How to choose a qualitative analysis approach for interview and focus group data
Scope: interview and focus group data in applied and academic research. All example data are invented.
Short answer: choose by what your question asks of the data and what you already bring to it, not by the name of a method. If you want to interpret how people make sense of something and have little prior structure, use thematic analysis (reflexive thematic analysis if you will do the interpretive work yourself). If a client brief sets practical questions, use template analysis. If you have a theory with defined categories, use directed or deductive content analysis. If you must judge a programme against evaluation questions, use the Framework Method with explicit criteria. If a decision cannot wait, use rapid qualitative analysis and say what you left out. Whichever you pick, the method named in your report has to match what you actually did.
This guide sets out the questions that decide the choice, gives a decision table by situation, works through one invented dataset four ways, and is plain about which methods SurveysAI Qualitative Analysis Workflow does not offer.
Four questions that decide the method
Method choice in qualitative research follows from the research question, the data, your assumptions about knowledge and the claims you intend to make (Braun & Clarke, 2021a, 2022). In applied work with interviews and focus groups, four practical questions cover most of that ground.
1. What kind of claim do you need to make?
| You need to... | Kind of question | Typical approach |
|---|---|---|
| interpret how people experience or make sense of something | interpretive | thematic analysis |
| answer a client's or a team's practical questions, with room for the unexpected | applied | template analysis |
| test, refine or extend an existing theory or model | theory-led | directed or deductive content analysis |
| judge how well a programme or policy works against agreed criteria | evaluative | Framework Method with evaluation logic |
| give decision-makers a structured overview quickly | time-bound applied | rapid qualitative analysis |
| build an explanatory theory of a social process | theory-building | grounded theory |
2. What do you bring to the data? Nothing but a question; a few sensitising ideas; a brief with expected topics; a codebook with definitions and rules; or evaluation questions with criteria. The more defined the structure, the further you move from inductive towards deductive work. A list of topics is not a codebook: a codebook defines each category and says when to use it and when not (MacQueen et al., 1998).
3. What does the reader need? A comparison of groups (staff and managers, site and site), traceability from each finding to a quotation, a judgement per evaluation question, or an interpretation that stands on its own argument. Comparison and traceability favour matrix-based methods; interpretive depth favours reflexive thematic analysis.
4. How much time do you have, and how much reduction can you admit? Many rapid approaches summarise each interview or group under predefined domains instead of coding every passage (Hamilton & Finley, 2019). That is legitimate when the reduction is stated and rigour is planned for (Vindrola-Padros & Johnson, 2020; Kowalski et al., 2024). It is not legitimate to present a rapid summary as full analysis.
A fifth question sits under all four: who interprets? In reflexive thematic analysis, the researcher's own reading and reflexivity are the method (Braun & Clarke, 2019). If nobody on your side will engage with the data, a method that depends on that engagement is the wrong choice.
Decision table by situation
| Situation | Approach | Why it fits | What you must not claim | Guide |
|---|---|---|---|---|
| Open interpretive question, little prior theory, you will own the interpretation | Thematic analysis, inductive | builds patterns of shared meaning from the data | that themes "emerged" by themselves; with AI assistance, that it is reflexive TA | — |
| Client brief with set questions, open to surprises, groups to compare | Template analysis (King) | a priori themes as a starting point, revised on the data | that the template was found rather than built | — |
| Theory or model with defined categories; comparability across groups or studies | Directed or deductive qualitative content analysis | categories from theory, applied by explicit rules; uncovered text analysed | that frequency equals importance; that "data confirmed the theory" without disconfirming evidence | Directed content analysis |
| Programme or policy evaluation with evaluation questions | Framework Method plus criteria and standards | case-by-theme matrices make cross-case comparison and traceability easy | that interviews demonstrate causal impact | Framework Method evaluation |
| Decision needed soon, questions known in advance | Rapid qualitative analysis | in its common form, structured summaries by domain and a matrix | full coding depth or a segment-level audit trail | Rapid qualitative analysis |
| Theory of a social process, with further data collection possible | Grounded theory | theoretical sampling and constant comparison | grounded theory, if you analysed a closed dataset without theoretical sampling | not offered in the app (see below) |
| Lived experience in a small, homogeneous sample | Interpretative phenomenological analysis | idiographic, case-by-case interpretation | — | not offered in the app |
| Language, stories or interaction as the object of study | Narrative, discourse or conversation analysis | they analyse how talk works, not only what it says | — | not offered in the app |
Key terms
| Method | One-sentence definition | Key source |
|---|---|---|
| Reflexive thematic analysis | the researcher develops patterns of shared meaning (themes) across a dataset, treating their own subjectivity as a resource | Braun & Clarke (2021a) |
| Template analysis | a hierarchical coding template, often starting from a priori themes, is developed on part of the data and revised as the rest is coded | Brooks et al. (2015) |
| Directed qualitative content analysis | categories derived from an existing theory are applied to the data, and text that does not fit is analysed to validate or extend the theory | Hsieh & Shannon (2005) |
| Framework Method | data are indexed under a thematic framework and summarised in case-by-theme matrices for comparison across and within cases | Gale et al. (2013) |
| Rapid qualitative analysis | a family of time-saving techniques; a common form summarises each interview by domain and compares cases in a matrix | Vindrola-Padros & Johnson (2020) |
| Grounded theory | data collection and analysis alternate, guided by theoretical sampling, to build a theory of a process | Charmaz (2024) |
One dataset, four questions
Methods differ less in their data than in the question they put to it. Take an invented example, used across our guides.
