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:
- The speaker turn. Each contribution is one unit. Best for content-analytic work where you want to know what was said about what. Loses the interaction.
- The exchange. A sequence of turns on one topic is one unit. Keeps disagreement visible. Harder to code consistently, because the boundaries are a judgement call.
- The group. Each group is one case, compared with other groups. Right when you ran six groups and the question is how they differ.
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:
- Konsens — others explicitly agree
- Widerspruch — someone explicitly disagrees
- Erweiterung — someone builds on it
- Unwidersprochen — said, and nobody responds
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:
- Label speakers consistently —
TN1:,TN2:,M:for the moderator. Automatic diarisation gets confused when people talk over each other, so plan to fix it by hand. - Mark overlaps. They are usually where the disagreement is.
- Pseudonymise carefully. Group members know each other. A detail that would not identify someone to an outsider — the ward, the shift, an incident — identifies them instantly to the four other people who were in the room, and they may read your thesis.
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.