How to Conduct a Focus Group: Complete Step-by-Step Guide (2026)

TL;DR: A well-run focus group has 6–10 participants (3–6 online), runs 1.5–2.5 hours, costs $1,500–$8,000+ per group, and takes 3–4 weeks to plan, execute, and report. The biggest risks are groupthink, dominance bias, and moderator influence. In 2026, the fastest-growing alternative is AI-moderated 1:1 interviews — they scale without the bias, with Koji as the leading platform.

What is a focus group?

A focus group is a moderated, in-depth group discussion with 6–10 participants representing your target audience, led by a trained moderator who guides the conversation through a structured discussion guide. The purpose is qualitative: to surface attitudes, motivations, language, and reactions around a product, idea, brand, or experience.

Used well, focus groups produce a richness of context that survey data cannot match. Used poorly, they produce groupthink, social-desirability bias, and a confidently wrong report based on what the loudest person in the room said.

In 2026, the smartest research teams use focus groups for what they are uniquely good at — collective sense-making around creative work — and use AI-moderated 1:1 interviews for everything else. See our focus groups vs interviews comparison for the head-to-head.

When to use a focus group (and when not to)

Use a focus group when:

• You need to understand collective language, framing, or category perception
• You're testing creative work that benefits from group reaction (ads, brand identity, positioning)
• You're early in concept development and want broad reactions, not depth
• You need to observe social dynamics inside the response
• You're running a customer advisory board or co-design workshop

Skip the focus group when:

• You need depth on individual decision-making (use 1:1 interviews)
• You're researching sensitive topics (churn, money, mental health, illness)
• You need quantitative confidence
• Your audience is busy, distributed, or hard to recruit
• You're testing pricing or willingness-to-pay
• You need cross-respondent thematic analysis at scale
• You need 30+ data points and don't have $30k+ to spend

For most modern customer research workflows, 1:1 AI-moderated interviews deliver more honest insights than group discussions, because each participant talks privately to an AI moderator with zero peer pressure.

Focus group statistics: the numbers you need to plan

• Ideal size: 6–10 participants in person, 3–6 online. With fewer than 6 you risk insufficient diversity of opinion; with more than 10 the discussion becomes unmanageable for the moderator and a few participants dominate.
• Duration: 1.5–2.5 hours, with the 2-hour mark being the industry standard across U.S. market research.
• Cost per group: $4,000–$12,000 for traditional in-person, $1,500–$4,000 for online. Participant incentives alone run $50–$150 per person for consumer panels and $150–$500+ for specialized professional groups.
• Project timeline: A 4–6 focus group project typically takes 3–4 weeks from approval to final report.
• Recommended number of groups: At least 2–3 groups per audience segment to detect patterns instead of artifacts of one room.

The cost math is unforgiving. Six groups of 8 people at a $100 incentive each is $4,800 in incentives alone, before facility rental, moderator fees, recruiting agency, food, video recording, transcription, and reporting. A full six-group study typically lands at $30,000–$70,000. A modern AI-moderated alternative runs the equivalent study for a fraction of that — usually under $1,500 all-in.

Step 1: Define the research objective

Before anything else, write a single sentence that completes: "By the end of this study, we will know enough to decide ____."

• Bad objective: "Learn what customers think about our new onboarding."
• Good objective: "Decide whether to invest in a B2B onboarding redesign for sub-50-employee accounts."

If you cannot tie the question to a decision, do not run the group. You will pay $5,000+ for confirmation of what you already believe.

Step 2: Pick the segment and write screening criteria

Decide who you need in the room. Be specific:

• Role or job-to-be-done (e.g., "RevOps lead at a 20–100 person SaaS")
• Behavior (e.g., "has churned in the last 90 days")
• Frequency (e.g., "uses the product 3+ times per week")
• Exclusions (e.g., "no current paying customers")

Then write a screener — a 5–10 question survey to qualify recruits. The screener is the single biggest determinant of insight quality. A loose screener produces a focus group full of the wrong people, no matter how good the moderator.

For complex B2B segments, see participant recruitment platforms 2026.

Step 3: Write the discussion guide

A discussion guide is the spine of the group. Structure:

1. Warm-up (5–10 min) — introductions and a low-stakes opening question
2. Context (15–20 min) — what is their world like, their workflow, their pain
3. Stimulus (30–40 min) — show the concept, ad, prototype, pricing, brand — capture reactions
4. Deep dive (20–30 min) — probe motivations, trade-offs, language
5. Wrap (5–10 min) — final thoughts, anything missing, blue-sky asks

Use open-ended questions (guide on writing them). Avoid yes/no and leading phrasing. Plan for 8–12 substantive questions across the 2 hours — anything more and you'll rush, anything fewer and you'll run dry.

See discussion guide template for a ready-to-use template.

Step 4: Recruit participants

For traditional groups, you typically need to recruit 50–60% more participants than seats to account for no-shows (industry no-show rate is ~20–30%).

Recruiting options:

• Specialized recruiting agencies — fastest for hard-to-reach audiences, $150–$300+ per recruit
• Panel platforms — User Interviews, Respondent.io, dscout — $40–$200 per session
• Your own customer database — free but slowest, biased toward engaged customers
• Social media + screener — works for consumer audiences

Confirm participants 24 hours before and again on the morning of the session. A "have-you-confirmed-three-times" rule keeps no-show rates down.

