Interview Transcription for Research, Verbatim Rules and Privacy

An interview is only as useful as the text it leaves behind. A PhD candidate coding thirty hours of expert conversations, an HR team documenting an internal investigation and a market researcher comparing focus groups in three countries all depend on the same thing: a transcript they can trust. Interview transcription looks simple from the outside. In practice it involves decisions about style, accuracy, privacy and language that shape every conclusion drawn from the material.

Why interview recordings are harder than they sound

Interviews are not speeches. People interrupt each other, trail off, change their minds mid sentence and lower their voice exactly when they say something sensitive. Add a café background, a weak laptop microphone or a participant speaking in their second language, and even experienced listeners need to replay passages several times. As a working rule, a clear one hour interview takes a professional transcriber around four hours. Poor audio, heavy accents or several speakers can push that well beyond six.

Automatic speech recognition helps with first drafts, but it fails in predictable places. It merges speakers, invents words where the audio is unclear and confidently misspells names, brands and technical terms. In research, an invented word is worse than a gap, because nobody knows it is there.

Choose the right verbatim style before you start

The style of transcript should follow from how you will analyse it. Changing your mind halfway through a project means re-transcribing hours of audio.

Whatever you choose, write it down in a short style sheet. It should cover speaker labels (for example "I" for interviewer and "P1" for participant), how often to insert timestamps, how to mark inaudible words, and whether to note non-verbal cues such as [laughs] or [long pause].

Privacy is part of the job, not an extra

Interview recordings often contain the most sensitive data a project holds. Participants talk about health, money, religion, colleagues and employers. In Europe, health data and religious or political views are special categories of personal data under Article 9 of the General Data Protection Regulation, and anyone who processes recordings on your behalf is a processor who needs a written agreement under Article 28.

Before sending a single file, check that your transcription provider will:

• sign a data processing agreement and a confidentiality agreement for every transcriber;
• transfer files through an encrypted portal, never by email attachment;
• pseudonymise names, places and employers in the transcript if your ethics approval requires it;
• delete audio and working files on a fixed schedule after delivery.

Free online tools rarely meet these conditions. Uploading a participant's voice to a service that keeps the data to train its models can breach the consent your participants gave you.

Interviews in a foreign language

International studies add another layer. If you interview engineers in Germany or patients in France, you need a transcript in the original language first, done by a native speaker who catches dialect, irony and technical vocabulary. Only then should the text be translated for analysis. Translating directly from audio removes the source text you need to check quotes, and reviewers of academic papers increasingly ask for it.

When the researcher does not speak the participant's language, the interview itself can be supported by an interpreter. Remote interpretation services make this practical for video interviews, and the recording then contains both languages, which the transcriber labels separately. Professional language transcription services handle this kind of bilingual audio routinely, while most automatic tools break down as soon as the language switches.

For publication, quotes translated from the transcript should be reviewed by a subject specialist. A participant's hedged "I suppose it could work" carries a very different weight from "it works", and a careless translation can flatten that nuance. This is where professional translation services earn their place in the research workflow.

What drives the price

Interview transcription is usually priced per audio minute. Expect the rate to rise with:

• the number of speakers and how much they overlap;
• audio quality and background noise;
• full verbatim instead of clean verbatim;
• specialist vocabulary in medicine, law or engineering;
• rare languages or strong regional dialects;
• short turnaround times.

The cheapest way to lower the bill is better audio. Use an external microphone, record each participant on a separate track for online calls, and ask people to introduce themselves at the start so the transcriber can identify voices.

FAQ

How long does it take to transcribe a one hour interview?

A professional usually needs about four hours of work for clear audio, and more for difficult recordings.

Is AI transcription good enough for qualitative research?

It can produce a draft, but every transcript used for analysis or quotation should be checked by a human against the recording.

Do I need a data processing agreement with my transcriber?

Under GDPR, yes, whenever the provider processes personal data in recordings on your behalf.

Good transcripts are invisible. Nobody notices them until a quote is challenged, and then they are the only thing that matters.