How to Run a Focus Group, Step by Step

When a focus group is the right method

Focus groups shine when you need range and reaction: how different people in a market respond to a concept, a message, or a category, and how those responses shift when they hear each other. They're the wrong tool for measuring prevalence (use a survey), for usability (watch individuals), or for deeply personal topics (use 1:1 interviews, where there's no audience to perform for).

The eight steps

1. Define one decision the group should inform

Not "learn about our customers" but "decide which of two positioning directions to develop." Every later choice (who to recruit, what to ask) falls out of this sentence. If you can't write it, you're not ready to spend the money.

2. Decide who must be in the room

Write a screener: the 4-6 criteria a participant must meet (role, behavior, category usage), plus quotas for the mix you need. Homogeneous-enough groups talk freely; too-mixed groups perform for each other. If two segments matter, run two groups rather than blending them.

3. Write the discussion guide

Twelve to sixteen main questions for an hour: openers, current behavior, the stimulus, the hard trade-offs, the close. (We keep a full library of 80 proven questions in the companion guide below.) Pilot the guide before the real session; the first draft always has a leading question or a dead end.

4. Recruit (the expensive part)

Plan 2-3 weeks. Sources: a recruiting firm (reliable, $100-200 per recruited participant), your own customer list (cheap, biased toward fans), or intercept/community recruiting (slow). Over-recruit by 20%: someone always no-shows. Incentives typically run $75-200 per consumer participant and $200-400 for hard-to-reach professionals.

5. Set the room

In person: a quiet room, a round table, name tents, recording (with consent), and snacks that aren't loud. Remote: a video call with cameras on, a co-moderator running tech, and stimuli pre-loaded as screen shares. Sixty to ninety minutes; after ninety, data quality falls off a cliff.

6. Moderate for disagreement

• Open with ground rules: no wrong answers, disagreement is useful, one voice at a time.
• Ask, then be quiet. The silence after a question is where honest answers live.
• Manage the dominator ("let's hear from someone who hasn't weighed in") and invite the quiet ("Maria, you were nodding...").
• Probe relentlessly: 'tell me more,' 'what makes you say that,' 'who sees it differently?'
• Never correct, defend, or sell. The moment participants sense a right answer, they'll give it to you.

7. Debrief within 24 hours

Top-of-mind impressions decay fast. Same day, write: the three strongest themes, the surprises, the verbatims worth quoting, and what the group disagreed about. Then do the slow pass on the transcript later.

8. Report decisions, not minutes

The deliverable is themes, supported by verbatims, ending in recommendations against the decision from step 1. Nobody needs a chronological summary of who said what.

What it really costs

The AI alternative (and when to use which)

An AI focus group runs the same structure (a recruited-to-spec panel, multi-round moderated discussion, themed report) with synthetic personas instead of recruited humans, in minutes instead of weeks. It will not give you real human evidence; it will give you a directional read, the objections, and a debugged guide. The decision rule most teams land on: AI panels for iteration, shortlisting, and decisions that would otherwise get no research; human groups for the final validation of high-stakes calls.

• Use AI first when: you're choosing between options, the budget is under $4,000, the timeline is days, or the topic is competitively sensitive.
• Go straight to humans when: the decision is a one-shot bet, regulators or boards need real evidence, or lived experience is the subject itself.
• Use both when stakes are high: AI rounds to converge on the strongest version, one human round to validate it.