Invented example. Rovina Regional Hospital (fictional) pilots a new rota on three inpatient wards: twelve-hour shifts and a shared team of nurses who move between wards as needed. After three months, a moderator runs two focus groups with the same topic guide: FG-A with eight ward nurses (N1–N8) and FG-B with six ward managers (W1–W6). Transcripts, the topic guide, the moderator's field notes and the hospital management's brief are available.
[A-12] Moderator: How did you find out about the new rota?
[A-13] N3: On the noticeboard. One day it was just up there. Nobody asked us,
and then they were surprised that we swap shifts among ourselves.
[A-14] N5: It suits me, actually. I have a small child, and twelve-hour shifts
mean more days off. If only we knew more than a week ahead.
[A-15] N1: Depends which ward you're on. On ours the swaps are kept track
of; on internal medicine they aren't.
[A-16] N3: Kept track of… covered up, more like. Someone from the shared team
comes in, doesn't know the patients, and I have to show her
everything anyway. So I'm doing two jobs.
[A-17] N6: Mhm. That's exactly it.
[B-31] W2: The nurses complain, but overtime has gone down. That's a fact.
[B-32] W4: It's gone down on the spreadsheet. When they swap a shift between
themselves, nobody records it as overtime.
[B-33] W2: But that's not the rota's fault.
The same lines answer different questions depending on the method:
| Approach | Question put to the data | What A-13 becomes |
|---|---|---|
| Template analysis | Management's brief: how is the pilot working, what helps and hinders, how do nurses and managers differ, what changes are acceptable before rollout? | a passage under "communication of the change", compared between groups |
| Reflexive thematic analysis | How do nurses experience and make sense of a change to how their work is organised? | material for a theme about being acted upon rather than consulted |
| Directed content analysis | Which job demands and job resources do participants link to the rota (codebook from the Job Demands–Resources model; Demerouti et al., 2001)? | lack of a resource (autonomy), plus uncovered text about nurses re-arranging the rota themselves |
| Framework Method in evaluation | EQ1 implementation against plan; EQ2 what helped and hindered; EQ3 observed changes; EQ4 conditions for rollout | a cell in the matrix for EQ1, "informing staff", next to a manager's row that says nothing about it |
Notice also what no method can do with these lines. B-31 and B-32 do not show whether overtime fell. They show that two managers disagree about what the overtime figures mean. That is a finding about perception and recording, not about impact.
Thematic analysis is three families, not one
Braun and Clarke's six-phase description (2006) made thematic analysis widely used, but the term covers approaches with different assumptions and quality criteria (Braun et al., 2019):
| Family | Examples | Codebook | Quality rests on |
|---|---|---|---|
| Coding reliability | Boyatzis (1998) | fixed, early | agreement between coders |
| Codebook | template analysis, Framework Method | early, working | transparency and usefulness |
| Reflexive | Braun & Clarke | none | reflexivity, depth and coherence of interpretation |
Mixing families, for example running reflexive TA with a fixed codebook and an intercoder agreement coefficient, is a practice Braun and Clarke criticise repeatedly (Braun & Clarke, 2021b). Two naming traps follow from this. "Applied thematic analysis" in the sense of Guest, MacQueen and Namey (2012) is its own approach, not the same as King's template analysis (Brooks et al., 2015). The app's "Template analysis for applied briefs" follows King and Brooks's template analysis, not Guest et al.'s method. And an AI-generated set of themes is not reflexive thematic analysis: a group of researchers including Braun and Clarke has rejected generative AI for reflexive qualitative research (Jowsey et al., 2026).
