Is a focus group a small survey?

No. Interaction and qualitative interpretation are central to the method.

Can researchers report counts?

They can use carefully defined descriptive counts, but these do not turn a purposive discussion into a population estimate.

Does agreement establish validity?

Not by itself. Researchers should examine context, dissent, moderation and alternative explanations.

AI Glossary / Statistics and probability / Foundation

Focus groups

Definition

Focus groups are moderated discussions in which a small group of participants explores a defined topic through questions, responses and interaction with one another. They are a qualitative research method for understanding experiences, meanings, language and social expectations. The discussion itself is part of the evidence: agreement, disagreement and the ways people explain or revise views can matter. A focus group is not a representative opinion poll, and counting favorable comments does not establish the proportion of a wider population that shares a view.

What it means

A structured questionnaire asks comparable questions of separate respondents, while a focus group lets participants react to each other's experiences and explanations. This can reveal assumptions that would remain hidden in a simple rating scale. Participants are often selected for relevant experience rather than through probability sampling. The resulting insight can help explain a problem, improve a research instrument or identify questions for further investigation. It should be interpreted in relation to recruitment, group composition, moderation and context, rather than treated as an automatic measure of market demand.

How it works

Begin with a focused research question and recruit people whose experience is relevant. Prepare an open-ended discussion guide, explain participation and recording arrangements, and use a moderator who can invite contributions without steering toward a preferred answer. Record the discussion appropriately and retain context when transcribing or summarizing it. Analyze the material systematically, documenting codes, emerging themes, disagreements and examples that challenge an initial interpretation. Compare across groups where the design calls for it. Group size and number depend on purpose, topic and practical conditions; no single count guarantees adequate insight. Protect identities in reporting and recognize that researchers cannot fully control what other participants later disclose.

Three practical examples

Product discovery: explore how intended users describe difficulties with an existing workflow before designing a survey or prototype.
Communication research: investigate how people interpret a proposed explanation, including confusing terms and conflicting expectations.
Employee research: examine shared experiences of a process, using appropriate recruitment and group composition so status differences do not unnecessarily suppress discussion.

Think of it this way

Think of listening to a carefully guided conversation around a table. One person's story prompts another to explain a different experience. The value lies in understanding those connections and differences, rather than turning the loudest voice into a vote for everyone outside the room.

Worked example

Consider an invented six-person discussion about an AI assistant. A coding exercise records eight relevant comments: participant A makes two comments about speed and one about control; B makes one about control; C makes one about clarity; D makes one about speed; E makes none on these topics; and F makes one each about control and clarity. There are three speed-coded comments, but only two participants made them. The difference illustrates why counting comments and counting speakers answer different descriptive questions. Neither count estimates how common the concern is in the customer population. Silence by E also cannot be coded automatically as agreement or disagreement. Analysts return to the transcript to examine what speed meant, whether participants were responding to a prompt and how their accounts differed. The arithmetic organizes the material; interpretation requires the surrounding discussion and the research design.

Illustrative case study

Fictional case: Juniper Experience Studio holds separate discussions with new and experienced users of a proposed AI feature. Participants interpret an automation label in different ways, with some expecting a draft and others expecting an immediate action. Researchers document the contrasting interpretations and revise wording for subsequent usability testing. They do not report that a percentage of all customers holds either expectation. This invented case demonstrates how focus groups can identify language and hypotheses worth testing without replacing direct observation of product use or a representative survey.

Common misconceptions and mistakes

Do not treat apparent consensus as proof that every participant privately agrees. Dominant speakers, social desirability and moderator wording can shape discussion. Do not count repeated remarks from one person as independent observations. A striking quotation is not necessarily a common view. Recruitment through convenient contacts can exclude important perspectives. Sensitive personal topics may be better explored through individual interviews. Automated transcription and AI-assisted summaries require checking against recordings and context; fluent summaries can omit dissent or invent coherence that the conversation did not contain.

Business use

Businesses can use focus groups to improve their understanding before committing to detailed measurement or design. State what the sessions were intended to explore and how participants were selected. Preserve a clear connection between themes and supporting material, while removing identifying details from shared reports. Distinguish observed accounts from researchers' interpretations and proposed actions. Combine the findings with other evidence when deciding prevalence, demand or causal effects. Budget time for analysis as well as moderation, because the usefulness of a discussion depends on how carefully its meaning is examined.

Check your understanding

One participant repeats a concern five times and another mentions it once. Does that establish six independent supporters?

Answer: No. There are two speakers and six mentions, with dependence and context. Neither number establishes the concern's prevalence outside the group.

Related terms

Usability testing
Mixed methods
Qualitative research
semi-structured interview
thematic analysis
purposive sampling
moderator
social desirability bias
research saturation

Learn next

Learn how to write open-ended discussion questions and maintain an analysis record. Then compare focus groups, individual interviews, observation and probability surveys to match each method to the kind of conclusion the project needs.

Sources

CDC, Collecting and Analyzing Qualitative Data: https://www.cdc.gov/field-epi-manual/php/chapters/qualitative-data.html CDC, Gather Credible Evidence: https://www.cdc.gov/evaluation/php/evaluation-framework-action-guide/step-4-gather-credible-evidence.html CDC, Practical Strategies for Qualitative Analytic Rigor: https://www.cdc.gov/pcd/issues/2022/22_0134.htm CDC, Using Qualitative Data, limitations of numerical generalization: https://stacks.cdc.gov/view/cdc/136927/cdc_136927_DS1.pdf

Previous: Focal stacking Next: Focus of expansion