Thematic analysis is popular because it is flexible: it works on interviews, focus groups, open survey responses, policy documents and forum posts, and it is not tied to a single epistemology. That same flexibility is why so many thematic analyses get criticised in vivas - without a stated approach, "themes emerged" reads as "I summarised what people said".
This guide walks through the six-phase process from Braun and Clarke's framework, shows what coding actually looks like on a real transcript extract, and covers the write-up decisions that examiners look at first.
Write up your qualitative findings faster
ThesisAI drafts structured results and discussion sections from your themes, with citations to the methodological literature pulled from real papers.
ابدأ بـ $1Decide Your Approach Before You Code
Three decisions have to be made explicitly and stated in your methodology, because they change what counts as a valid theme.
| Decision | Option A | Option B |
|---|---|---|
| Where do themes come from? | Inductive - driven by the data, codes generated bottom-up | Deductive - driven by theory, a codebook prepared in advance |
| What level do you read at? | Semantic - what participants explicitly said | Latent - underlying assumptions and ideology behind what they said |
| What is your stance on the data? | Realist - accounts report experience directly | Constructionist - accounts construct meaning in context |
Most theses land on inductive, semantic and broadly realist, which is a perfectly defensible combination. What is not defensible is not knowing which you did.
The Six Phases
Phase 1: Familiarisation
Transcribe or read every item at least twice before coding anything, taking notes on impressions rather than assigning labels. If someone else transcribed your interviews, listen to the audio while reading. Familiarisation is not a formality - the recall you build here is what lets you spot a pattern in phase three.
Phase 2: Generating initial codes
Work systematically through every item, labelling any segment relevant to your research question. Codes are short, descriptive, and numerous - 60 to 150 across a typical interview set is normal. Code generously; you can merge later, but you cannot recover a pattern you never labelled.
Transcript extract
"I check my email before I get out of bed, which is ridiculous. But if I don't, there's forty things waiting and I spend the whole morning behind. So it's not really a choice, it's just... damage control."
Codes applied
Pre-work work · Self-criticism about habit · Volume as pressure · Constrained agency ("not really a choice") · Coping framed defensively
Five codes from three sentences. Note that they are not all at the same level - "pre-work work" is descriptive, "constrained agency" is interpretive. That is fine at this stage.
Phase 3: Searching for themes
Now collate codes into candidate themes. A theme is not a topic - it is a pattern of shared meaning organised around a central idea. This is where most analyses go wrong.
Weak theme: "Email". This is a topic bucket. Everything about email goes in it, and it tells the reader nothing.
Strong theme: "Boundary work as damage control". A central organising idea - participants describe out-of-hours work not as ambition or enthusiasm but as defensive containment of a workload they cannot control. Codes about pre-work email, weekend catch-up and constrained agency all sit under it because they share that meaning.
The test: can you state the theme as a sentence that makes a claim? "Email" cannot. "Participants frame out-of-hours work as defensive rather than voluntary" can.
Phase 4: Reviewing themes
Two passes. First, read all the coded extracts for each theme - do they form a coherent pattern, or does the theme need splitting? Second, read the whole data set again - does the thematic map represent it, and did you miss anything? Themes routinely collapse, merge or split at this stage, and that is the phase working, not failing.
Phase 5: Defining and naming themes
Write a short paragraph for each theme stating what it captures, what it does not, and how it relates to your research question. If you cannot describe a theme without using the word "and" twice, it is probably two themes.
Theme definition example
Boundary work as damage control. Captures accounts in which participants describe work outside contracted hours as necessary containment of an unmanageable inbound volume, typically accompanied by self-criticism. Does not include accounts of voluntary overwork framed positively, which sit under "discretionary effort". Directly addresses RQ2 on how remote workers rationalise availability expectations.
Phase 6: Producing the report
Each theme gets a subsection: a claim, evidence, and analysis. The most common failure here is a results section that is 80 percent quotation and 20 percent linking sentences. Quotes illustrate the analytic point; they do not substitute for it.
How to Present Themes in the Results Section
- Open with the claim. State what the theme is before showing a quote, so the reader knows what to look for in it.
- Use two or three quotes per theme, from different participants. Repeatedly quoting your most articulate participant makes the theme look like one person's view.
- Analyse after every quote. A sentence explaining what this extract demonstrates and how it connects to the theme's central idea.
- Include disconfirming cases. Reporting the participants who did not fit strengthens credibility considerably.
- Be careful with counts. "Six of twelve participants" is acceptable in some traditions and criticised in others as quasi-quantification. Follow your field's convention, and never present percentages from a sample of fifteen.
Establishing Rigour
Thematic analysis attracts the criticism that it is unsystematic. These are the standard defences, and you should describe whichever you used:
- An audit trail - dated records of codebook versions and the decisions that changed them.
- Reflexivity - an explicit statement of your position relative to the participants and topic, and how it may have shaped coding.
- Second coding - a colleague codes a subset independently. Note that in reflexive thematic analysis, inter-rater reliability statistics are considered inappropriate, since the analyst's interpretation is treated as a resource rather than a source of error. In codebook or coding-reliability variants, they are expected.
- Thick description of context - enough setting detail that a reader can judge transferability to their own case.
- Member checking - returning themes to participants, where it is practical and ethically appropriate.
Common Mistakes
- Themes that are just the interview questions. If your six themes map one-to-one onto your six questions, you have organised the data, not analysed it.
- "Themes emerged." Themes are constructed by an analyst making decisions. Describe the decisions.
- Too many themes. Beyond about six, themes stop being patterns and start being categories. Consider subthemes.
- Coding only the vivid quotes. Systematic coding means every relevant segment, including the boring ones.
- Skipping phase four. The review phase is where weak themes get caught. Analyses that go straight from grouping to writing are the ones that fall apart under questioning.
FAQs About Thematic Analysis
How many interviews do I need?
There is no fixed number. For a master's thesis, 8 to 15 interviews is a common range; doctoral studies often run 20 to 40. Justify your sample by the point at which new data stopped producing new codes, and describe how you judged that.
Do I need NVivo or ATLAS.ti?
No. Software organises codes and speeds up retrieval; it does not analyse. Many good theses are coded in a spreadsheet or in a word processor with comments. Use software if your data set is large or your institution expects it.
What is the difference between thematic analysis and content analysis?
Content analysis typically counts occurrences of predefined categories and can be quantitative. Thematic analysis interprets patterns of meaning and does not depend on frequency - a theme mentioned by three participants can matter more than one mentioned by ten.
Can a theme come from a single participant?
Generally no, since a theme is a pattern across the data set. A striking single account is better reported as a deviant or negative case, which is analytically valuable in its own right.
How is it different from grounded theory?
Grounded theory is a complete methodology aimed at producing an explanatory theory, with its own sampling and analytic procedures. Thematic analysis is a method for identifying patterns, usable within many methodologies, and does not commit you to building a theory.
Done properly, thematic analysis is slow, iterative and heavily documented. That documentation is what turns a set of interesting quotes into findings a committee will accept.