Focus Groups

In Brief

A focus group is a facilitated discussion with 6–10 participants designed to generate qualitative feedback on a specific topic. A moderator guides the conversation while participants react to each other’s ideas, revealing group buying dynamics, shared language, and points of agreement or disagreement. The output is a set of themes, customer language, and insights into how people influence each other’s opinions — though results must be interpreted carefully, because outgoing participants tend to dominate and groupthink can distort the outcome.

Description

Focus groups are a facilitated small-group discussion — typically 6–10 participants over 60–90 minutes — that larger companies have long used to hear the “voice of the customer.” They are a divergent qualitative tool: group conversations chain associations among participants, surfacing shared language and unarticulated needs in a way one-on-one interviews often do not (Krueger & Casey, Focus Groups: A Practical Guide for Applied Research, 5th ed., 2015).

The method has its origins in WWII-era research at Columbia University’s Bureau of Applied Social Research, where Robert K. Merton, Marjorie Fiske, and Patricia L. Kendall worked with Paul Lazarsfeld on the effectiveness of military propaganda films. They coined the term focused interview — the procedure was later adapted and renamed focus group as it spread into commercial market research. The method was codified in their 1956 manual The Focused Interview (Merton, Fiske & Kendall, Free Press, 1956).

Focus groups are most useful for idea generation rather than verification. According to Carol-Ann Morgan of B2B International, typical situations where focus groups fit include:

• Unraveling a complex process from the basics, such as a complicated buying journey
• Identifying customer needs where multiple factors interact to drive motivation
• Understanding working practices, such as how a particular product is used in context
• Testing new products when you need to show participants something tangible
• Exploring a concept with stimulus aids
• Exploring drivers of satisfaction for customers, staff, or suppliers
• Examining perceptions of a brand and the service elements associated with it

There is a long-running skeptical view of the method, summarized in industry guidance from Nielsen Norman Group on the use and misuse of focus groups: groups are widely misused for evaluative work — confirming a design or measuring willingness to pay — where one-on-one interviews and behavioral methods produce more reliable signal. Treat focus groups as a tool for hearing how people talk about a problem, not for proving they will buy a solution. More broadly, Harvard Business Professor Gerald Zaltman has argued that the bulk of cognition is unconscious, which limits what any verbal method — focus groups included — can surface (Zaltman, How Customers Think, Harvard Business School Press, 2003).

Lean version: If you don’t have a research budget, you can run a scrappy focus group by inviting 4–6 existing users or prospects to a 45-minute Zoom call. No venue, no recruitment firm, no incentives needed beyond genuine interest in the topic. You moderate, an AI transcription tool takes notes, and you get 80% of the insight at 5% of the cost. The key difference from individual interviews is the group dynamic — participants build on each other’s ideas and challenge assumptions in ways they would not one-on-one.

AI has introduced new variations on the format. Platforms now offer AI-moderated discussion groups where a chatbot guides participants through structured prompts, collects responses, and follows up on interesting threads. The most effective approach today is a hybrid: a skilled human facilitator runs the live session while AI handles transcription, theme extraction, and cross-session comparison. AI also speeds up individual interview analysis, which has eroded one of the historical time-saving advantages of focus groups over one-on-one methods.

How to

Prep AI Prompt

1. Pick a single, clear purpose. One product or one issue per session. If you cannot state the learning goal in a sentence, you are not ready to run the group.
2. Narrow your target audience. Talk to one segment at a time. Mixing distinct customer types in one room produces conversation that averages over real differences.
3. Decide whether you need a control group. Running a parallel group of non-customers or a different segment lets you contrast opinions against a wider context.
4. Refrain from ulterior motives. Do not run a focus group to validate a decision you have already made or to soft-launch a sales pitch. The signal will be skewed before the first question lands.
5. Find a co-facilitator. A second person handles note-taking, equipment, and side conversations so the moderator can stay present in the discussion.
6. Choose a comfortable venue and recording method. Comfort and psychological safety are critical. For virtual sessions, default to a meeting tool participants already know.
7. Prepare up to ten open-ended questions. Establish rapport early, avoid jargon, and save anything embarrassing or threatening for late in the session — once participants feel comfortable.

Analysis AI Prompt

18. Transcribe the session. AI transcription tools (Otter.ai, Looppanel, Recall.ai, Dovetail) handle this in minutes; clean obvious errors before coding.
19. Code and cluster the responses. Use affinity mapping — sticky notes or a digital equivalent — to group related quotes into themes. Capture the customer’s actual language, not your paraphrase.
20. Look for divergent opinions and outliers. A focus group’s strength is showing how people influence each other; pay particular attention to moments where someone changes their mind, pushes back, or goes quiet.
21. Separate group-influenced views from individual beliefs. When a theme looks important, plan a follow-up one-on-one interview to check whether participants hold the view independently or only adopted it under social pressure.
22. Triangulate with other data. Compare themes against your existing customer interviews, support tickets, and behavioral data before treating any single theme as a finding.