Qualitative data — interview transcripts, field notes, open-ended survey answers and documents — tells you how people understand their experiences, what something means to them, and why they act as they do. It captures context, reasoning and emotion that numbers alone usually miss, and it can explain the story behind a pattern that a survey only measures.
This guide walks through the analysis of interview data step by step: preparing transcripts, reading closely, coding, building themes, checking your reading against the data, and writing up findings honestly. It assumes no specialist software and no prior experience, only a set of transcripts and the patience to read them carefully more than once.
Qualitative data is strong where meaning matters. It can show how veterinary students experience their first clinical placement, why small farmers adopt one feeding practice and reject another, or what patients find reassuring about a consultation. It reveals processes, turning points and contradictions in ways that a questionnaire, with its fixed answers, rarely can.
It cannot, on its own, prove how common a view is in a wider population, or establish that one factor causes another. Twenty interviews that all describe the same difficulty do not mean that difficulty affects most people in the population — they mean it matters deeply to the people you spoke to. Respect that boundary when you write up your findings, and your conclusions will be stronger for it.
Good analysis starts before any coding. Transcribe interviews accurately, including pauses, laughter and hesitations where they carry meaning, and check each transcript against the recording at least once. Give every participant a pseudonym, remove names of people and places that could identify them, and store recordings and transcripts separately in a secure location.
Then organise the material so you can work with it. Keep one clean copy of each transcript, a short note on the setting of each interview, and a simple log of decisions you make along the way. This audit trail — what you coded, what you merged, what you set aside — is what lets you, a supervisor or a reviewer retrace your steps later, and it costs almost nothing to keep.
Resist the urge to open your coding tool or highlighter on the first pass. Read every transcript from start to finish, slowly, as a whole conversation rather than a collection of quotes. Note your first impressions in the margin: what surprised you, what repeated, what seemed tense or unfinished, and where one participant contradicted another.
Read a second time with your research question in mind, and write a short summary — no more than a paragraph — for each interview. These summaries keep individual voices intact and stop you from treating every participant as interchangeable. They also make the next stage, coding, far faster, because you already know the shape of each conversation before you start breaking it into parts.
Open coding means going through the text line by line or passage by passage and giving each meaningful segment a short label that captures what is happening in it. A label might be "fear of making a mistake", "learning by watching", or "cost as a barrier". Keep labels close to the participants' own words at this stage, and do not worry if you end up with many of them.
Code generously rather than sparingly. A segment can carry more than one code, and a code that later proves unimportant can always be dropped, while a passage you skipped is easily forgotten. Work through one transcript fully before moving to the next, and revisit earlier transcripts when a new code appears, so that all interviews are coded to the same standard.
Once coding is done, the task changes from breaking data apart to putting it back together. Group codes that belong together into categories: for example, "fear of making a mistake", "avoiding difficult cases" and "asking a senior for help" might form a category about managing uncertainty. Categories describe what is in the data; themes go one step further and say something about it.
A theme is a pattern of meaning that answers part of your research question, such as "new graduates manage uncertainty by borrowing the confidence of others". Test each candidate theme by asking whether it is supported by more than one participant, whether it holds across different interviews, and whether you can state it in a single clear sentence. If you cannot, it is probably still a category, not yet a theme.
Thematic analysis is the most widely used approach for interview data, and its steps are straightforward. First, familiarise yourself with the data by reading and re-reading. Second, generate initial codes across the whole data set. Third, search for themes by grouping codes. Fourth, review the themes against the coded extracts and then against the full data set.
Fifth, define and name each theme, writing a short description of what it covers and what it does not. Sixth, write up the analysis, selecting vivid extracts that illustrate each theme and explaining how the themes relate to one another and to your research question. Keep a note of each step as you go, because a method section that describes what you actually did is far more convincing than one that only names the approach.
A theme that sounds persuasive in a summary can fall apart when you return to the transcripts, so check every theme twice. First, read all the extracts grouped under a theme together and ask whether they genuinely belong together, or whether two different ideas have been forced into one label. Split or merge themes until each one feels coherent.
Second, re-read the full data set with the themes in mind and ask whether anything important has been left out, or whether a theme distorts what a participant actually meant. If a participant's words contradict your theme, do not quietly drop that participant. The contradiction may be the most interesting finding in your study, and it certainly deserves a place in your write-up.
Disagreement inside your data is normal, and it strengthens rather than weakens an analysis when it is handled openly. When two participants describe the same situation in opposite ways, report both, and consider what might explain the difference: different roles, different experience, different settings. A theme that only fits half your participants needs to say so plainly.
