Transcription decisions that shape what you can claim, what focus group data demands that interview data does not, and the gap between what people say and what they do.
Analysing interviews means working with accounts — what people say about what they do, think and experienced — and keeping that distinction visible in your claims. Focus groups add a second layer, because what is said is shaped by who else is in the room, which means the interaction is data rather than noise. Both start with a transcription decision that quietly determines what you will be able to argue later.
Transcription is usually treated as an administrative chore. It is actually the first analytic act, because it decides what survives into your dataset.
A clean verbatim transcript removes stumbles, repetitions and fillers. It reads well and suits analysis of content. But it removes hesitation — and if part of your finding is that people hesitated before answering a particular question, you have deleted your own evidence.
A fuller transcript preserves pauses, overlaps, emphasis and false starts. It takes considerably longer, is harder to read, and is necessary if how something was said is part of what you are analysing. Approaches concerned with language as action generally require it.
Decide before you start, apply it consistently, and state the convention in the methodology chapter. Two further points worth attention:
Transcribing at least some of your own interviews is worth the time even if you outsource the rest. It is where familiarisation actually happens.
Interviews give you what people are willing and able to tell you, in a particular setting, to a particular person, about themselves. That is genuinely valuable, and it is not a direct record of what happened.
People misremember, reconstruct events to make sense in hindsight, and present themselves favourably — not usually through dishonesty but because that is how accounts work. And who you are shapes what you are told: an insider gets different material from an outsider, and a researcher perceived as connected to management gets different answers from one who is not.
None of this undermines interview research. It does determine how findings should be worded. “Participants reported that supervision was inconsistent” is defensible. “Supervision was inconsistent” is a claim about the world that interviews alone did not establish. Examiners notice the difference, and the fix is a wording habit rather than more data.
Where the gap between account and practice is itself interesting, say so. That people describe following a protocol they demonstrably do not follow is a finding, not a problem with your data.
The most common error in focus group analysis is treating the transcript as several interviews that happened simultaneously. If that were the aim, the design would be wrong — separate interviews would have produced better data with less trouble.
The point of a focus group is the interaction: how a view is offered, challenged, defended, modified or abandoned when others are present. So the analysis has to attend to things interview analysis does not.
Look at what happens between contributions, not just within them. Where does agreement build, and does it build genuinely or through nobody wanting to dissent first? Where does someone soften a position after a challenge? Who speaks after whom, and what does that ordering do?
An analysis reporting only what was said, with no account of the group dynamic, has thrown away the reason for choosing the method — and that is a reasonable question for an examiner to ask.
Dominance. One participant setting the frame early, with others positioning relative to them. Note it; do not pretend the transcript is a set of independent views.
Conformity. Groups drift towards a socially acceptable position, particularly where participants know each other or work together. Silence is not agreement, and should not be reported as consensus.
Composition. Mixing hierarchical levels — managers with those they manage — reliably suppresses material. If it happened, it is a limitation to state rather than a fact to omit.
Attribution is the main one. Overlapping speech in a group of eight is genuinely hard to assign correctly, and misattribution damages any claim about how positions developed. A seating plan, an audio setup that separates voices, and participants stating their name early all help. Video helps most and raises its own consent questions.
The analytic approach itself — thematic, content, grounded theory, phenomenological — comes from your design rather than from the data type. What interview and focus group material adds is a set of practical demands.
Extracts are evidence for your interpretation. They do not make the argument by themselves, and a chapter built from long quotations joined by linking sentences has handed the analytic work to the reader.
Practical conventions worth following:
Read each transcript whole before coding anything, then code systematically using the approach your design specifies — thematic, content, grounded theory or phenomenological. Work on the codes rather than the transcripts to develop themes, check those against the whole dataset, and look deliberately for material that contradicts them. Keep memos throughout: that is where the interpretation actually forms.
It depends what you intend to claim. Clean verbatim removes fillers and false starts, reads better and suits analysis of content. Fuller transcription preserving pauses, overlaps and emphasis is necessary if how something was said is part of your analysis, and is generally required for approaches treating language as action. Decide before you begin, apply it consistently, and state the convention in your methodology chapter.
The interaction is data, not background. A focus group analysis that reports only what was said, without attending to how views were challenged, defended or modified in the room, has discarded the reason for choosing the method. You also have to account for group effects — dominance, conformity, and the influence of who was in the room together — and to treat silence as silence rather than as agreement.
Yes, as a first pass, but it has to be checked against the audio. The errors are systematic rather than random — accents, overlapping speech, technical vocabulary and names are handled worst — which means uncorrected errors can quietly distort a theme. Transcribing at least some interviews yourself is also worth the time, because it is where familiarisation genuinely happens.
Enough for a reader to judge whether your interpretation is warranted, and not so many that the chapter becomes a collection of extracts with commentary. Draw from several participants rather than repeatedly from the most articulate one, choose extracts that are representative rather than merely striking, and make sure each one is doing work in the argument rather than illustrating a point already made.
Problems in interview data are usually visible in the first two transcripts, while there is still time to change the guide. Describe your study and send where you have got to, and a PhD in your field will read it with you.
Leave an email or a WhatsApp number — whichever you prefer — and tick how we should reach you. We reach out within 30 minutes.
