A focus group is a qualitative data-collection method in which a moderator leads a small, purposively assembled group of participants through a structured discussion, and the resulting interaction between participants — not just their individual answers — is treated as data. That last point is what separates a focus group from a set of individual interviews run in the same room: a well-run focus group generates disagreement, elaboration, and challenge between participants that no individual interview can produce, and analyzing that interaction is a distinct skill from coding interview transcripts.
This guide covers the full method: when a group discussion is the right choice over one-on-one interviews, how to decide on group size and composition, how to build a moderator’s guide, how to manage group dynamics (dominant talkers, silent participants, groupthink), and how focus group data is analyzed differently from interview data. For the individual-interview counterpart to this method, see CASRAI’s guide to interviews as a research method and the companion guide to designing semi-structured interviews.
When a focus group beats individual interviews
The choice between focus groups and individual interviews is not about which method is more rigorous — it is about what kind of data the research question actually needs. Individual interviews are the better choice when you need depth on one person’s private experience, when the topic is sensitive enough that group disclosure would suppress honest answers, or when you need each participant’s answer to be uncontaminated by what others said. Focus groups are the better choice when the research question is about shared or contested understanding — how a group makes sense of something together, where consensus breaks down, and what language or framing people reach for when talking to peers rather than a researcher.
| Situation | Better method | Why |
|---|---|---|
| Understanding how a community talks about and normalizes a shared experience | Focus group | Social interaction surfaces the shared vocabulary, norms, and disagreements that a one-on-one interview cannot elicit |
| Highly sensitive or stigmatized personal experience (trauma, illness, financial hardship) | Individual interview | Participants disclose more, and more honestly, without peers present |
| Testing reactions to a concept, product, message, or policy draft | Focus group | Group discussion reveals which framings generate agreement, confusion, or objection in real time |
| Needing an individually attributable, uncontaminated account of one person’s decision process | Individual interview | Group answers are influenced by what was said before them (conformity and social-desirability effects) |
| Limited fieldwork time and a need to hear from many participants quickly | Focus group | One two-hour session with eight participants covers more voices per researcher-hour than eight individual interviews |
| A power imbalance exists between participants (staff and their manager, patients and their clinician) | Individual interview | Mixing power levels in one group suppresses the lower-power participants’ honest input |
The two methods are also frequently combined in the same study: focus groups to map the range of views and shared language during an early, exploratory phase, followed by individual interviews to pursue depth on specific accounts that the group discussion surfaced. See CASRAI’s guide to triangulation in research for how combining methods like this strengthens a qualitative design.
Group composition and size
Two decisions dominate focus group design before a single question is written: how many participants per group, and how homogeneous each group should be.
Group size
Most methodological guidance, including the widely used practitioner reference by Richard Krueger and Mary Anne Casey, converges on a working range of six to ten participants per group, with many applied studies settling on six to eight. The reasoning is practical, not arbitrary:
| Group size | Trade-off |
|---|---|
| Fewer than 5-6 | Too little diversity of opinion to generate real discussion; the session behaves more like a small joint interview and is vulnerable to one or two participants dominating entirely |
| 6-8 (the common default) | Enough participants for a range of views to surface, small enough that every participant gets meaningful speaking time within a 60-90 minute session and the moderator can track who has and hasn’t spoken |
| 9-10 | Workable for topics where breadth of reaction matters more than depth per person; harder to give quieter participants space and to keep the discussion on track |
| More than 10-12 | Generally avoided — the session fragments into side conversations, transcription and interaction-analysis become substantially harder, and individual participants get too little airtime to contribute meaningfully |
Plan to recruit above your target size, since no-shows are routine — recruiting 8-10 to seat a group of 6-8 is a common working margin.
How many groups
A single focus group is rarely sufficient for a defensible qualitative claim, because you cannot tell whether what you observed is a property of the topic or a property of that particular group’s dynamics (a dominant personality, an unusually homogeneous or contentious mix). Applied qualitative research commonly runs three to five groups per distinct sub-population in the study, continuing until later groups stop producing genuinely new themes — the same logic of thematic saturation used in interview-based designs. If the study compares sub-populations (for example, clinicians and patients, or staff at different seniority levels), each sub-population typically needs its own set of groups rather than being mixed together, for the same power-imbalance reason noted above.
