Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Marketing Research › Customer Research

Running a Focus Group That Produces Evidence

How to run a focus group: when a group beats interviews, setting the mix, a guide built for a group, moderating dominant voices, consent and honest analysis.
Tutorialintermediate10 min readUpdated

What a group shows that an interview does not

The qualitative versus quantitative comparison lists group discussions among the qualitative methods and warns against writing "we need focus groups" as the first sentence of a brief. The method follows the question, and a group suits a narrow set of questions well.

A group is good at reactions to stimulus, such as a name, a label, an advert or a pack, where hearing one person's reaction prompts another to say what they noticed. It is good at how people talk about a category among peers: the shorthand they share, the comparisons they make and the points they argue over. And it shows where the language of different kinds of customer diverges, if each group is made up of one kind.

A group is poor at private behaviour, at individual decision journeys and at anything sensitive. Nobody reconstructs the evening they nearly cancelled in front of strangers the way they would in a one-to-one conversation, and the customer interviews tutorial owns that kind of reconstruction. A group is also not a standing panel of advisers: the customer advisory board guide is explicit that a board is not a focus group, because its value is continuity with the same informed customers. Nor can a group be simulated: the guide to synthetic personas and simulated respondents describes a simulated panel as the prompt's framing voiced several ways, not focus group insight.

Before you start

Have a written brief, built as the research design guide describes, naming the decision, the questions and the kinds of people whose reactions matter. The research brief template has the fields. If the brief cannot say what the group must show, a group will produce a lively discussion and no finding.

Read the tutorial on recruiting research participants first. A group depends on recruitment more than an interview does, because one wrong participant shapes what everyone else says.

Know the professional principles the site's research guides set out from the MRS and ESOMAR codes: participation is voluntary and informed, people are told who is asking and why, answers are confidential unless people agree otherwise, research is never a disguise for selling, and vulnerable people get extra care. Decide with whoever owns data protection how recordings, notes and the participant list are handled, and when they are deleted.

Decide who moderates. As general practice, the moderator should be neutral about the answer and practised at managing a room; someone with a stake in the concept tends to defend it. Decide who observes, and how they will watch without being seen or heard by participants.

The steps

  1. Decide what the group must show and whether a group can show it

    Write the one or two things the decision needs from the group, such as "how home painters read the colour names and which ones they would not say aloud in a shop". Then test each against the question in the previous section: is this about shared language or reaction to stimulus, which a group shows, or about an individual's private decision, which it hides? Move anything in the second category to interviews. A group answering a question it cannot answer still produces quotes, and the quotes mislead.

  2. Set the composition of each group

    Put people who share the experience that matters in the same group, and keep groups apart where their experience or standing differs. A professional and an amateur in one room, a long-standing customer and a first-time buyer, or a manager and the person who reports to them, will usually defer to whoever seems to know more. The business's Segmentation and its personas are useful places to find the dividing lines, provided the line changes the answer to the question. Run separate groups for each kind of person the decision compares, so differences show between groups rather than being smoothed over inside one.

  3. Recruit through a screener and invite reserves

    Describe the participants by what they did and when, choose a frame and write a neutral screener, as the recruitment tutorial sets out. Add the usual exclusions: people who work for the business or a competitor, people who work in research or marketing in the category, and people who have taken part in research recently. Screen out people who know each other, because friends talk to each other rather than to the group. Invite reserves from the same profile, because a group in which people fail to appear changes character, and fill a gap from the same group's reserves, never from whoever is available.

  4. Write a guide built for a group

    A group guide differs from an interview guide in a few ways, as general practice. It moves from the general to the specific, so people talk about the category in their own words before they see any of the business's material. It builds in moments where everyone writes or marks something on their own before the discussion starts. And it plans for who speaks: questions that go round the group, questions put to someone who has not spoken, and prompts for disagreement such as "who saw it differently?". The jobs-to-be-done framework helps with the opening section: ask what people were trying to get done the last time they bought in the category, so later reactions can be read against it.

  5. Prepare the stimulus and the room or the call

    Stimulus is whatever people react to: names, labels, adverts, packs or messages. Show each in the same form to every group, and rotate the order between groups so the first item does not always set the standard. When the stimulus is a message, ask the questions the messaging hierarchy guide puts first in its qualitative tests: what did you understand this to mean, what did you doubt, and what would you want to know next. Comprehension before preference. For a room, arrange seating so people face each other and the moderator can see everyone. For an online group, check that everyone can join, see the stimulus and speak, and agree how people signal that they want to come in.

  6. Open with consent and ground rules

    At the start, say who is running the session and for whom, what it is for, that it is recorded if it is, who will see the recording, that there are no right answers and that anyone may stop at any time. Confirm that every participant agrees to the recording before it starts. A recording of a group captures several people's voices and often their faces, and as the personal data glossary entry explains, recordings that identify people are Personal Data under the GDPR. Then set the ground rules: one person at a time, disagreement is welcome, what others say stays in the room, and nobody is selling anything today.

  7. Have people write before they discuss

    Before each important discussion, ask everyone to write their own first reaction: a word, a sentence, a mark on a scale or a choice between items. Collect or photograph the sheets, or have people type into a private field in an online session. As general practice, this is a strong defence against group effects. A first reaction written in private cannot be swayed by the first person to speak, and it gives the analysis a record of what each person thought before the room formed a view. Then ask people to read out or describe what they wrote, starting with someone other than the most confident person in the room.

  8. Moderate: draw out the quiet and contain the dominant

    As general practice, the moderator's job is to hear from everyone, not to reach agreement. Bring in people who have not spoken by name and with an easy question. When one person dominates, thank them, then turn to the group: "that's one view; who sees it differently?". Treat quick agreement with suspicion and probe it: ask what someone would change, or what a friend who disliked it would say. Do not explain or defend the material, because once the moderator argues for a concept, the group stops criticising it. Keep a note of moments where the discussion turned on one person's comment, because those are the moments the analysis must treat with care.

