Thematic analysis, step by step
How to move from transcripts to codes to defensible themes, and how to tell a real theme from a topic summary.
10 min read · Reviewed 2026 · Written and reviewed by Chris McGarrigle
Thematic analysis is the most commonly used qualitative approach at dissertation level and the most commonly done badly. The usual failure is presenting topic headings as themes.
A theme is not a subject that came up. A theme is a pattern of shared meaning that says something about your research question.
The working sequence
Widely taught accounts of thematic analysis follow broadly these stages; check the version your department teaches.
- Familiarisation: read and reread, noting first impressions.
- Initial coding: label meaningful segments systematically.
- Searching for themes: cluster codes into candidate patterns.
- Reviewing: test each candidate against the coded extracts and the full dataset.
- Defining and naming: write a sentence stating what each theme argues.
- Writing up: build the narrative with evidence.
Theme or topic?
'Workload' is a topic. 'Participants treated unmanageable workload as evidence of personal failure rather than system design' is a theme: it has a subject and a claim. If your theme name could be a chapter in a textbook, it is probably a topic.
Evidence and counter-evidence
Each theme needs enough extracts to show it is genuinely patterned, across more than one participant. Cases that do not fit should be reported, not deleted — deviant cases strengthen credibility rather than undermining it.
Link themes back to your objectives
Every theme should answer part of your research question. A theme that answers nothing you asked is interesting but belongs in your future research section.
Common mistakes
- Themes that are simply the interview questions rearranged.
- Counting occurrences and calling it analysis.
- One long quotation per theme with no interpretation.
- Silently dropping data that contradicts the pattern.
Try this: The theme sentence test
- Write each candidate theme as a full sentence with a claim.
- List the participants whose data supports it.
- Find one extract that complicates it.
- If you cannot write the sentence or find three participants, it is not yet a theme.
Do it in SuperDiss
Cluster codes into themes and check each one links to an objective.
Open the Theme FinderStudents also ask
- What are the steps of thematic analysis?
- Familiarise yourself with the data, generate initial codes, search for themes, review themes against the coded extracts and the whole dataset, define and name themes, then write up with evidence.
- What is the difference between a code and a theme?
- A code labels one interesting feature of the data. A theme is an argument that draws several codes together and says something about the research question.
- How many themes should I have?
- Typically three to six. More than six usually means codes have been promoted to themes without enough synthesis.
Related research: analyse findings
Move from raw data to themes, results and a discussion that answers the question.
How to write a results chapter
How to present dissertation findings clearly and objectively, organised by research question or theme, without slipping into discussion.
How to write a discussion chapter
The difference between results and discussion, a reliable structure, and how to make claims your data can support.
Continue your dissertation journey
- How to write a results chapter
- How to write a discussion chapter
- Reliability and validity in dissertation research
- How to build a literature matrix
- How to write a limitations section
Browse every guide in dissertation analysis.
Read next
How to write a discussion chapter
The difference between results and discussion, a reliable structure, and how to make claims your data can support.
How to build a literature matrix
A literature matrix turns scattered reading into a synthesis engine. Here is what to put in each column and how to read patterns out of it.
This guide is general educational content. Your institution's regulations, handbook and supervisor's instructions always take precedence. Where a named framework is mentioned, consult the original source text rather than relying on this summary.