Negative cases — participants whose experience runs against the emerging pattern — deserve special attention. Actively search for them rather than hoping they do not exist, and let them refine your theme: perhaps the pattern holds only for beginners, or only in one kind of setting. An analysis that acknowledges its exceptions is more trustworthy than one that claims a uniformity the data do not support.
Quotations are evidence, not decoration. Choose extracts that show the theme clearly and let the reader hear the participant's voice, rather than the most dramatic line you can find. Keep quotations intact enough that their meaning survives: trimming a sentence is acceptable, but editing that changes what a person meant is not, and any omission should be marked clearly.
Protect your participants in every quotation. Remove identifying details, be careful with rare combinations of facts that could reveal someone even without a name, and remember that consent to be interviewed is not unlimited permission to publish anything said. If a quotation could harm or embarrass a participant, either anonymise it further or leave it out, however useful it seems.
A strong write-up shows the reader how you reached your conclusions. Present each theme in turn, explain what it means, support it with two or three well-chosen extracts, and comment on how widespread or varied it was across participants. Then connect the theme back to your research question, and where relevant, to earlier literature, noting where your findings agree and where they differ.
Avoid the common trap of asserting a theme and then offering a single quotation as proof. One voice cannot carry a pattern, and a paragraph of interpretation with no data behind it asks the reader to take your word for it. Balance is the test: every claim about the data should be traceable to extracts the reader can see, and every extract should be followed by your interpretation of what it shows.
Imagine three newly qualified veterinary graduates describing their first months in practice. The first says she avoided difficult cases and asked a senior colleague to stand beside her. The second says he rehearsed procedures the night before and kept a notebook of what went wrong. The third says she felt she was "pretending to be a vet" until a senior vet praised her handling of a case.
Three different behaviours, but one shared pattern: each graduate managed the gap between qualification and confidence by borrowing support from elsewhere — a colleague's presence, a private routine, or someone else's approval. Naming the theme "borrowing confidence" captures that pattern in a way none of the individual codes could, and each interview then supplies a different kind of evidence for it. That move, from separate codes to a shared pattern, is the heart of qualitative analysis.
Before you write up, run through this checklist honestly. Have you read every transcript at least twice? Are all transcripts coded to the same standard, with later codes applied back to earlier interviews? Can you state each theme in one sentence, support it with extracts from more than one participant, and explain what does not fit? If any answer is no, that stage still needs work.
Also check the practical safeguards: are transcripts anonymised and stored securely, are quotations checked against the original wording, and does your method section describe what you actually did, step by step? A clean audit trail and honest reporting of exceptions will satisfy a reviewer far more than a polished-sounding method section that the data cannot support. When all of these boxes are ticked, your analysis is ready to write up.
| Step | What you do | Output |
|---|---|---|
| Familiarisation | Read transcripts fully, note first impressions | One summary per interview |
| Open coding | Label meaningful passages close to the data | A set of codes across all interviews |
| Grouping | Sort codes into categories, then themes | Candidate themes with extracts |
| Reviewing | Check themes against extracts and full data | Revised, coherent themes |
| Defining | Name each theme and describe its scope | Theme definitions for the write-up |
| Writing up | Present themes with evidence and interpretation | Findings section of the paper |
These external resources explain qualitative methods and tools in more depth:
How many interviews do I need for a qualitative study?
There is no fixed number. What matters is whether your data answer your research question in enough depth, and whether new interviews keep adding new ideas or start repeating what you have already heard. For a focused study, a small, carefully chosen sample analysed deeply is stronger than a large sample skimmed quickly.
Do I need specialist software to code qualitative data?
No. Software such as NVivo or MAXQDA helps organise large data sets, but coding can be done well with a word processor, a spreadsheet, or even printed transcripts and coloured pens. The quality of the analysis depends on careful reading and honest checking, not on the tool you use.
What is the difference between a code and a theme?
A code is a short label for one idea in one passage of data. A theme is a broader pattern of meaning across the data set that answers part of your research question. Codes describe what is there; themes interpret what it means, and a theme is usually supported by several codes and several participants.
How do I know when my analysis is finished?
When every theme is coherent, supported by extracts from more than one participant, checked against the full data set, and stated clearly enough to write up — and when new reading of the data stops changing the themes. If a theme still shifts every time you re-read, it needs more work before you write it up.
Can I use quotations from participants in my published paper?
Yes, provided participants consented to quotation, identifying details are removed, and quotations are presented fairly without changing their meaning. Check your consent forms and your institution's requirements, and remove any detail that could identify a participant even without a name.
What should I do if my themes do not match what I expected to find?
Report what the data actually show, not what you hoped to find. Unexpected themes are often the most valuable part of a qualitative study. Explain the surprise honestly, check that it is supported by the data rather than a single striking quotation, and discuss what it might mean for practice or future research.