Last updated: 7 October 2026
Quick Answer: Thematic analysis is a qualitative method for finding, interpreting and reporting patterns of shared meaning, called themes, across interview or text data. You do it in the six phases Braun and Clarke first set out in 2006: familiarize, code, generate themes, review, define and name, then write up. It works by hand, in a spreadsheet or in online analysis tools.
Thematic analysis is how many researchers make sense of interviews, focus groups and open-ended answers. It turns hundreds of pages of transcript into a handful of themes, each one a pattern of shared meaning backed by quotes.
Its modern form traces to one paper. Virginia Braun and Victoria Clarke's 2006 article gave the method a name, a procedure and a set of decisions to make explicitly, and most guides since build on it.
What Are the Six Phases of Thematic Analysis?
The six phases of thematic analysis are familiarizing, coding, generating initial themes, developing and reviewing themes, refining and naming themes, and writing up. Braun and Clarke have renamed several phases since 2006, so older guides use different labels for the same steps.
| Phase | 2006 name | Current name | What you do |
|---|---|---|---|
| 1 | Familiarizing yourself with your data | Familiarizing yourself with the dataset | Read and reread every transcript, noting first impressions |
| 2 | Generating initial codes | Coding | Label meaningful segments across the whole dataset, in two or more rounds |
| 3 | Searching for themes | Generating initial themes | Cluster codes into candidate patterns of shared meaning |
| 4 | Reviewing themes | Developing and reviewing themes | Test candidates against the coded data and full dataset; split, merge or drop |
| 5 | Defining and naming themes | Refining, defining and naming themes | Work out each theme's scope and story, then give it an informative name |
| 6 | Producing the report | Writing up | Weave the analytic story together with quotes and existing literature |
The phases are sequential but not rigid. On their guide to doing reflexive thematic analysis, Braun and Clarke describe analysis as recursive, moving back and forth, and note that experienced analysts with small datasets may blur phases together.
What Did Braun and Clarke Publish in 2006?
In 2006, Braun and Clarke published "Using thematic analysis in psychology" in Qualitative Research in Psychology, volume 3, issue 2, pages 77 to 101. It described thematic analysis as widely used but poorly defined, and offered a clear, theoretically flexible procedure.
That paper became one of the most cited in science. Nature's 2025 feature on the most-cited papers of the twenty-first century includes it, and the authors report that it ranked third.
Their thinking has moved on since. They now call their approach reflexive thematic analysis and point readers to recent work, including their 2022 textbook, Thematic Analysis: A Practical Guide.
What Is Reflexive Thematic Analysis?
Reflexive thematic analysis is Braun and Clarke's own version of the method, in which the researcher's interpretation is treated as a resource rather than a bias to control. Themes are patterns of shared meaning built around a central organizing concept, developed late in analysis.
It is one of three broad families, which Braun and Clarke set out in their overview of understanding thematic analysis.
| Approach | How coding works | What a theme is |
|---|---|---|
| Coding reliability | Fixed codebook, several independent coders, agreement scores | Often a topic summary set early |
| Codebook (template, framework, matrix) | Structured codebook, no agreement scores | Often a topic summary charted in a grid |
| Reflexive | Open coding that evolves as understanding deepens | A pattern of shared meaning, developed from codes |
Picking a family matters because each implies different quality checks. Agreement scores make sense in coding reliability work and make no sense in reflexive work.
What Is Inductive Thematic Analysis?
Inductive thematic analysis lets the content of the data direct coding and theme development, while deductive analysis is directed by existing concepts or theory. Most projects lean one way rather than sitting at either extreme.
A second choice runs alongside it. Semantic coding stays close to what people explicitly said, while latent coding looks at the assumptions underneath. Braun and Clarke treat both as continua, and note that inductive, semantic work often clusters together, as does deductive, latent work.
A brand team asking why customers left would usually start inductively. A team testing whether an existing model, such as a brand equity framework, explains its interviews would work deductively.