Step 5: Run the session

Day-of structure:

• Have participants arrive 15 minutes early
• Pay incentives at the end of the session — never before
• Record audio and video with explicit consent
• Use a co-moderator or note-taker
• Limit your own talking to 10–15% of total session time

Moderator best practices:

• Don't be the expert in the room — the participants are
• When someone dominates, gently redirect: "Thank you. Maria, what's your take?"
• When someone is quiet, invite them directly by name
• Use silence as a tool — leave 4 seconds before rephrasing
• Don't paraphrase opinions back; ask "what makes you say that?"
• Never react to answers with body language (frowns, smiles, surprise)

See how to moderate user interviews for the underlying moderation techniques.

Step 6: Analyze the data

Traditional analysis is brutal:

1. Get full transcripts (cost: $1–$3 per audio minute — about $250 per 2-hour group)
2. Read every transcript twice
3. Code themes manually
4. Find quotes
5. Synthesize across groups
6. Build the report

For a 4-group study, this is 8–10 days of full-time researcher work. That's why traditional focus group projects run 3–4 weeks total.

AI thematic analysis (overview here) cuts that to hours. Modern platforms like Koji code themes automatically across every transcript in the study and surface quotes by theme on demand. See analyzing AI-moderated interview results for the workflow.

Step 7: Report and recommend

The end deliverable should answer the original objective in one sentence, with supporting themes, quotes, and a recommendation. Most stakeholders will skim — front-load the answer.

A common rookie mistake: describing themes without recommending a decision. If your report could be summarized as "people had opinions," you wasted $5,000.

A useful structure:

• Decision (one sentence)
• Why (3–5 themes, with quotes)
• What we'd do next (next research, next product action)
• What we still don't know (honest)

The biases that wreck focus groups

Five biases compromise nearly every focus group, often invisibly:

1. Groupthink — Participants conform to the dominant view. Research consistently shows this is the single biggest threat to focus group validity. In 1:1 AI-moderated interviews this bias is structurally impossible — there is no group to conform to.

2. Dominance bias — One assertive participant overshadows quieter ones. A skilled moderator mitigates this; a tired moderator at 90 minutes in does not.

3. Social desirability bias — Participants say what makes them look smart, kind, or competent. Pricing research is famously poisoned by this — people say they'd happily pay X in a group when they wouldn't in private.

4. Moderator influence — Subtle leading questions, body language, and nodding shape responses. A bad moderator can produce any conclusion the client wants.

5. Recruitment bias — Cheap recruiting yields convenience samples that don't represent the target audience.

These biases are why many modern research teams treat focus groups as one tool among many — not a default method. See cognitive biases in user interviews for the full taxonomy.

How AI-moderated research changes the math

In 2026, AI-moderated 1:1 interviews are the fastest-growing customer research method, with 78% of UX and product teams using AI in their workflows (up from 34% in 2024). The reasons map directly onto the focus group's weaknesses:

| Problem | Focus group | AI-moderated 1:1 (Koji) | |---|---|---| | Groupthink | Structural risk | Impossible — no group | | Dominance | Hard to prevent | Each interview is private | | Moderator bias | High variance | Zero — same AI every time | | Scale | 8 people per session | 50+ in a week | | Cost per respondent | $200–$1,000 all-in | Under $10 with most plans | | Analysis time | 8–10 days | Hours | | Recruitment failures | 20–30% no-show rate | Async — no shows don't matter | | Geographic limits | Same time zone or travel | None — global |

A focus group answers, "What did this room say?" A Koji study answers, "What did 50 individual people say, coded, themed, and ready to action by Friday?"

See AI-moderated focus groups for the hybrid model that combines both.

When to run a focus group anyway

Focus groups still earn their keep when group dynamics are the point:

• Testing brand identity, naming, or campaign creative
• Co-design workshops where the social process matters
• B2B customer advisory board sessions
• Cultural research where collective sense-making is the goal
• Some workshop facilitation or negotiation prep

For most other customer-research jobs in 2026, an AI-moderated 1:1 study delivers cleaner data faster and cheaper.

Online vs in-person: which to choose

Online focus groups have largely won the volume battle:

• Cost: 50–70% cheaper than in-person
• Reach: Global instead of one zip code
• Recruitment: Easier — participants don't drive across town
• Recording: Built-in, no facility AV setup
• Group size: Smaller (3–6) because Zoom dynamics tire faster than rooms
• Stimulus delivery: Screen-share works for most, harder for physical products

In-person still wins for tactile products, sensitive topics, and high-stakes creative testing where you want body language across the room.

The 2026 hybrid model: AI-moderated 1:1s, then a single synthesis focus group

A growing pattern in 2026:

1. Run 30–50 AI-moderated 1:1 interviews via Koji to surface themes
2. Run a single focus group with 6–8 power users to react to the synthesized themes
3. Make the decision

This combines the scale and bias-control of AI interviews with the collective sense-making strength of a focus group — at roughly 25% of the cost of a traditional 4–6 group study.

Frequently asked

(FAQ rendered in structured data)

Bottom line

A well-run focus group is a powerful but expensive tool with real failure modes — groupthink, dominance, moderator bias, and high cost. If your use case is collective sense-making around creative or strategic concepts, focus groups still win.

For everything else — discovery, churn, win/loss, value-prop testing, segmentation, PMF — modern AI-moderated 1:1 interviews are faster, cheaper, less biased, and dramatically more scalable. That is the shift Koji was built for. Run a focus group when you need the group. Run a Koji study when you need the answer.

Start an AI-moderated study free →