Thematic analysis, step by step

What is thematic analysis, and which version are you using?

Thematic analysis is a way of finding, organising and interpreting patterns of meaning across a dataset. It is flexible: it is not tied to one theory, and it works with interviews, focus groups, open survey responses, diaries or documents. That flexibility is why it is popular, and also why it is often done badly.

The most important thing to understand is that “thematic analysis” is not one method. Braun and Clarke have written several papers pointing out that many studies cite their 2006 paper while doing something quite different from it. In a 2021 paper in Qualitative Research in Psychology (“One size fits all?”) they described three broad families:

Decide which one you are doing, and say so in your methodology chapter. Mixing them, for example citing Braun and Clarke and then reporting a kappa score to prove your themes are “reliable”, is one of the problems examiners familiar with the literature will pick up. How reliability is judged in quantitative studies is a different question, covered in our reliability and validity guide.

Choices to make before coding

• Inductive or deductive. Are codes driven mainly by the data, or by an existing theory or framework? Most studies mix both; say which leads.
• Semantic or latent. Are you coding what participants explicitly say, or the assumptions and ideas beneath it?
• Philosophical position. Broadly realist (the data tells you about participants’ reality) or constructionist (the data shows how meaning is constructed). This affects how you write claims.

What are the six phases of thematic analysis?

Braun and Clarke describe the phases as a guide rather than a strict sequence. You will move back and forth, especially between phases 3, 4 and 5. The names changed slightly between the 2006 paper and the 2022 book, and the changes are deliberate: they stress that themes are built by the analyst, not found.

Phase 1: Familiarisation

Transcribe your own interviews if you can; it is slow, but it is the deepest familiarisation you will get. If you use a transcription service or automatic transcription, check every transcript against the recording. Read each transcript at least twice. Keep brief notes of ideas, surprises and your own reactions. These notes are the start of your reflexive record. Planning and recording the interviews themselves is covered in our semi-structured interviews guide.

If interviews were in Tamil, Hindi or another language, decide early whether to analyse in the original language and translate only the extracts you quote, or to translate everything first. Analysing in the original usually keeps meaning better. Either way, say what you did.

Phase 2: Coding

Work through the whole dataset, labelling every segment relevant to your research question. A code is a short label that captures what is interesting about the segment: “feeling judged by senior nurses” rather than “emotions”. Code everything relevant, including what does not fit your early ideas. Expect to revise codes as you go; go through the data at least twice.

NVivo, ATLAS.ti and MAXQDA are the main commercial tools; Taguette and QualCoder are free. Word comments or a spreadsheet also work for smaller datasets. Software organises coding; it does not do the analysis.

Phase 3: Generating initial themes

Group codes that seem to share a central idea into candidate themes. A theme, in Braun and Clarke’s sense, is a pattern of shared meaning organised around a central concept. It is not a bucket for everything said about a topic. Draw a thematic map; seeing the relationships helps.

Phase 4: Developing and reviewing themes

Check each candidate theme at two levels. First, against its coded extracts: do they fit together? Second, against the whole dataset: does the theme reflect the data, and does anything important sit outside all your themes? Merge, split, discard. Many good analyses change substantially here.

Phase 5: Refining, defining and naming themes

Write a few sentences for each theme saying what it is about, what it is not about, and how it relates to other themes. If you cannot state a theme’s central idea in two sentences, it is probably not yet a theme. Choose names that say something (“Proving I belong”) rather than labels (“Identity”). Sub-themes are fine, but keep them few.

Phase 6: Writing up

In reflexive TA, writing is part of analysis, not a report added afterwards. Present each theme as an argument: what it is, illustrated with extracts, with your interpretation around each extract. Extracts support the analysis; they do not replace it. Then connect the themes to your research question and to earlier work, which the discussion guide covers.

What is the difference between a theme and a topic summary?

This is the most common problem in PhD thematic analyses, and one that Braun and Clarke have written about repeatedly. A topic summary collects everything participants said about a subject, often an interview question: “Challenges faced”, “Support received”, “Suggestions for improvement”. It summarises; it does not interpret.

A theme, in reflexive TA, has a central organising idea that says something about the data. Compare:

• Topic summary: “Challenges faced by first-generation PhD scholars”, followed by a list of financial, academic and family challenges.
• Theme: “Carrying the family’s investment”, capturing how scholars described financial, academic and family pressures as a single sense of debt to parents who had sacrificed for their education.

The second tells the reader something they did not know from the interview schedule. A quick test: if your theme names match your interview questions, you probably have topic summaries.

What mistakes do examiners look for?

Braun and Clarke’s 2021 “One size fits all?” paper lists common problems in published TA. The ones we see most often in Indian theses:

• Themes that “emerged”. Themes do not emerge on their own; you develop them. Using “emerged” suggests a passive view of analysis that does not fit reflexive TA.
• Citing Braun and Clarke for a coding reliability procedure. Kappa scores, multiple coders reaching “agreement”, and a fixed codebook belong to different approaches.
• Claiming saturation. In their 2021 paper “To saturate or not to saturate?” in Qualitative Research in Sport, Exercise and Health, Braun and Clarke argue that saturation does not fit reflexive TA, because meaning is generated by the analyst, not waiting to be exhausted. Justify sample size by the aims, the richness of the data and practical limits.
• Too many themes. Ten themes with four sub-themes each is usually a list of codes. Most good analyses have two to six main themes.
• Extracts with no analysis. Long quotations joined by “as one participant said” is description, not analysis.
• No reflexivity. In reflexive TA your perspective shapes the analysis. Describe your position (a teacher studying teachers, an insider in the community) and how you reflected on it.

• The methodology chapter names the version of TA you used (reflexive, codebook, coding reliability) and the paper or book you followed.
• You state your orientation: inductive or deductive, semantic or latent, and your philosophical position.
• Each theme has a central organising idea, not just a topic heading such as “Challenges” or “Benefits”.
• Themes are described as developed or constructed by you, not as having “emerged”.
• Each theme is illustrated with extracts from several participants, with analytic commentary around each extract.
• The number of themes is manageable, typically two to six main themes for a study or chapter.
• Your sample size is justified by the study’s aims and the richness of data, rather than by a claim of saturation alone.
• You describe your own position and how you reflected on it.

The 2006 paper also includes a 15-point checklist of criteria for good thematic analysis, which examiners sometimes use. Byrne’s 2022 worked example in Quality & Quantity is a useful model of how to describe each phase in a methods section.

How do you write thematic analysis in a PhD thesis?

In the methodology chapter

Name the approach and cite the version you followed: for reflexive TA, Braun and Clarke’s 2006 paper plus their later work (the 2019 “Reflecting on reflexive thematic analysis” paper and the 2022 book are the usual citations). Describe how you carried out each phase in your study, not a textbook summary. State your orientation (inductive or deductive, semantic or latent), your position as a researcher, how data was transcribed and translated, and what software you used.

In the findings chapter

Start with an overview: a table or thematic map of themes and sub-themes. Then give each theme its own section. For each, define the central idea, present extracts from several participants (identified by pseudonym or code), and interpret each extract. Some examiners also like to see how many participants contributed to each theme; if you give counts, explain that in reflexive TA frequency does not decide importance.

Using AI tools for coding

AI tools that generate codes or themes from transcripts are now widely marketed. Reflexive TA depends on the researcher’s interpretation, so handing it to a tool undermines the method, and uploading interview transcripts to a public AI service can breach the confidentiality promised in your consent form and ethics approval. See our guide to AI tools in a PhD before using any.