Reflexive Thematic Analysis: Braun and Clarke's Approach in Practice

Short answer: Reflexive thematic analysis (reflexive TA) is Virginia Braun and Victoria Clarke's version of thematic analysis, in which themes are patterns of shared meaning that the researcher actively builds through interpretation, not topics that "emerge" from the data. You work through six recursive phases (familiarisation, coding, generating initial themes, developing and reviewing themes, refining and naming themes, writing up), you treat your own perspective as a resource rather than a bias to remove, and you do not use inter-coder agreement scores to prove quality. Use it when you want an interpretive account of what experiences mean. Use a codebook or coding reliability approach instead when you need fixed categories that several coders apply the same way.

This guide explains what makes reflexive TA different from other kinds of thematic analysis, walks through each phase with practical examples from product and user research, lists the mistakes reviewers flag most often, and shows where software (including Koji) helps and where it cannot replace the researcher.

Why "reflexive" thematic analysis exists

Braun and Clarke's 2006 paper, "Using thematic analysis in psychology", is one of the most cited academic papers of the century. The University of Auckland reported in 2025 that it has been cited between 100,000 and 230,000 times, depending on the database. That popularity created a problem. Many papers cite Braun and Clarke while doing something quite different from what the authors meant: counting topics, measuring coder agreement, or presenting a list of interview questions as "themes".

So the authors started calling their approach reflexive TA, to separate it from two other families of thematic analysis:

In a 2019 handbook chapter, Braun, Clarke, Hayfield and Terry warned that "the plurality of TA is often not recognised by editors, reviewers or authors, who promote 'coding reliability measures' as universal requirements of quality TA." In other words, if you choose reflexive TA, a kappa score is not a sign of quality. It is a sign you are mixing two incompatible approaches.

Topic summaries vs themes: the distinction that matters most

The most common error in published reflexive TA is presenting topic summaries as themes. A topic summary collects everything people said about a subject. A theme in reflexive TA says something about that subject.

• Topic summary: "Onboarding" (everything participants said about getting started).
• Theme: "Setup feels like proving yourself to the product" (a central idea that explains why several different onboarding complaints feel the same to users).

A quick test: if your theme name could have been written before you ran a single interview, it is probably a topic. If it could only have come from the data and your interpretation of it, it is more likely a theme.

The six phases of reflexive TA

Braun and Clarke describe six phases. They are not a strict sequence: you will loop back, especially between phases 3 and 5. Their later writing renames some phases from the 2006 version to stress that themes are generated and developed, not searched for.

Phase 1: Familiarisation

Read and re-read every transcript, and listen to audio where you have it. Make short notes on anything that strikes you: surprises, contradictions, things participants struggled to say. Do not code yet.

Product research example: after twelve churn interviews you notice that several people describe cancelling "before it became a problem", even though none of them reported a concrete problem.

Phase 2: Coding

Work systematically through the whole dataset and attach codes to segments that might matter for your research question. Codes can be semantic (close to what people said, such as "price increase") or latent (an interpretation of what lies underneath, such as "anticipating future regret"). In reflexive TA you code alone or as a team that discusses meaning, and you can revise codes as you go. The codebook is not fixed, and you do not measure agreement.

Phase 3: Generating initial themes

Look across codes for clusters that share a central organising concept. Draw candidate theme maps. Expect to have more candidate themes than you will keep.

Phase 4: Developing and reviewing themes

Check each candidate theme against the coded extracts and then against the full dataset. Ask: does this theme have a clear central idea? Is there enough rich data behind it? Does it overlap too much with another theme? Merge, split or discard as needed.

Phase 5: Refining, defining and naming themes

Write a short definition of each theme: what it is about, what it is not about, and how it relates to the other themes. Choose names that carry the central idea. A participant's own phrase often works well.

Phase 6: Writing up

In reflexive TA, writing is part of the analysis, not a report produced at the end. Combine vivid extracts with your interpretation. Each extract should illustrate a point you are making, not stand in for one.

Reflexivity in practice

Being reflexive means accounting for how your position shapes what you see. For a product team, that might include:

• Role: a PM who championed a feature will read complaints about it differently from a researcher who did not.
• Prior beliefs: write down what you expect to find before Phase 1, then revisit it before Phase 6.
• Proximity: if you know some participants as customers, note how that may colour your reading.

Keep a reflexive journal. A few lines after each analysis session is enough. Braun and Clarke's 2023 chapter "Doing reflexive thematic analysis" encourages readers to strive to be a "thoughtful, engaged, knowing researcher". In practice that means knowing why you made each analytic choice and being able to explain it.

Common mistakes reviewers flag

Braun and Clarke have reviewed published TA studies in several journals. In Health Psychology Review (2023) they examined 100 papers from five health psychology journals, most of them citing reflexive TA, and identified 10 common areas of problematic practice, with 20 recommendations for authors. They repeated the exercise for 31 papers in Health Promotion International (2024) and 20 papers in Palliative Medicine (2024), where they introduced the Reflexive Thematic Analysis Reporting Guidelines (RTARG).