Homogeneity vs. heterogeneity
Within a group, most guidance favors relative homogeneity on the characteristics that matter for the research question — similar role, similar experience level, similar demographic position relative to the topic — while allowing enough variation in opinion and background to generate real discussion rather than simple agreement. Homogeneity on the wrong dimension (recruiting only people who already agree with each other) produces a group that confirms itself rather than surfaces the range of views the method exists to capture; homogeneity on the right dimension (comparable status or experience) is what allows participants to speak freely in the first place.
Building a moderator’s guide
A moderator’s guide is the focus group equivalent of an interview guide, but its structure is built around managing a group rather than a single respondent. The standard shape, sometimes called the funnel approach, moves from broad and easy to specific and probing:
| Segment | Purpose | Typical share of session time |
|---|---|---|
| Welcome and ground rules | Explain the purpose, confidentiality expectations, recording, and the norm that there are no wrong answers and disagreement is welcome | ~5% |
| Opening question | An easy, non-threatening question everyone can answer briefly (often a round-robin) to get every voice into the room before the topic gets harder | ~5% |
| Introductory questions | General orientation to the topic, still relatively easy | ~10% |
| Transition questions | Move the group from general orientation toward the specific research questions | ~15% |
| Key questions | The 3-5 questions the study most needs answered; the moderator’s guide should flag these explicitly so time is protected for them even if the session runs long | ~50% |
| Ending question | A summary or “anything we haven’t asked” question, often after a brief recap of what the moderator heard, to surface anything unsaid and to close the discussion deliberately | ~15% |
A practical worked example of a key-question sequence, illustrating the funnel from general to specific within a single topic block:
- Opening/general: “Tell us briefly about your experience with [the topic] — just a sentence or two to start.”
- Transition: “When you think about [the topic], what’s the first thing that comes to mind?”
- Key question: “What has made [the process/experience] easier, and what has made it harder?”
- Probe (used selectively, not scripted for every answer): “Can you say more about that?” / “Did others have a similar experience, or different?”
- Key question: “If you could change one thing about [the topic], what would it be, and why?”
Two design principles distinguish a good moderator’s guide from a good interview guide. First, questions are written to invite discussion between participants, not just answers to the moderator — phrasing like “what do others think about that?” is a deliberate tool, not an aside. Second, the guide is deliberately kept to a small number of key questions (typically 5-8 total, including openers and closers) because group discussion consumes far more time per question than an individual interview does; a guide written at interview-length density will not fit the session.
Moderating: managing group dynamics
Facilitation skill, not just question design, determines whether a focus group produces usable data. The moderator’s core job is to keep the discussion balanced and on-topic without steering participants toward a particular answer.
| Dynamic | Moderator technique |
|---|---|
| One or two participants dominate the discussion | Set the norm explicitly at the outset that everyone will be asked for their view; redirect with a direct, neutral prompt (“thank you — let’s hear from some others on this one”) rather than cutting the dominant speaker off abruptly |
| A participant stays silent or gives only minimal answers | Use direct, low-stakes invitations by name (“[Name], what’s your take?”) and round-robin formats for at least the opening and one key question, so silence is never mistaken for agreement |
| The group converges quickly on an apparent consensus | Probe for dissent explicitly (“has anyone had a different experience?”); rapid, unprompted consensus in a focus group is a known risk (a group-level version of social-desirability bias, sometimes discussed alongside groupthink) and should be treated as a signal to probe, not as a clean finding |
| The discussion drifts off-topic | Acknowledge the point briefly, then bridge back explicitly (“that’s useful — bringing it back to X for a moment…”) rather than ignoring it, which can shut down future contributions |
| A co-moderator or note-taker is present | Assign them to track speaker order, non-verbal reactions, and timing, freeing the primary moderator to focus entirely on the discussion itself |
Every session should be audio- or video-recorded with participant consent, and a note-taker (ideally separate from the moderator) should log who is speaking, not just what is said — this speaker-attribution record is what makes interaction analysis possible afterward. See CASRAI’s guide to research ethics in qualitative research for consent and confidentiality requirements specific to group settings, where confidentiality can only be requested of participants, not guaranteed by the researcher, since other participants also hear what is said.
Analyzing focus group data: interaction as data, not just content
This is the step where treating a focus group like a set of individual interviews produces a weaker analysis. Two layers of analysis are available, and a rigorous focus group study uses both.