  9. Debrief observers straight after

    Observers form strong impressions, usually of the most quotable participant, and those impressions travel faster than the analysis. Gather them straight after each group and ask three questions: what surprised you, what did people say in their own words, and where did the group disagree? Write the answers down with the group's code and the date. The debrief captures observations while they are fresh and makes each observer's reading visible, so it can be checked against the record rather than repeated in meetings as the finding. Booking the debrief before the first group is one line of the research session readiness checklist, the gate ticked once recruits are confirmed.

  10. Analyse by group and by person, never by show of hands

    Start from the written first reactions, then the transcript. For each theme, note which groups it came from and which individuals, and whether it appeared before or after someone influential spoke. A theme that appears in the private sheets of people in different groups is stronger than one that appeared once and was agreed with. Never report a count of hands or a share of the room as a measure; prevalence needs a survey, as the qualitative versus quantitative comparison states. The AI for research synthesis guide sets out an evidence table, with the participants each theme comes from and the excerpts for and against it, which suits group material well. Attribute quotes by group, not by name.

Groups or interviews: choosing

Most questions point clearly to one method once they are written down. The table sets out where each wins.

The question is aboutA groupInterviews
How people talk about a category among peersStrong: shared language and disagreement surface in the discussionWeaker: each person's words, without the exchange
Reactions to names, labels, adverts or messagesStrong, if first reactions are written before discussionStrong, and free of group influence, but slower to cover many items
An individual's decision journeyWeak: people summarise rather than reconstructStrong: the core of the interviews tutorial
Private behaviour or sensitive topicsUnsuitable: people hold back in front of strangersSuitable, with extra care
Differences between kinds of customerStrong, with separate groups for each kindStrong, with interviews planned per group
How many people share a viewUnsuitable: a group is not a sampleUnsuitable: needs a survey
Senior or hard-to-gather participantsOften impractical to assemble at one timeUsually easier to schedule one by one

Where the table points both ways, interviews are the safer default, because they avoid group effects and can be scheduled around each participant. A group earns its place when the exchange between people is itself the evidence. The customer research hub sets both beside surveys and observation, with the recruiting and synthesis work they share.

A paint maker runs groups with decorators and home painters (illustrative)

Illustrative example — not a real result

Nymbrecht Paints tests colour names and tin labels in separate groups

Nymbrecht Paints, an invented paint maker, is renaming 24 colours and redesigning its tin labels, and wants to hear how the people who buy paint read the names and labels. The brief compares professional decorators and home painters, so recruitment fills 2 groups of each, 6 people per group, through a screener that asks which of 10 home improvement jobs people did in the past year and screens out anyone working in paint retail.

The first decorators' group goes wrong early. One decorator with long experience gives a firm verdict on the first label, and the next 4 people agree with him almost word for word. The moderator changes the guide for the remaining groups: before each label is discussed, every participant writes a first reaction on a card and picks the 3 colour names they would be comfortable saying aloud at a trade counter.

The written cards change the picture. Decorators read the labels for coverage and drying information and dismiss names that sound playful, while home painters read the names first and find the technical panel hard to follow. Several names that the first group appeared to reject were in fact picked on the cards by 3 of the 6 decorators before the discussion began. The report sets out where the 2 kinds of buyer read the labels differently, attributes each quote to a group code, gives no counts of hands, and recommends a short survey of home painters before the final names are chosen. All recordings are deleted on the date agreed with the company's data protection owner.

Common mistakes

  • Choosing a group before writing the question. A group answers questions about shared language and reactions; it hides individual decisions and private behaviour.
  • Mixing people who defer to each other. Experts and novices, or managers and their staff, in one room produce the expert's view repeated.
  • Discussion before private reactions. Without a written first reaction, the analysis cannot tell what people thought from what the room decided.
  • A moderator who defends the material. Once the concept is being argued for, the group stops criticising it.
  • Counting hands. A share of a room is not a share of the market; prevalence needs a measured sample.
  • Observers' impressions as findings. The most quotable participant becomes the story unless the debrief and the evidence table check it.
  • Recording without everyone's agreement. Every voice on a group recording is personal data, so each participant agrees before recording starts.

Frequently asked questions

Is a focus group cheaper than running interviews?

It can look that way, because several people are heard in one session, but the saving is smaller than it seems. Each person speaks for a fraction of the time, the group shapes what each says, and recruitment for a group is harder because everyone must fit the profile and turn up at once. Choose a group when the question is about shared language or reactions among peers, and interviews when it is about individual decisions, private behaviour or anything sensitive.

Can we use a show of hands to see how many people liked a concept?

No. A group is a small, purposive set of people who influence each other, so a count of hands describes that room on that day, not the market. The qualitative versus quantitative comparison makes the same point about interviews: prevalence needs a measured sample. Report what people said and why, attributed by group, and if the decision depends on how many people share a reaction, follow up with a survey.

Do we need consent from everyone before recording a group?

Yes. The customer interviews tutorial asks for agreement before recording starts, and in a group that means every participant, because a recording captures everyone's voice and often their face, and recordings are personal data. Tell people in the invitation and again at the start what is recorded, who will see it and how to withdraw, and agree the wording, the lawful basis and how long the recording is kept with whoever owns data protection; this page is educational, not legal advice.

Terms used on this page

Frameworks

Where this page is used

Also related

Sources

  1. Market Research Society — MRS
  2. ESOMAR — ESOMAR
  3. Regulation (EU) 2016/679 (GDPR) — EUR-Lex

Related