Inductive or deductive is only part of the decision
The inductive/deductive label hides separate decisions: whether your categories were specified before you engaged with the data, whether the category system may change in response to the data, and which family of analysis you work in. Interpretation is involved either way. Hybrid designs that combine a priori and data-driven codes are well established, provided the origin of each code stays visible (Fereday & Muir-Cochrane, 2006). The distinctions are worked through in Inductive vs deductive coding: what the labels decide.
The name must match what you did
The strength of a method's name should match the process you carried out. These are the claims most often overstated:
- Grounded theory requires theoretical sampling: collecting further data in response to emerging categories (Charmaz, 2024). Coding a closed set of transcripts with grounded theory techniques is at most "grounded-theory-informed".
- Reflexive thematic analysis depends on the researcher's interpretive work. If a tool proposes codes and themes, say so, and say who decided.
- Content analysis counts can be reported, but frequency is not importance (Sandelowski, 2001). Ten mentions from one talkative participant are one voice.
- The Framework Method organises and summarises data in matrices (Gale et al., 2013). An evaluative judgement needs explicit criteria and standards on top (Davidson, 2005), and interviews alone are not causal evidence.
- Rapid qualitative analysis reduces the process. The report should state what was not done (Vindrola-Padros & Johnson, 2020).
- A second reader that is an AI model is model critique, not human validation, not a second field witness and not intercoder reliability.
Focus groups change the unit of analysis
With focus groups, define early whether your unit is the speaker or the group. For claims about how views are distributed, count speakers or groups, not turns. A short agreeing turn such as A-17 ("Mhm. That's exactly it.") expresses N6's alignment with N3, but it is not a separate substantive account: record it as agreement and do not count it as a new statement. Interaction, disagreement and silence are data in their own right (Kitzinger, 1994; Kidd & Parshall, 2000). Compare groups that were run separately, such as nurses and managers, rather than pooling them as one sample. See How to analyse focus group data.
Named methods and the closest analysis in the app
SurveysAI Qualitative Analysis Workflow offers five analyses. If you are looking for a specific method, this is the closest option in the app and what to keep in mind:
| You are looking for | Closest analysis in the app | What to keep in mind |
|---|---|---|
| Grounded theory (e.g., Charmaz) | AI-assisted inductive thematic analysis | it does not build a theory and has no theoretical sampling; it is not grounded theory |
| Reflexive thematic analysis (Braun & Clarke) | AI-assisted inductive thematic analysis, option "I will decide at key points (the researcher interprets)" | the model proposes codes and candidate themes, and you decide the themes; the result is not reflexive thematic analysis |
| Template analysis (King) | Template analysis for applied briefs | follows King and Brooks's template analysis |
| Content analysis with a theory (Mayring; Hsieh & Shannon) | Deductive qualitative content analysis | the report names the method by the codebook mode you choose: "Deductive qualitative content analysis (Mayring) with a locked codebook", methodologically a locked-codebook variant informed by Mayring's deductive category application; or "Directed content analysis (Hsieh & Shannon)" when new categories are allowed |
| Framework Method in evaluation | Framework Method evaluation | interviews show perceived effects, not demonstrated ones |
| Rapid qualitative analysis | Rapid qualitative analysis | not every segment is coded; the report states this |
| IPA, narrative, discourse or conversation analysis | none | not offered |
If your study needs one of the methods in the last row, analyse it by hand or with software built for it. A thematic summary of a conversation-analytic dataset is not conversation analysis.
