Back to blog
8 min readby Mathias

Analysing Focus Groups: the Interaction Is the Data

Why focus groups are not group interviews, choosing between turn, exchange and group as your unit of analysis, handling the dominant speaker, and coding agreement and disagreement explicitly.

Focus GroupsGroup DiscussionMethodologyCoding

Focus groups are usually analysed as if they were several interviews recorded at once. They are not, and treating them that way discards the one thing that makes them worth running: the interaction.

What a focus group actually gives you

In an interview, a participant tells you their view. In a focus group, you watch a view get formed, contested and revised in front of you. Someone says something, someone else disagrees, a third person finds a middle position, and the first speaker adjusts. None of that exists in a one-to-one interview.

So the analytically interesting material is often not what anyone said, but where the group converged, where it split, and what nobody was willing to say until someone else said it first.

Three units of analysis, and you must choose

This is the decision that determines everything downstream, and it is usually made by accident:

Most software defaults to speaker turns because it is the only one that can be derived automatically. That is a reasonable default and a poor excuse — if the exchange is your unit, you need to mark boundaries yourself, and you should say so in the methods section.

The dominant-speaker problem

Every focus group has someone who talks twice as much as anyone else. Code by frequency and their view becomes "the group's view", which is a straightforward measurement error.

Two defences. Report contribution distribution — how many turns and how many words per participant. It takes a minute and immediately shows whether you had a discussion or a monologue with an audience. And count participants, not mentions: a theme raised eleven times by one person is not more prevalent than one raised once each by four people. It may be less.

Coding agreement and disagreement explicitly

The most useful thing you can add to a focus-group codebook is a small second dimension applied alongside your content codes:

That last one is the interesting category. A statement that draws no response at all may mean quiet agreement, or it may mean nobody wanted to touch it. Either reading matters, and neither is visible if you only code content.

Practicalities of the transcript

Focus group transcripts are harder than interview transcripts, and the difficulty is mostly clerical:

That last point is genuinely underrated. Our pseudonymisation tool is free on every plan and handles German inflection, which manual find-and-replace almost always misses.

Reporting

Say how many groups, how many participants each, and how they were composed. Say which unit of analysis you used and why. Report contribution distribution. And when you quote, give the group and speaker (FG2, TN4) so a reader can see whether your three strongest quotes all came from the same enthusiastic person.

On building a codebook → · See a worked example analysis

Try Themera on your own data

3 analyses free. No credit card. EU-hosted and GDPR-native.

Start free