Content analysis (what was said)
The transcript is coded for themes using the same qualitative coding logic used for interview data — see CASRAI’s guides to thematic analysis and coding qualitative interview data for the mechanics. The coding unit is typically the individual utterance or turn of speech, tagged by speaker so that patterns within and across groups can be compared.
Interaction analysis (how it was said, and in response to what)
What content analysis alone misses is the group-specific data: whether a view was volunteered unprompted or only emerged after another participant raised it, whether a statement was met with visible agreement, silence, or open disagreement from others in the room, and where the group’s language or framing shifted over the course of the discussion as participants built on or challenged each other. Analysts who study group interaction explicitly code for these sequences — agreement, disagreement, elaboration, and challenge between speakers — as a second, separate analytical pass over the same transcript, using the note-taker’s speaker-attribution record to reconstruct who was responding to whom. This is the layer of evidence a focus group is uniquely positioned to produce, since it does not exist in individual-interview data at all.
Unit of analysis: the individual or the group?
A closely related design decision is whether the group or the individual participant is the unit of analysis. Some studies report themes at the level of the group (“three of the five groups raised concern X”), which is appropriate when the research question is about shared or contested group-level understanding. Others still report at the individual level within groups, which is appropriate when individual-level variation matters but group setting was chosen for practical or disclosure reasons. This should be decided during design, not after transcription, since it affects how the moderator’s guide and note-taking are structured.
Limitations and known biases
| Limitation | What it means for the findings |
|---|---|
| Groupthink / conformity effects | Participants may shift their stated view toward the emerging group consensus rather than voice a genuine minority position, especially on socially charged topics |
| Dominant-participant bias | Without active moderation, the transcript can over-represent the views of the most vocal one or two participants relative to the quieter majority |
| Not generalizable in the statistical sense | Focus group findings describe the range and texture of views present in the groups studied; they are not a representative sample from which population-level proportions can be estimated — see CASRAI’s comparison of qualitative and quantitative research for why “how many people said X” is generally not a valid way to report focus group results |
| Recruitment and selection effects | People willing to talk about a topic in front of strangers are not a neutral cross-section of the population; who agrees to participate can itself bias what views are represented |
None of these limitations disqualify the method — they describe what it is and is not good evidence for, which is exactly why the “when a focus group beats individual interviews” comparison above matters at the design stage, not after data collection.
Frequently asked questions
How many people should be in a focus group?
Most guidance recommends six to ten participants per group, with six to eight as a common working default. Recruit above the target size (often 8-10 to seat 6-8) since no-shows are routine.
How many focus groups do you need for a study?
There is no fixed number, but three to five groups per distinct sub-population is a common range in applied qualitative research, continuing until additional groups stop producing new themes — the same saturation logic used for interview-based sampling.
What is the difference between a focus group and a group interview?
The terms are sometimes used loosely, but a focus group is specifically designed to generate and analyze interaction between participants — agreement, disagreement, and elaboration — as data, using a moderator trained to facilitate that interaction. A group interview that simply asks each participant the same question in turn, without using the group dynamic itself as a data source, is closer to several sequential individual interviews conducted in one room.
Can focus groups be conducted online?
Yes — video-conference focus groups are now common and follow the same design principles (group size, moderator’s guide, funnel structure). The practical adjustments are mainly in moderation technique: managing turn-taking without physical cues, muting/unmuting protocols, and being alert to unequal participation caused by connectivity or unfamiliarity with the platform rather than the topic itself.
How is focus group data different from interview data during analysis?
Focus group transcripts support a second analytical layer that individual interviews cannot: interaction analysis, which codes how participants responded to one another (agreement, disagreement, building on a prior comment) rather than only what each individual said. See the “analyzing focus group data” section above.
Related CASRAI guides
For the individual-level counterpart to this method, see interviews as a research method and designing semi-structured interviews. For what happens to a focus group transcript afterward, see coding qualitative interview data and thematic analysis. For how a focus group study is written up, see writing the methodology section of a qualitative research paper. For participant recruitment and sampling logic, see the dictionary entry on sampling methods and the comparison of purposive vs. convenience sampling. For consent and confidentiality specific to group settings, see research ethics in qualitative research. For the broader qualitative-vs-quantitative framing, see qualitative research, qualitative data collection techniques, and the comparison of qualitative research vs. quantitative research.