What the five analyses in the app do
All five start from the same required inputs: one or more interview or focus-group transcripts and an analysis brief. Optional inputs include a topic guide, field notes, case or speaker details and a starting frame. Each analysis produces quotations traceable to the turn and speaker, a report with an executive summary, and a method section that says who decided what.
| Analysis in the app | Choose it when | Extra input | Main outputs |
|---|---|---|---|
| Template analysis for applied briefs | an applied brief with concrete questions, room for the unexpected, groups to compare | — | hierarchical template (a priori and data-derived, versioned), coded dataset, answers by analytic question, group comparison, critical reading of findings |
| AI-assisted inductive thematic analysis | an interpretive question with little prior structure, and a researcher who will take over the interpretation | optional starting frame used as sensitising concepts | working codes, candidate and reviewed themes with a central organising concept, thematic map; the method section states who made the analytic decisions |
| Deductive qualitative content analysis | a theory or model with defined categories; comparability across groups | your codebook, and the choice "Only my codebook" or "My codebook plus new categories from the data" | expectations from the theory, decision rules from a pilot, coded data, residual (text the codebook does not cover), category-by-group matrix, conclusions against the theory |
| Framework Method evaluation | formative or process evaluation with evaluation questions; systematic comparison of cases | optional "Evaluation questions or criteria" and "Theory of change" | evaluation matrix with criteria and standards (supplied, or proposed and labelled as such), analytic framework, case-by-category matrices, answer and judgement per evaluation question |
| Rapid qualitative analysis | a fast result for a decision; questions known in advance | — | domains, structured summary of each group by domain, group-by-domain matrix with a synthesis; the report states that not every segment is coded |
At the start you also choose who decides at key points: "I will decide at key points" or "Leave it to the workflow". In the second case the workflow accepts its own proposals and the report says so.
What checking does and does not mean. An AI critic reads the key outputs of an analysis by the rules of its method, not every step. The critic is the same AI model that wrote them, not a human and not a second model, so it is not independent or human validation: it can miss problems and can be wrong. Quotations are checked word for word against the transcripts by program code. Points that stay unresolved are listed rather than dropped. The app does not produce agreement statistics between coders. It produces draft analytic outputs; you remain responsible for reviewing, revising and owning the interpretation.
Data handling. Transcripts are processed by external AI providers. Check that this is covered by your participants' consent and your ethics approval before you upload anything.
When none of these is the right approach
- You have one short interview. The five cross-case approaches in this guide are usually a poor fit. Use a close case-oriented analysis or a case summary, and consider an idiographic or narrative approach if that matches your question.
- Your question is causal ("did the programme reduce overtime?"). Interviews give perceived effects and the mechanisms participants describe. You need a design such as contribution analysis or realist evaluation, and usually other data (Mayne, 2012; Pawson & Tilley, 1997).
- You need agreement statistics between independent human coders. That is a coding-reliability design with trained coders. An AI model is not a second independent coder.
- Your data are mixed and unstructured (interviews, documents and observation without a shared structure). Matrix methods struggle with highly heterogeneous data (Gale et al., 2013).
- Language or interaction is the object of study. Use discourse, narrative or conversation analysis.
Choosing in practice
- Write your question in one sentence and label it: interpretive, applied, theory-led, evaluative or time-bound.
- List what you bring: nothing, sensitising ideas, a brief, a codebook with definitions, evaluation questions with criteria.
- Decide your unit of analysis (speaker, group, site) and how you will compare.
- Write down what your report will not be able to claim.
- Pick the approach from the decision table, then read its guide before you start.
Try it on your transcripts
If your situation matches one of the five analyses above, create an account with the matching analysis, then upload your transcripts and brief:
- Template analysis for applied briefs
- AI-assisted inductive thematic analysis
- Deductive qualitative content analysis
- Framework Method evaluation
- Rapid qualitative analysis
If the analysis is not preselected, choose it by name in the start dialog. Read the report as a draft that you check against the transcripts.
References
- Boyatzis, R. E. (1998). Transforming qualitative information: Thematic analysis and code development. SAGE.
- Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa
- Braun, V., & Clarke, V. (2019). Reflecting on reflexive thematic analysis. Qualitative Research in Sport, Exercise and Health, 11(4), 589–597. https://doi.org/10.1080/2159676X.2019.1628806
- Braun, V., & Clarke, V. (2021a). Thematic analysis: A practical guide. SAGE.
