How to Do a Thematic Analysis

RESEARCH & DISSERTATION All Homework Solutions · expert essay, research and dissertation help

Thematic analysis is the most widely used method for analysing qualitative data in student research. It is flexible, but flexibility is not the same as anything goes.

Thematic analysis identifies patterns of meaning across a dataset — interview transcripts, focus groups, open survey responses or documents. Its accessibility makes it popular in dissertations across psychology, health, education, business and the social sciences. It is also frequently done poorly: themes that are really just topic headings, codes applied without reflection, and findings that summarise what participants said without interpreting it. This guide follows the widely used six-phase approach associated with Braun and Clarke and explains what good practice looks like at each step.

Decide what kind of thematic analysis you are doing

Thematic analysis is a family of approaches rather than a single method. Reflexive thematic analysis treats themes as patterns the researcher actively develops through interpretation. Codebook approaches use a structured framework, sometimes developed in advance. Coding reliability approaches emphasise agreement between multiple coders. Each rests on different assumptions about knowledge and the researcher’s role. Choose one deliberately, cite its key sources, and make sure your procedures match it — mixing elements inconsistently is a common criticism in examiners’ reports.

Stuck on your assignment? Our expert writers can help.

Get help →

Choose inductive or deductive coding

Inductive analysis works bottom-up, letting codes emerge from the data. Deductive analysis works top-down, using existing theory or your research questions to guide what you look for. Many studies combine both. Explain your choice in the methodology and how it relates to your research question. A study exploring an under-researched experience suits an inductive approach; a study testing how well an existing framework fits a new setting suits a deductive one.

Phase one: familiarise yourself with the data

Read and reread the entire dataset before coding. If you collected interviews, listen to recordings as well as reading transcripts, because tone and emphasis carry meaning. Make brief notes on initial impressions, striking moments and possible patterns. Familiarisation is easy to rush, but it is the foundation for everything that follows; analysts who skip it tend to code superficially.

Phase two: generate initial codes

Work systematically through the data, labelling segments that are relevant to your research question. A code is a short label capturing what is interesting about a piece of data. Codes can be semantic, describing what is said explicitly, or latent, capturing underlying assumptions and meanings.

Extract: “I didn’t want to bother the nurses, they were run off their feet.”
Semantic code: reluctance to ask for help.
Latent code: patient positioning self as burden on stretched system.

Code the whole dataset with equal attention, not just the passages that first caught your eye.

Use software or manual methods

Qualitative analysis software such as NVivo, ATLAS.ti or MAXQDA helps organise codes and retrieve extracts, especially with larger datasets. Many students analyse effectively using word processors, spreadsheets or printed transcripts with coloured pens. The software does not do the analysis; it only manages it. Choose whatever lets you work consistently and keep a clear record of your coding.

Phase three: generate initial themes

Look across your codes for broader patterns. A theme captures something important about the data in relation to the research question and represents a patterned meaning across the dataset. Group related codes into candidate themes, often using a table or a visual map. Some codes will become themes, others sub-themes, and some will be discarded as not relevant.

Themes are not topics

“Communication” is a topic. “Communication as a gatekeeper to feeling safe” is a theme — it makes an interpretive claim about what the data shows.

Phase four: review themes

Check your candidate themes at two levels. First, read all the coded extracts for each theme: do they form a coherent pattern? Second, check themes against the whole dataset: do they capture its meaning accurately in relation to your question? Expect to merge, split, rename or abandon themes at this stage. Good themes have clear boundaries and do not overlap heavily with each other.

Phase five: define and name themes

Write a short definition for each theme explaining its central idea, what it includes and what it excludes. Then give it a concise, informative name. If you cannot describe a theme’s essence in a couple of sentences, it may need further refinement. Theme names that use a participant’s phrase can be vivid, but make sure they still communicate the analytic point.

Phase six: write up the findings

Present each theme with a clear explanation, analytic commentary and well-chosen extracts from a range of participants. Your interpretation should lead and the quotes should illustrate it. Show how themes relate to one another and to the research question. A thematic map or table summarising themes and sub-themes helps readers see the overall structure at a glance.

Be reflexive throughout

In reflexive thematic analysis, your perspective is a resource rather than a bias to eliminate, but it must be acknowledged. Keep a reflexive journal noting your assumptions, reactions and analytic decisions. In your write-up, describe your position in relation to the topic and participants and how it may have shaped interpretation. This transparency is a key marker of quality.

Demonstrate rigour

Show how you ensured quality: a detailed description of each phase, an audit trail of codes and theme development, examples of coding in an appendix, and discussion of how you handled cases that did not fit. Use quality criteria appropriate to your approach — for reflexive work, these focus on depth, coherence and transparency rather than inter-rater agreement.

Common mistakes to avoid

  • Using interview questions as themes.
  • Themes that only summarise content without interpretation.
  • Quoting heavily with little analytic commentary.
  • Claiming themes “emerged” without describing the researcher’s role.
  • Mixing incompatible versions of thematic analysis.
  • Ignoring data that contradicts the main themes.