Thematic Analysis Step by Step: From Transcripts to Themes
Thematic analysis turns interview and open-ended data into themes that answer your research question. This guide follows the widely used six-phase approach of Braun and Clarke, with practical tips for each phase.
Before you start: decide your approach
Decide whether your analysis will be inductive (codes come from the data) or deductive (you code against a framework), and whether themes will capture meaning at the surface (semantic) or beneath it (latent). Write your research question down. It guides every later choice.
Phase 1: Familiarise yourself with the data
Read the transcripts fully, more than once. If you transcribed them yourself, note first impressions. Check transcripts against recordings, anonymise names and keep a short analytic diary. To explore the text quickly, our free word frequency counter and keyword in context viewer show what comes up most and how key terms are used.
Phase 2: Generate initial codes
A code is a short label for a meaningful piece of data. Work through the whole dataset and code every segment relevant to your question. Use participants’ own words where helpful. Expect many codes at first; you will refine them later. Software such as NVivo, MAXQDA or ATLAS.ti keeps codes linked to the original text, but a well-organised spreadsheet works for small projects.
Phase 3: Search for themes
Group related codes into candidate themes. A theme is a pattern of meaning, not just a topic. “Barriers” is a topic; “trust in the health system is built through continuity of care” is closer to a theme. Sketch a map showing how codes cluster. Some codes will not fit anywhere; keep them aside for now.
Phase 4: Review the themes
Check candidate themes at two levels. First, do the coded extracts for each theme form a coherent pattern? Second, does each theme fit the whole dataset and answer your question? Split themes that are too broad, merge ones that overlap and drop those with too little data.
Phase 5: Define and name the themes
Write a short description of each theme: what it is about, what it includes and how it relates to your question. Choose clear, informative names. This is also the moment to select the best illustrative quotations.
Phase 6: Write up
Present themes as a story, not a list. For each theme, state the point, give two or three well-chosen quotations with participant identifiers and explain what they show. Tie the themes back to your research question and the literature.
How to make your analysis credible
- Keep a codebook with definitions and example quotations.
- Write memos recording why you merged, split or renamed codes.
- Where two people code, check agreement and discuss differences.
- Be transparent about your role and assumptions (reflexivity).
- Avoid presenting counts as proof; themes are about meaning, not frequency alone.
Thematic analysis or something else?
If you need to count categories systematically, see content analysis. If you want to build theory, see grounded theory.
A worked mini-example
Suppose a participant in a study of working parents says: “I never know if I’m going to be called in on my day off, so I can’t plan anything with my kids.” You might apply these codes: unpredictable schedule, difficulty planning family time and work intruding on days off. Across other interviews, those codes group with “last-minute shift changes” and “no say in rota” under a candidate theme such as lack of control over working time shapes family life. The theme is a statement of meaning that answers the research question, not just the label “scheduling”.
What a codebook entry looks like
- Code name: Unpredictable schedule
- Definition: The participant describes working hours that change at short notice.
- Include when: short-notice changes, being called in, cancelled shifts.
- Exclude when: a fixed but long working day (code as long hours).
- Example quote: “I never know if I’m going to be called in on my day off.”
Reflexive or codebook thematic analysis?
Braun and Clarke describe different styles. Reflexive thematic analysis treats coding as an interpretive and subjective process, so themes are generated by the researcher and inter-coder agreement is not the goal. Codebook approaches use a structured coding frame and are often chosen when teams need to code consistently, for example in applied evaluations. State which one you used and why, because reviewers expect the method to match the philosophy.
How to write up the findings
- Open each theme with a clear sentence that states its main point.
- Support it with two or three short quotations, each labelled with a participant identifier such as P7.
- Explain what each quotation shows instead of letting the quote speak alone.
- Show how the themes relate to each other and to the research question.
- Describe your analysis in the methods section: software, phases, coding approach and how you checked quality.
Judging the quality of your analysis
Qualitative researchers often use the criteria of credibility, transferability, dependability and confirmability. In practice this means keeping an audit trail, describing the context so readers can judge how far findings apply, checking interpretations with the data (and sometimes with participants), and being open about your own influence on the analysis.
Common mistakes
- Presenting topics as themes. “Childcare” is a topic. “Childcare costs force parents to choose between working and caring” is a theme.
- Saying themes “emerged” from the data. Themes are produced through your interpretation, so describe your active role.
- Using one quotation per theme. A theme should be supported by data from several participants.
- Coding only what is interesting. Code systematically across the whole dataset first.
- Counting quotations as proof. Prevalence can be reported, but the strength of a theme comes from its meaning.
Frequently asked questions
How many interviews do I need for thematic analysis?
There is no fixed number. Small studies often use 8 to 20 interviews. What matters is whether the data are rich enough to answer your question.
What is the difference between a code and a theme?
A code labels a specific idea. A theme groups related codes into a broader pattern of meaning that answers your research question.
Do I need software like NVivo?
No. For small datasets a spreadsheet can work. Software becomes helpful for larger projects and for tracking codes back to the original text.
How many themes should I have?
Usually a handful of well-developed themes. More themes often means they are still topics and need merging.
Can I use AI tools to help?
AI can help you explore data or suggest candidate codes, but you remain responsible for checking every code and theme against the data, and you should follow your ethics approval and data-protection rules before sharing any participant data.
Related tools and services
Want an expert to check or run your analysis?
Send us your question and data description. We reply with a clear plan and a quote.
Discuss your project on WhatsApp