How to transcribe a focus group: labels, crosstalk and checking who said what

A one-to-one interview has two voices. A focus group has a moderator and six to eight participants, some of whom talk at the same time, laugh together or agree with a single "mm-hm". The transcript has to show who said what, or the analysis falls apart. This guide covers the whole path, from the seating plan to a transcript you can code.

Before the session: make the voices easy to tell apart

Most focus group transcripts go wrong in the room, not in the software.

• Get consent from everyone, on paper and on the recording. Ask each participant to agree before you start, and repeat it at the start of the recording. The rules differ between countries and institutions, so check what your ethics board or client requires.
• Draw a seating plan. Write down where each person sits and give them a label: M for the moderator, P1, P2 and so on for participants. When two voices sound alike, the seating plan is what tells them apart.
• One recorder in the middle, not at the moderator's end. A phone or recorder at the centre of the table hears everyone at a similar distance. If you can, add a second recorder at the other end as a backup.
• Ask people to say their label the first time they speak. "P3, I think..." takes a second and saves minutes of guessing later.
• Agree on one rule for the session: one person at a time. It will not always hold, but it helps the moderator steer and helps everyone who listens to the recording later.

Record a short test

Thirty seconds, played back on headphones, tells you whether the person furthest from the recorder can still be heard. Background music, a coffee machine or an open window make every voice harder to separate.

Transcribe the recording

1. Upload the file. Audio and video work as they are, including MP4 from an online session. You can upload several sessions at once.
2. Set the number of speakers if you know it, or leave it on Auto. Speakers are separated automatically.
3. Add names of products, places and specialist terms to the glossary first, so they come back spelled right.

The speaker setting goes up to six. With a moderator and seven participants, leave it on Auto and plan for more fixing in the next step: merging voices and moving lines. The seating plan and the labels people said aloud are what make that quick.

A 90-minute session needs a paid plan: the free plan takes files up to 30 minutes. See the cost example below.

Give every speaker their label

Open the transcript next to your seating plan.

• Rename each speaker once. Change "Speaker 4" to "P4" and the label changes across the whole transcript.
• Merge a voice that was split in two. In a long group discussion, one person is sometimes recognised as two speakers. Merge them into one.
• Move a line to the right person. When two people talk over each other, a line can land with the wrong speaker. Click the line to hear it, then move it.
• If you cannot tell who it is, write P? A guess that turns out wrong is worse than an honest question mark.

Mark crosstalk and group reactions

These conventions are common in focus group transcripts. Pick yours, write them at the top of the transcript and use them in every session.

You add these tags yourself, in the editor or after export. Our free focus group transcript template has a participant key, these conventions and a transcript table ready to fill in.

Check before you code

A machine transcript is a draft, and a focus group is the hardest kind of recording to transcribe.

• Start with the underlined words. Words the engine was unsure about are underlined with dots, and every transcript shows the share of uncertain words. Click a line and the recording plays from that moment.
• Listen to every passage you plan to quote, and check that it is attributed to the right participant.
• Check a quote shows whether a sentence was said exactly like that in the recording, and takes you to that moment. It is on every plan, including Free.

Verbatim or clean verbatim

Decide before you start checking. Verbatim keeps fillers, repetitions and false starts; clean verbatim leaves them out without rewording anything. In SonaFact the transcript stays word for word and the clean view hides "um", "uh" and stuttered repeats. Our transcription style guide shows both styles side by side, if your team needs to agree on one.

Export for analysis

Export as DOCX or PDF, or as files for NVivo, MAXQDA and ATLAS.ti, with the speaker label at the start of every turn: how the import works.

Keep the key that links labels to real names apart from the transcripts. Our privacy checklist lists the questions worth asking about where participants' recordings are stored. When the project is over, you can delete each recording, or let SonaFact delete recordings automatically after the time you choose.

What it costs: three sessions of 90 minutes

You can also pay once for 30 days, with no renewal. Prices in your currency are on the pricing page.

Checklist

• Consent from every participant, on paper and on the recording.
• Seating plan with labels: M, P1, P2...
• Recorder in the middle of the table, test played back.
• Speakers renamed, split voices merged, misplaced lines moved.
• Crosstalk, group reactions and unclear passages marked.
• Every quote checked against the recording.

For a one-to-one interview, see how to transcribe an interview.

Transcribe your first focus group recording: free for files up to 30 minutes, no card.

Try it on your own recording

The free plan gives you 3 transcripts a day, up to 30 minutes each, without a card — enough to judge the transcript, the speakers and the editor yourself.