Thematic analysis is one of the most widely used approaches to qualitative data, and also one of the most frequently done badly. Students often present a list of topics that appeared in their interviews and call them themes, without showing how they moved from raw data to interpretation. Done properly, thematic analysis is a rigorous, transparent process: you work systematically through your data, build codes, develop patterns, and arrive at themes that answer your research question. This guide teaches Braun and Clarke’s six-phase approach as it is meant to be used, with a worked example you can follow in your own project.
Braun and Clarke’s approach, set out in their influential work on thematic analysis, is deliberately flexible: it can be used with interviews, focus groups, open-ended survey responses, diaries and other textual data, and with different theoretical positions. That flexibility is a strength, but it also means you must make and record your own choices: whether your analysis is primarily inductive, driven by the data, or deductive, shaped by an existing theory; whether you code at a semantic level, close to participants’ words, or a latent level, concerned with underlying meanings; and whether you aim for a rich description of the whole dataset or a detailed account of one aspect. State those choices in your methodology, because a reader cannot judge your themes without knowing what kind of analysis produced them.
Phase 1 and 2: Familiarisation and generating initial codes
Phase 1 is familiarisation: reading and re-reading your data until you know it well. If you conducted interviews, transcribing them yourself, where feasible, is itself part of this phase, because transcription forces close listening. As you read, note initial ideas, surprising moments, contradictions and anything that seems relevant to your question. These notes are not yet codes or themes; they are the beginning of an analytic diary that will help you later, when you need to explain why a theme developed as it did.
Phase 2 is generating initial codes. A code is a short label that captures something interesting about a segment of data, relevant to your research question. Codes are the building blocks of the analysis: they are more specific than themes, and a single extract may carry several codes. Work systematically through the whole dataset, giving equal attention to each item, rather than coding the first few interviews thoroughly and skimming the rest. Keep codes close to the data at this stage. A code such as “feeling judged by lecturers” is more useful than a vague label such as “problems”, because it preserves what the participant actually said and felt.
Phase 3 and 4: Searching for and reviewing themes
Phase 3 is searching for themes: sorting codes into candidate themes and gathering the extracts that belong to each. A theme, in Braun and Clarke’s sense, is not simply a topic that recurs. It is a pattern of shared meaning, organised around a central idea, that helps answer the research question. Several codes may combine into one theme, a single code may become a theme on its own, and some codes may be set aside as miscellaneous for the time being. Visual tools help here: tables, mind maps or thematic maps that show how codes group together and how candidate themes relate to each other.
Phase 4 is reviewing themes, and it happens at two levels. First, check each candidate theme against its coded extracts: do they form a coherent pattern, or does the theme contain extracts that do not belong together? If a theme is incoherent, split it, merge it with another, or discard it. Second, check the candidate themes against the whole dataset: does the thematic map reflect the data as a whole, and have you missed anything important? This is also where you consider whether a theme is genuinely supported across the dataset or rests on a single vivid extract. A theme supported by one participant’s striking quotation may still be valid, but you must be honest about its basis and not present it as a shared pattern.
Reviewing is iterative. You will move back and forth between phases, refining codes, regrouping extracts and redrawing the thematic map. That movement is not a sign that you have done something wrong; it is the analysis working. What matters is that you can account for the decisions you made, ideally in a brief audit trail or reflexive notes, so that your supervisor or examiner can see that the themes were developed, not simply asserted.
Phase 5 and 6: Defining, naming and writing up
Phase 5 is defining and naming themes. For each theme, write a short definition: what it is about, what it includes and excludes, and what story it tells in relation to your research question. A good theme name is concise and evocative, but it must also be accurate. A name such as “Learning in the shadows” may capture a theme about hidden peer support, but only if the definition makes clear what kind of support, from whom, and with what effect. If you cannot define a theme in a paragraph, it probably needs further work, or it may be two themes rather than one.
Phase 6 is producing the report. A thematic analysis write-up does more than list themes and attach quotations. For each theme, you present the analytic story: what the pattern is, how it appears in the data, how it varies, and what it means for your research question. Quotations should illustrate and evidence the analysis, not replace it. Choose extracts that are vivid and representative, introduce them with context, and then analyse them — explaining what the participant’s words show, rather than simply repeating them in different words. A write-up that is all quotation and no analysis has not completed the method.
A worked example: from codes to themes
Imagine a study of first-year students’ experience of feedback on written assignments. Early coding produces codes such as “feedback arrives too late to use”, “comments feel generic”, “not knowing what marker wants”, “asking friends to explain feedback”, “fear of asking lecturer”, and “feedback as judgement of ability”. At first glance, these are separate observations. Grouping them, however, reveals patterns. “Feedback arrives too late”, “comments feel generic” and “not knowing what marker wants” share a concern with timing and clarity: feedback that cannot be acted on. “Asking friends” and “fear of asking lecturer” share a concern with seeking help: informal support versus formal channels. “Feedback as judgement” stands apart, concerned with identity and self-worth.