What Is an Example of Thematic Analysis?
A well-documented thematic analysis example is David Byrne's worked example of reflexive thematic analysis in Quality & Quantity, built on eleven semi-structured interviews of about 25 to 30 minutes.
Byrne studied educators' attitudes to wellbeing education. Early codes such as "positivity regarding the wellbeing curriculum" were later split into student and educator versions, and one theme, "the influence of time", went beyond describing time pressure as a barrier.
A market research version, for illustration, might analyze 30 interviews with people who canceled a meal-kit subscription:
- Code: Label segments such as "cooking felt like a chore" or "checked my bank app before renewing".
- Cluster: Group codes that share a meaning, not just a topic.
- Name: Replace a topic label like "Reasons for canceling" with a claim like "Canceling to regain control of spending".
How Is Thematic Analysis Different From Content Analysis?
Thematic analysis interprets what patterns in the data mean, while content analysis often describes and counts what appears, though the two terms have long overlapped.
Braun and Clarke note that the terms were historically used interchangeably, and that content analysis has most in common with the structured coding reliability and codebook forms of thematic analysis. Reflexive thematic analysis is the most distinct from it.
Use content analysis when the question is about frequency, such as how many respondents mention price. Use thematic analysis when the question is about meaning, such as what price stands for in people's decisions.
What Are Common Mistakes in Thematic Analysis?
In reflexive thematic analysis, the most common mistake is presenting topic summaries as themes, Braun and Clarke wrote in a 2023 commentary on good practice. Other errors follow from it.
- Topic summaries: A theme named "Experiences of X" or "Barriers to Y" lists what people said about a topic without a unifying idea.
- Themes that mirror the guide: If a theme maps closely onto a question you asked, or could have been written before analysis, it is probably a topic summary.
- Positivism creep: Reaching for researcher bias or agreement scores inside an approach that rejects both.
- Mixing approaches unknowingly: Combining codebook procedures with reflexive claims without saying why.
- Not owning your perspective: Reflexive analysis expects you to state how your position shaped the reading.
When Should You Use Thematic Analysis in Qualitative Research?
Use thematic analysis in qualitative research when you want patterns of meaning across a dataset, such as semi-structured interviews, focus groups or open-ended survey answers.
Braun and Clarke recommend it over interpretative phenomenological analysis (IPA) for larger, more diverse datasets and for data that is not first-person accounts of experience. It sits alongside other methods in the wider qualitative research toolkit.
Another approach is the better choice in three cases:
- Small experiential studies: IPA suits a small group where each individual account matters.
- Building a theory: Grounded theory, with theoretical sampling, aims to produce a theory, which thematic analysis does not.
- Counting mentions: Content analysis suits questions about how often something is said.
Can AI Do Thematic Analysis?
AI can assist thematic analysis, especially when applying a defined codebook, but a 2026 blinded comparison found its output still needed human verification.
In that study in PLOS Digital Health, large language models matched blinded human coders when applying a 10-code codebook to a focus-group transcript, but were variable at open-ended theme generation. The authors concluded the models can augment analysis with human checking.
On AI-moderated interviews, Alchemic codes themes while fieldwork runs, links each finding to the respondent, verbatim or voice clip behind it, and has expert researchers oversee the analysis. Tools for open-ended answers are compared in AI open-end coding software, and desktop options in qualitative data analysis software.
Analysis can only find patterns in the voices you reached. WhatsApp-native interviews run as text inside a chat, with voice notes supported, and Alchemic publishes 60+ languages including Hindi and Spanish.
For a thesis using reflexive thematic analysis, code by hand or in desktop software, because the coding is how you come to know the data. In applied work, the bigger risk sits upstream, in guide questions that invite topic summaries. Once it has the client's brief, Alchemic designs and tailors the discussion guide to handle these risks before fielding, rather than leaving that work to the buyer.