At a Glance
A methods section that says "we coded the interviews thematically" leaves a reviewer unsure what was done and by what standard to judge it. Approaches to qualitative analysis rest on different ideas about what knowledge is and how a researcher relates to it. This page helps you choose one, say so, and use the checks that belong to it. A worked example is on from interview notes to findings.
At a Glance
| Approach | Best For | Codes Come From | Researcher's Role | Typical Output |
|---|---|---|---|---|
| Reflexive thematic analysis | Meaning and experience across a set of accounts | Developed during reading, revised throughout | Central and interpretive; acknowledged | Themes with a central idea each |
| Codebook or content analysis | Fixed questions, team coding, counting categories | A codebook, set in advance or developed early | Applies agreed definitions | Coded categories, sometimes counts |
| Framework analysis | Team work against set questions | Framework built by the team, often partly in advance | Organizes and compares | Matrix of cases by categories |
| Grounded theory | Explaining a process with no good existing theory | The data, through constant comparison | Builds the theory; variants differ on how | A conceptual explanation |
| Narrative analysis | How people tell and make sense of events | The structure and framing of each account | Interprets accounts kept whole | Readings of stories |
| Rapid qualitative analysis | A decision due soon, focused questions | A template set in advance | Summarizes quickly | Matrix or short report |
One Question, Six Ways
One Question Under Different Traditions
Take one fictional evaluation question: why do some adults leave an evening learning program in the first month? Each approach would work differently, and the findings would differ in kind.
- Reflexive thematic analysis: reads all the interviews, builds themes such as "falling behind and feeling exposed," and writes about them with the analyst's own position acknowledged. No fixed codebook is required.
- Codebook-based content analysis: defines codes such as timing and care in advance, trains coders, and reports how often each is applied. Consistency between coders may be examined.
- Framework analysis: builds a matrix of participants by categories, so the team can see who said what under each category and compare leavers with stayers.
- Grounded theory: follows early leads by choosing the next interviewees to test them, and produces a concept such as how falling behind and being helped to catch up shape staying.
- Narrative analysis: reads each account as a story with turning points, and compares how leavers and stayers narrate the first weeks.
- Rapid analysis: fills a template after each interview and reports the main domains in days, noting what was not examined.
None of these is the right answer. A commissioner who needs a list of barriers by next week and one who wants an explanation of the process need different things.
Matching Checks to the Tradition
Quality criteria are not interchangeable. Using the wrong one is a common cause of methods sections that contradict themselves.
| Check | Fits | Does Not Fit |
|---|---|---|
| Intercoder agreement statistic | Codebook or content analysis treated as a consistency task; some framework work | Reflexive thematic analysis, where coding is interpretive and a fixed codebook is not used |
| Saturation | Grounded theory (theoretical saturation); some interview studies with a stated meaning | Reflexive thematic analysis, where Braun and Clarke argue stopping is situated and subjective; narrative work focused on individual accounts |
| Audit trail of excerpts, codes and memos | All approaches, in different forms | None |
| Negative case analysis | Most approaches | None, though the form differs; see negative case analysis |
| Reflexive memos on position | All approaches; central to reflexive and constructivist work | None; see positionality and reflexivity |
| Counting coded passages | Content analysis with defined units | Any approach where counts would mislead about prevalence |
Saunders and colleagues show that saturation is understood in several ways across methodologies and that the logic of predicting the unobserved from the observed is uncertain. Do not borrow it as a universal rule. See qualitative saturation.
Choosing and Combining
Choosing Under Real Constraints
Ask four questions in order.
What decision does this inform, and when? A decision in two weeks narrows the choices toward a rapid or framework approach.
What is the question? A "what are the barriers?" question suits a framework or codebook approach. A "how does this happen?" question suits grounded theory. A "what does this mean to people?" question suits reflexive thematic or narrative analysis.
Who will analyze? A team that must code consistently needs shared definitions. A single analyst can work interpretively but must show reflexive practice.
What will the reviewers expect? Some commissioners ask for numbers of coded passages or agreement figures. Explain what these would and would not mean, and agree the standard in the terms of reference. See evaluation questions.
When Approaches Must Be Combined
Combining is common and workable. A team might run a rapid analysis to inform an urgent decision and a fuller thematic analysis later. Or use a framework matrix to organize the data and then write interpretive themes from it. When you combine, say which step belongs to which approach, and do not apply one approach's checks to another's output. For mixing qualitative and quantitative strands, see joint displays and conflicting results.
An AI tool does not choose a tradition for you. It will code or summarize in whatever way the prompt implies, and a prompt that supplies a codebook asks for deductive coding. Name the approach before you prompt, and trace what comes back; see checking AI-assisted qualitative coding.
Common Mistakes
Naming a tradition and using another. Calling an analysis "grounded theory" because it was inductive does not make it so.
Reporting an agreement score for reflexive analysis. The score does not match the method.
Citing saturation for any sample. Say which kind, in which approach, and how it was checked.
Letting software settle the approach. The tool's workflow is not a methodology.
Hiding the analyst. In interpretive approaches the analyst's position is part of the method. Record it.