From those groupings, three candidate themes emerge. The first might be defined as “Feedback that cannot be used”: students receive comments after the relevant learning has moved on, or in language too general to guide revision. The second might be “Hidden help-seeking”: students turn to peers rather than lecturers, because peer help feels safer and faster. The third might be “Feedback and self-worth”: students read comments as judgements of their ability, not just of a single piece of work, which shapes whether they seek further feedback at all. Each theme is supported by multiple codes and extracts, each tells a distinct part of the story, and together they answer the research question about how students experience feedback.
Notice what did not happen in this example. Topics were not mistaken for themes: “feedback” itself is a topic, not a theme, because it does not state a pattern of meaning. Frequency alone did not decide importance: a code that appeared often but said little new did not automatically become a theme, while a less frequent but conceptually important pattern was retained. And the themes were checked against the whole dataset before being finalised, so that a theme reflecting only the first two interviews would have been revised. Following those disciplines in your own analysis will protect you against the most common criticisms of thematic work.
Manual workflows, NVivo and quality checks
Thematic analysis can be done manually, with printed transcripts, highlighters and tables, or with software such as NVivo. Manual analysis suits smaller datasets and students who think best on paper; it forces close reading, but it becomes unwieldy as the dataset grows, and tracking codes across many transcripts by hand is error-prone. NVivo, by contrast, helps you organise, code, retrieve and compare extracts, search across the dataset, and build visual maps of codes and themes. It does not analyse the data for you: the judgements about what counts as a code or a theme remain yours, and software cannot rescue an analysis that has not been thought through.
Whichever workflow you choose, keep the process traceable. Save versions of your code list and thematic map as they develop, note decisions and changes, and keep a reflexive diary about your own assumptions and how they may have shaped coding. Reflexivity is particularly important if you have personal experience of the topic you are studying: your familiarity may help you notice subtle meanings, but it may also lead you to hear what you expect. Acknowledging that position honestly strengthens, rather than weakens, a qualitative study.
Finally, check your analysis against common quality markers: have you coded the whole dataset, not just striking extracts? Do your themes tell a coherent story that answers the question? Are quotations analysed, not merely presented? Have you considered negative cases — data that does not fit a theme — rather than ignoring them? If you can answer yes to those questions, and show your working in the methodology and write-up, your thematic analysis will stand up well to scrutiny.
How thematic analysis help works
If you are unsure where to start, begin by clarifying your research question and your analytic approach, then code a single transcript fully before moving on to the rest. That first, careful pass will reveal whether your codes are too broad, too narrow, or drifting from the question, and it is easier to adjust after one transcript than after ten. If you are part-way through and your themes feel muddled, return to Phase 4: check each theme against its extracts, and against the dataset as a whole, before writing further.
United Academic Help supports students with qualitative analysis as part of wider dissertation and assignment help, including guidance on coding, theme development, NVivo organisation and write-up structure. Share your question, your data stage and your current code list or thematic map, and we can help you move from a collection of codes to a set of themes you can defend.
Most of the workflow above can be run inside dedicated software once your transcripts and field notes pile up. Our NVivo data analysis help covers coding, codebooks and node structures for exactly this kind of analysis, whether you want the coding done with you or checked before you write up.
Thematic analysis starts from accurate transcripts. If your data comes from interviews, our guide to transcribing interviews covers verbatim and intelligent-verbatim conventions, timestamps and anonymisation — the groundwork to complete before you start coding.
Frequently Asked Questions
How many themes should a thematic analysis have? There is no fixed number. A small undergraduate project might present three to five themes; a larger dissertation might present more, or fewer, deeper themes. What matters is that each theme is coherent, distinct and supported by the data, not that you reach a particular count.
Is thematic analysis the same as content analysis? No, although both work with textual data. Content analysis typically counts and categorises manifest content, often quantitatively. Thematic analysis, in Braun and Clarke’s approach, identifies patterns of meaning and interprets them in relation to the research question, which may include latent as well as semantic meanings.
Do I need NVivo to do thematic analysis? No. NVivo is a useful organising tool, especially for larger datasets, but thematic analysis predates it and can be done manually. Choose the workflow that fits your dataset size and your way of working, and justify the choice briefly in your methodology.
Can I use thematic analysis with a single interview? It is possible, but themes drawn from one interview cannot be presented as shared patterns across participants. Be honest about the basis of your themes, and consider whether a different analytic approach, such as narrative or interpretative phenomenological analysis, fits a single-case study better.
What is the most common mistake in thematic analysis? Presenting topics as themes, and supporting them with unanalysed quotations. A theme states a pattern of meaning; a topic merely names an area of discussion. If your theme name could be a chapter heading with no argument in it, it probably needs refining.
Thematic analysis rewards patience and honesty: code the whole dataset, build themes from patterns of meaning rather than frequency alone, review them against the data, and write them up as analysed arguments, not lists of quotations. Do that, and your analysis will show the thinking behind the findings, which is exactly what markers are looking for.