Framework · Research · 3 min read
Six phases of thematic analysis
Braun and Clarke's six phases for turning interview data into themes: read everything, code everything, then build, check, name and write up patterns of shared meaning you can defend.
- 1Familiarise
Read every transcript in full before you code anything.
Transcripts, not memory - 2Code
Label every relevant segment, in two or more rounds.
"Exports the plan to a spreadsheet" - 3Initial themes
Group codes into candidate patterns of shared meaning.
"What idea ties these codes?"
- 4Review
Check each candidate against its extracts, then against the whole data set.
Split, merge or drop - 5Define, name
Write what each theme is and isn't, and name its central idea.
A name that says something - 6Write up
Tell the story of the themes, woven together with extracts.
Every claim points to a quote
Sequential but recursive: go back to any phase when the data pushes you.
Use it when
- You have five or more interview transcripts and need more than a list of what people said.
- Several people will read the same data and you want one shared, traceable way to get from quotes to themes.
- A stakeholder asks "how did you get to that?" and you want every theme to point back to the moment it came from.
How to use it
- Familiarise yourself with the data set. Read and re-read every transcript, including the sessions you ran yourself. Work from transcripts, not memory. Before you start, write down what you expect and what you hope to find. Output: notes on first observations and a short reflexivity note.
- Code the whole data set. Go a segment at a time and ask "what is this about?" Use descriptive codes close to the words and interpretive codes for your reading. Do two rounds or more, and keep a code list with a description and an example per code. Output: every relevant segment coded, with participant and timestamp.
- Generate initial themes. Look across the codes for broader patterns of shared meaning, and gather the extracts for each candidate. Output: a handful of candidate themes, each with its codes and extracts.
- Develop and review themes. Check each candidate against its extracts, then against the whole data set. Look for data that argues against it. Split, merge or drop. Output: themes that hold up, plus the contradicting data you found.
- Refine, define and name themes. For each theme, write what it is about and what it isn't, in one or two sentences. Give it a name that says something. Then take a break and come back with fresh eyes. Output: a definition and an informative name per theme.
- Write up. Tell the story of the themes together with extracts. Say which approach you used and what the team brought in. Output: a write-up where every theme points to its evidence.
Worked example
A team interviews 8 supervisors who use an internal shift-planning tool. After coding, one pile of codes is labelled "Exporting". That's a topic. Reviewing the extracts, the team sees one idea behind most of them: supervisors don't trust the plan view to show who is actually available.
| Topic summary (not a theme) | Theme |
|---|---|
| "Exporting": some export to a spreadsheet, some print, some want a PDF | Supervisors rebuild the plan elsewhere because the tool doesn't show who is really available, so publishing feels like a risk |
| "Notifications": feedback on the alert settings | Supervisors mute alerts during planning because they arrive while the plan is still wrong |
Two extracts don't fit: one supervisor exports only to share with a manager who has no access. The team reports that as a limit of the theme instead of dropping it.
Watch out for
- Coding only the memorable bits. Skimming and coding the quotes you remember gives you your memory back, not the data. Code the whole set.
- Themes named like interview questions. If the name could be a section of your discussion guide, it's a topic summary. Find the central idea.
- "Themes emerged". Say "we developed three themes". It's more honest and reminds everyone that the team shaped the reading.
- Ignoring data that contradicts a theme. Look for it in phase 4 and report it. Counting how often a word appears isn't analysis either.
Template
Use one row per coded segment in a spreadsheet; for six to ten interviews that's enough.
Reflexivity note
We expect to find:
We hope to find:
Our role in the product:
Code list
Code | Description | Example extract
Coded segments
Participant | Timestamp | Extract | Code(s)
Theme
Name (says something):
Central idea (one sentence):
What it is not:
Codes in it:
Key extracts (participant, timestamp):
Data that argues against it:
- Thematic Analysis: A Practical GuideVirginia Braun and Victoria ClarkeBook · 2021 (Amazon)
- Doing reflexive TAthematicanalysis.net (external site)
- Using thematic analysis in psychologyBraun and Clarke, 2006 (external site)
- Analyze Qualitative Data: Thematic AnalysisNN/g (external site)
"See on Amazon" links are affiliate links. As an Amazon Associate I earn from qualifying purchases, at no extra cost to you.