Transcribing a focus group: how to tell many voices apart

Why is a focus group harder to transcribe than an interview?

In a one-on-one interview, two people take turns. In a focus group, you have several voices at once, and people cut in, agree out loud, laugh together, and finish each other's sentences. Short remarks like "exactly" or "yes, I think so" come up all the time, and they are the hardest lines to pin to a specific person.

The room itself doesn't help. One recorder in the middle of the table picks up the people sitting close to it and loses the ones further away. Chairs creak, cups clink, and people turn their heads while they talk. Audio that sounded fine when you were sitting in the room can be tough to follow when you play it back.

Plan for this before the group sits down. Any lively discussion is going to leave you with a few unclear spots in the transcript. You won't fix all of them, so make sure you know where they are, and never build your analysis around a quote from one of those passages.

How do you set up the room and the recording?

Seat everyone in a circle or a horseshoe, and put a tent card in front of each person with their participant code or first name. Put your main recorder in the middle. If you have a second recorder, set it near the far end of the table so you have a backup and a second angle on the room. Run a quick test recording and listen back on headphones before you start.

Go over a few ground rules at the start. Ask people to speak one at a time, wait for the moderator if the conversation gets heated, and say their name or code the first time they speak, if your protocol allows it. Those simple habits keep the discussion in order and make the audio much easier to work with.

Sketch a seating chart as people sit down. While the group is talking, jot down a quick note with a timestamp whenever something happens that the microphone will miss, like a nod, a facial expression, or someone pointing at a visual aid. Those notes will help you make sense of the transcript later.

What should you tell participants about consent and confidentiality?

Let participants know that you are recording the audio and transcribing it. Tell them who will see the files, how you will protect their identities, and how long you will keep the recording. Get their explicit agreement to be recorded, and have another way for someone to take part if they say no. Rules around recording and personal data depend on the country and the institution, so follow whatever your ethics committee or legal adviser tells you.

Be upfront about the limits of confidentiality. While you can secure the transcript, you cannot stop group members from repeating things once they leave the room. Ask participants to keep what others say private, and explain clearly that this is a shared request rather than a guarantee.

How do you label speakers in the transcript?

Give everyone a consistent code, like M for the moderator and P1 through P8 for the participants, and keep the key in a separate file. Take real names out of your working transcript. If your analysis needs details like age band or job role, put those in a separate table linked only by the participant codes.

When you upload your audio to KolWrite, it creates a transcript with timestamps and speaker labels. As you go through it, change the generic labels to your participant codes, and those updates will carry over into every file you export. Still, treat automatic speaker separation as a rough draft. With a whole group talking, software will sometimes lump two similar-sounding people together, split one person into two speakers, or tack a quick interjection onto whoever was just talking.

If you cannot tell who said something, tag it as an unknown speaker. Pinning a quote on the wrong person does much more damage to your analysis than leaving a line unattributed.

How do you check a transcript with overlap?

Play the recording all the way through with the transcript open in front of you. Pause whenever the speaker changes to check the label and fix the wording. Give the most time to fast-moving exchanges, since both the text and the speaker labels are least reliable there.

Use a consistent notation to mark overlapping speech, and note whenever two people talked at once and one of them could not be heard. Decide with your team how to handle backchannel remarks, since some analyses count them and others leave them out. Automatic transcription can also mishear specific details, so check every name, number, and direct quote against the audio.

Make sure every quote you plan to publish is tied to the right participant code. Your readers are counting on you to get that right.

How do you analyze a group transcript?

Read the whole transcript from start to finish before you start coding it by theme. With a focus group, how people interact is part of what you are looking for. Pay attention to where people agreed, who changed their mind after listening to someone else, and which topics brought out tension or silence. Note the timestamps for those moments so you can go straight back to the audio and listen again.

Set up a simple matrix with your themes as rows and participant codes as columns, then drop short quotes into the cells. That lets you see right away whether a theme came from the whole group or just one loud voice, and it shows you which participants stayed quiet. You can also export the transcript as a DOCX, TXT, or JSON file to pull it into your qualitative analysis tool.

When you write up your findings, report what the group said as a group. A focus group shows you how people talk through a topic together. It is not a survey, so avoid turning the discussion into headcounts and generalizing from them.