not a decision-making one.
Focus groups have fallen out of favor in product design circles, and for good reason — they're expensive, logistically painful, and vulnerable to bias. But they still have a specific, narrow use case: generating raw material for further research.
Say you're redesigning a help center. You don't know what words your users use to describe their problems. Running a focus group with 8 customer support agents reveals that customers say "refund" while your internal team says "credit adjustment." That vocabulary gap would torpedo your card sort labels if you hadn't caught it. The group format works here because participants build on each other's examples: one agent mentions "refund," another adds "chargeback," a third brings up "store credit." You walk out with 30 candidate card labels in 90 minutes.
Where focus groups fail is anywhere social pressure matters. Ask a group "Would you look for billing under Account Settings or Payments?" and you'll get consensus — but it's the consensus of whoever spoke first and sounded most confident. That's not data. That's one person's opinion with an audience.
Recruit deliberately. Mix roles and experience levels within each group, but keep groups internally coherent enough that participants can relate to each other. Don't put C-suite executives in the same room as junior support agents — the power dynamic kills candid discussion.
Write a discussion guide, not a script. Plan 5-7 open-ended questions that move from broad ("How do you currently find information on our site?") to specific ("What would you call a section about changing your subscription?"). Leave room for follow-up questions based on what participants actually say.
Manage the room. Your moderator's primary job is equalizing airtime. Directly invite quiet participants to share: "Sarah, you looked like you had a reaction to that — what are you thinking?" Cut off dominators politely but firmly. A good moderator produces data from 8 people, not a monologue from 1.
Record everything. Transcribe sessions and code them for recurring themes, terminology, and disagreements. The disagreements are often more valuable than the agreements — they reveal where your IA has genuine ambiguity that a card sort needs to resolve.
The most productive workflow uses focus groups upstream of card sorting, never downstream. Run focus groups first to generate card labels, identify content gaps, and understand how different user segments frame the same tasks. Then take those labels into a card sort where each participant works individually, free from social influence.
This two-step approach gives you the best of both methods: the generative richness of group discussion with the unbiased categorization data of individual sorting. Skip the focus group, and you risk building a card sort with labels that make sense to your team but confuse your users. Skip the card sort, and you're building IA on group opinions instead of individual mental models.
One thing focus groups cannot do: tell you how users will actually navigate your product. For that, you need moderated testing or tree testing with individual participants making real choices.
How many participants should be in a focus group? A focus group typically includes 6-10 participants. Fewer than 6 limits the range of perspectives and can feel like an awkward interview. More than 10 makes it difficult for everyone to contribute and increases the risk of dominant personalities steering the conversation. For UX research, running 3-4 separate groups of 6-8 participants provides enough data to identify recurring themes.
When should you use a focus group instead of individual interviews? Focus groups work best when you want participants to react to each other's ideas and build on shared experiences, such as generating vocabulary for card labels or exploring how different user segments talk about the same concepts. Use individual interviews when you need unbiased individual opinions, when the topic is sensitive, or when you want to avoid groupthink influencing responses.
Can focus groups replace card sorting studies? No. Focus groups generate qualitative insights about how users think and talk about content, but they cannot replace the structured data that card sorting provides. Focus groups are useful for generating card labels and identifying content gaps before a card sort, but the actual categorization and information architecture decisions should be validated through card sorting with individual participants to avoid group bias.
Explore related concepts, comparisons, and guides