The problems that recur, translated for applied research teams:

1. Topic summaries labelled as themes. Fix: name the central idea, not the subject.
2. "Themes emerged". Fix: say you developed or generated them. Themes are the product of analysis.
3. Mixing paradigms. Claiming reflexive TA while reporting coder agreement or "objective" coding. Fix: choose one approach and report it consistently.
4. Citing only the 2006 paper. The method has developed since. David Byrne's worked example in Quality & Quantity (2022) notes that "confusion persists as to how to implement this specific approach to TA appropriately", and points readers to Braun and Clarke's later writing.
5. Saturation claims. Data saturation fits poorly with reflexive TA, because meaning is generated by the researcher rather than used up by the data. Justify sample size by information power, depth and the purpose of the study.
6. Too many themes, too thin. Fix: fewer themes, each with a clear central idea and rich evidence.

When to use reflexive TA (and when not to)

Good fit:

• Exploratory discovery research where you want to understand meaning, motivation or experience.
• Interview sets small enough that one researcher or a small team can know every transcript deeply.
• Studies where the "why" matters more than the "how many".

Poor fit:

• Tracking the same categories across many studies or waves. A codebook approach gives you comparability.
• Large teams that must code thousands of responses consistently. Coding reliability or codebook TA is more defensible.
• Questions about prevalence ("what percentage of users mention X?"). Use structured questions or content analysis instead.

How Koji helps

Koji does not do reflexive TA for you, and nothing should. Reflexive TA depends on a researcher's interpretation. What Koji changes is the work around the interpretation: collecting rich data quickly and making it easy to get back into.

• Data collection without scheduling. Koji's AI interviewer runs voice or text interviews that participants take through a link whenever suits them, and it asks follow-up questions based on each answer. You set follow-up depth per question in the research brief (none, or one to three follow-ups), so you get conversational depth from more participants than you could moderate yourself.
• Familiarisation (Phase 1). Every interview produces a full transcript you can read in the app or download, and each finding in the report links back to the interview it came from, so you can always return to the context.
• A first pass, clearly labelled as one. Koji's automatic analysis summarises each interview and builds a real-time study report with themes, supporting quotes, sentiment and pain points. Treat this the way a reflexive TA researcher would treat any outside summary: as a prompt to check against the data, not as your themes. It is closer to a codebook-style topic summary, which is often exactly what helps in Phase 3 to see what you might be missing. Your own themes, definitions and reflexive journal live in your write-up.
• Structured questions for the "how many". When you need counts alongside meaning, add structured questions (rating scales, single and multiple choice, ranking, yes/no) to the same interview, so your reflexive analysis is not pushed into answering prevalence questions it was not designed for.
• Export for your own coding. Export all conversations as CSV or JSON and code them in your preferred tool or a spreadsheet. If you connect Koji to an AI assistant, you can also pull transcripts into that assistant to search them.

Compared with a manual workflow of scheduling calls, recording, transcribing and then coding, the main saving is in the first and last steps: data arrives transcribed and searchable, so more of your time goes into phases 2 to 6. Compared with legacy survey tools, the difference is depth: open-text survey boxes rarely give reflexive TA enough material, while follow-up questions do.

FAQ

What is the difference between thematic analysis and reflexive thematic analysis? Thematic analysis is a family of methods. Reflexive TA is Braun and Clarke's version, in which themes are patterns of shared meaning that the researcher actively develops, and the researcher's subjectivity is treated as a resource. Other versions, such as coding reliability and codebook TA, treat themes as topics and judge quality by consistency between coders or a fixed codebook.

Can I use inter-rater reliability in reflexive thematic analysis? Braun and Clarke advise against it. Measuring coder agreement assumes there is one correct coding to converge on, which conflicts with reflexive TA's view that coding is interpretive. Discussing codes with colleagues to deepen interpretation is fine. Reporting a kappa score is not.

How many interviews do I need for reflexive thematic analysis? There is no fixed number, and Braun and Clarke caution against claiming data saturation. Justify sample size by the richness of the data, the specificity of your question and how much interpretation the analysis requires.

What are the six phases of reflexive thematic analysis? Familiarisation, coding, generating initial themes, developing and reviewing themes, refining, defining and naming themes, and writing up. The phases are recursive: expect to move back and forth.

Can AI do reflexive thematic analysis? AI can transcribe, summarise and suggest topic groupings, which speeds up familiarisation and helps you search your data. It cannot supply the researcher's reflexive interpretation that defines the method. Treat AI-generated themes as one input to check against the data.

Related Resources

• The Complete Guide to Thematic Analysis
• Structured Questions Guide
• Qualitative Research Codebook
• Inter-Rater Reliability in Qualitative Research
• Data Saturation in Qualitative Research
• Thematic Analysis vs Content Analysis

Sources

• Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101.
• University of Auckland (2025). Psychology professor's paper is one of century's most-cited.
• Braun, V., Clarke, V., Hayfield, N., & Terry, G. (2019). Thematic analysis. In Handbook of Research Methods in Health Social Sciences. Springer.
• Braun, V., & Clarke, V. (2023). Is thematic analysis used well in health psychology? Health Psychology Review, 17(4), 695–718.
• Braun, V., & Clarke, V. (2024). A critical review of the reporting of reflexive thematic analysis in Health Promotion International. 39(3).
• Braun, V., & Clarke, V. (2024). Supporting best practice in reflexive thematic analysis reporting in Palliative Medicine.
• Byrne, D. (2022). A worked example of Braun and Clarke's approach to reflexive thematic analysis. Quality & Quantity, 56(3), 1391–1412.