Thematic Analysis: From Interview Extracts to Themes

Use thematic analysis to develop a defensible interpretation of qualitative data. Compare approaches and follow a worked extract-to-code-to-theme example.

Thematic analysis develops an account of patterns of meaning in qualitative data. It is a family of approaches, not a single procedure that becomes rigorous by adding labels to transcripts. State which approach you are using and keep your analysis consistent with it.

Choose the approach before choosing a coding workflow

Braun and Clarke distinguish reflexive thematic analysis from other approaches. In reflexive TA, the researcher’s interpretation is part of the analysis. A fixed codebook, independent-coder agreement and a consensus score are not interchangeable requirements for that approach.

Analytic choice	What to make explicit
Reflexive thematic analysis	Your question, interpretive position, engagement with the data and development of themes
Codebook-based team analysis	How the coding framework is developed, used and revised; how analysts handle differences
Descriptive content categorisation	The categories, units and counting rules; what those counts can and cannot mean

Do not label every spreadsheet of comments “Braun–Clarke analysis”. A practical team codebook can be useful without presenting it as a requirement of reflexive TA. Their editor and reviewer guidance identifies this methodological mismatch.

Work through six phases iteratively

The authors describe familiarisation, coding, developing initial themes, reviewing and developing themes, refining and naming them, and writing the analysis. Their guide to doing reflexive TA treats the process as recursive: interpretation can send you back to earlier material.

For a product study, keep the research question visible as you work. Record why a theme explains a meaningful pattern, which extracts support it and what challenges it. A heading such as “onboarding” names a topic; it does not yet explain what is happening.

A worked extract-to-theme example

The following extracts are invented for teaching. They are not participant quotations or client findings.

Example extract	Initial code	Analytic question
“I saved it, but I still sent a screenshot to my manager to check.”	Seeking confirmation after completion	What makes completion feel uncertain?
“The green tick tells me the file arrived, not whether somebody approved it.”	Distinguishing receipt from approval	Which system states do people need to tell apart?
“I did not need approval; I only wanted to know where the file went.”	Needing traceability rather than authorisation	Does the proposed interpretation fit different circumstances?

A possible candidate theme is completion without a clear next state creates verification work. It connects the extracts more specifically than “trust” or “feedback”. However, the third extract challenges an interpretation framed solely around approval. A real analysis would return to the full dataset and the study context before settling on a theme.

Download the analysis worksheet to keep extracts, codes, interpretive notes and contrary evidence together. Its fields support an audit trail; completing them does not validate the interpretation automatically.

Keep frequency, importance and uncertainty separate

A theme is not necessarily important because its label occurs most often. Several extracts from one participant do not equal independent accounts from several people. In a purposively recruited interview sample, counts do not estimate market prevalence.

Use a participant-by-evidence matrix if comparison across cases helps answer the question. State how the matrix relates to your analytic approach, and avoid reducing a nuanced interpretation to a vote.

Use AI assistance with traceable evidence

AI may help locate candidate passages or suggest alternative readings. Check every extract against the original, review context and record the analyst’s decisions. A generated theme name does not demonstrate that a pattern exists in the dataset.

For tool choices, see AI tools for qualitative research. For verification, use validating AI-generated insights. Neither replaces methodological judgement or the study’s consent and data arrangements.

Write findings that support a decision

For each finding, present the interpretive claim, evidence, contrary cases and scope. Separate what the study supports from the product response you propose. A theme about verification work might motivate several designs, not just an approval banner.

Use affinity mapping for organising material when appropriate, and insight to impact to make the evidence-to-decision chain inspectable. This page covers thematic analysis across qualitative product and UX research; a second general coding guide is not needed for the same task.

Glasgow Research helps B2B SaaS teams turn research into product decisions. Discuss your study.