- Braun, V., & Clarke, V. (2021b). One size fits all? What counts as quality practice in (reflexive) thematic analysis? Qualitative Research in Psychology, 18(3), 328–352. https://doi.org/10.1080/14780887.2020.1769238
- Braun, V., & Clarke, V. (2022). Conceptual and design thinking for thematic analysis. Qualitative Psychology, 9(1), 3–26. https://doi.org/10.1037/qup0000196
- Braun, V., Clarke, V., Hayfield, N., & Terry, G. (2019). Thematic analysis. In P. Liamputtong (Ed.), Handbook of research methods in health social sciences (pp. 843–860). Springer. https://doi.org/10.1007/978-981-10-5251-4_103
- Brooks, J., McCluskey, S., Turley, E., & King, N. (2015). The utility of template analysis in qualitative psychology research. Qualitative Research in Psychology, 12(2), 202–222. https://doi.org/10.1080/14780887.2014.955224
- Charmaz, K. (2024). Constructing grounded theory (3rd ed.). SAGE.
- Davidson, E. J. (2005). Evaluation methodology basics: The nuts and bolts of sound evaluation. SAGE.
- Demerouti, E., Bakker, A. B., Nachreiner, F., & Schaufeli, W. B. (2001). The job demands–resources model of burnout. Journal of Applied Psychology, 86(3), 499–512. https://doi.org/10.1037/0021-9010.86.3.499
- Fereday, J., & Muir-Cochrane, E. (2006). Demonstrating rigor using thematic analysis: A hybrid approach of inductive and deductive coding and theme development. International Journal of Qualitative Methods, 5(1), 80–92. https://doi.org/10.1177/160940690600500107
- Gale, N. K., Heath, G., Cameron, E., Rashid, S., & Redwood, S. (2013). Using the Framework Method for the analysis of qualitative data in multi-disciplinary health research. BMC Medical Research Methodology, 13, Article 117. https://doi.org/10.1186/1471-2288-13-117
- Guest, G., MacQueen, K. M., & Namey, E. E. (2012). Applied thematic analysis. SAGE.
- Hamilton, A. B., & Finley, E. P. (2019). Qualitative methods in implementation research: An introduction. Psychiatry Research, 280, Article 112516. https://doi.org/10.1016/j.psychres.2019.112516
- Hsieh, H.-F., & Shannon, S. E. (2005). Three approaches to qualitative content analysis. Qualitative Health Research, 15(9), 1277–1288. https://doi.org/10.1177/1049732305276687
- Jowsey, T., Braun, V., Clarke, V., Lupton, D., & Fine, M. (2026). We reject the use of generative artificial intelligence for reflexive qualitative research. Qualitative Inquiry, 32(10), 931–935. https://doi.org/10.1177/10778004251401851
- Kidd, P. S., & Parshall, M. B. (2000). Getting the focus and the group: Enhancing analytical rigor in focus group research. Qualitative Health Research, 10(3), 293–308. https://doi.org/10.1177/104973200129118453
- Kitzinger, J. (1994). The methodology of focus groups: The importance of interaction between research participants. Sociology of Health & Illness, 16(1), 103–121. https://doi.org/10.1111/1467-9566.ep11347023
- Kowalski, C. P., Nevedal, A. L., Finley, E. P., Young, J. P., Lewinski, A. A., Midboe, A. M., & Hamilton, A. B. (2024). Planning for and assessing rigor in rapid qualitative analysis (PARRQA): A consensus-based framework for designing, conducting, and reporting. Implementation Science, 19, Article 71. https://doi.org/10.1186/s13012-024-01397-1
- MacQueen, K. M., McLellan, E., Kay, K., & Milstein, B. (1998). Codebook development for team-based qualitative analysis. Cultural Anthropology Methods, 10(2), 31–36. https://doi.org/10.1177/1525822X980100020301
- Mayne, J. (2012). Contribution analysis: Coming of age? Evaluation, 18(3), 270–280. https://doi.org/10.1177/1356389012451663
- Pawson, R., & Tilley, N. (1997). Realistic evaluation. SAGE.
- Sandelowski, M. (2001). Real qualitative researchers do not count: The use of numbers in qualitative research. Research in Nursing & Health, 24(3), 230–240. https://doi.org/10.1002/nur.1025
- Vindrola-Padros, C., & Johnson, G. A. (2020). Rapid techniques in qualitative research: A critical review of the literature. Qualitative Health Research, 30(10), 1596–1604. https://doi.org/10.1177/1049732320921835