Framework analysis (also called the framework method) is a way of organising qualitative data into a matrix: one row per case, one column per theme, and a short, referenced summary in each cell. It was developed by Ritchie and Spencer for applied policy research and is now common in health and evaluation studies. This guide covers the stages, what the matrix is for, when the method fits better than thematic analysis, and what software can and cannot do along the way.
Where the method comes from
Ritchie and Spencer described the approach in a 1994 chapter on qualitative data analysis for applied policy research. Their setting explains the design: commissioned studies, research questions fixed in advance, teams of several analysts, and a client who wants to see how conclusions were reached. The method is therefore built around a visible, inspectable structure.
Framework analysis is not tied to one epistemological position. Gale and colleagues (2013) point out that it can support inductive, deductive or combined analysis, and that it is often wrongly regarded as purely deductive because it is so systematic. It is a procedure for managing and comparing data, not a theory of what data are.
The five stages
The original description has five stages. The stages are iterative: you will go back to earlier ones when the material demands it.
- Familiarisation. You read, and ideally listen to, a good part of the material before coding anything, and note recurring ideas and your first impressions.
- Identifying a thematic framework. You draft an index of themes and subthemes. It combines issues from your research questions and interview guide with ideas that emerged during familiarisation.
- Indexing. You work through every transcript and label passages with the framework. Passages that fit nowhere are useful: they show where the framework needs revising.
- Charting. You move the indexed material into the matrix. Each cell holds a summary of what one case said about one theme, ideally close to the participant's own words and with a reference back to the transcript.
- Mapping and interpretation. You read across rows and down columns, compare cases, look for patterns, ranges and clusters, and develop explanations or typologies.
You will also see a seven-stage version by Gale and colleagues, which adds transcription at the start and splits the middle: initial coding, developing a working analytical framework, applying it, and charting. It describes the same logic in more detail, especially for teams that code the first transcripts independently and then agree on a shared framework.
What the matrix gives you
Reading across a row shows everything one participant said, so a person's account stays intact. Reading down a column shows how all participants talked about one theme. Contradictions, deviant cases and empty cells become visible, which is hard to achieve with coded segments alone. Because the matrix is a concrete artefact, a co-analyst, supervisor or commissioner can follow your reasoning from the cell back to the transcript.
Typical mistake: treating the matrix as a counting table. Samples in qualitative studies are usually purposive, so "seven of twelve mentioned this" is not a prevalence. Use the matrix to compare and explain, and be careful with numbers.
Framework analysis or thematic analysis?
They are not opposites. Framework analysis is often used to carry out a thematic analysis in a team and in a structured way. The difference lies in what the method adds: a case-by-theme matrix and an explicit, shared framework. Reflexive thematic analysis, by contrast, treats themes as something you actively construct, and does not rely on a codebook or a matrix (see reflexive vs codebook thematic analysis). If you work alone on an open question and want to stay close to interpretation, the six phases in thematic analysis after Braun and Clarke may serve you better.
Framework analysis tends to fit when the questions are set before fieldwork, when several people analyse, when you need to compare groups or sites, or when an outside party needs an audit trail. It fits less well with very heterogeneous material, or with questions that call for narrative or discourse analysis. For a small project with one analyst, a full matrix can be more overhead than benefit.
What software can and cannot do here
Indexing is the stage where software helps most, and charting and interpretation are where it helps least. In Themera every code is backed by a verbatim quote, and quotes that are not exactly in the material are rejected automatically, so each indexed passage can be traced to the text. A coverage indicator shows which responses received no code, a sign that the framework may be missing something. You can edit the codebook and re-code the material, which matches the revise-and-reapply loop of stages two and three. The results can be exported as CSV or DOCX, and a spreadsheet is a common place to build a matrix from them.
What the tool does not do is write the cells for you in a way you could cite unchecked, or interpret the matrix. Deciding what a summary should preserve, and what a pattern means, stays your task. If you run a team analysis, the intercoder agreement check (blind re-coding, Cohen's kappa) can serve as one comparison point when you agree on the framework. Pseudonymise transcripts before analysis, and ask your supervisor or data protection office what your institution requires. For the basics of building a code system, see how to code qualitative data.
Wording for your methods section
A template with placeholders, not a description of a real study. Remove whatever does not match what you did.
Data were analysed using the framework method (Ritchie & Spencer, 1994; Gale et al., 2013). After familiarisation with the transcripts, [number] researchers independently coded [number] transcripts. A working analytical framework of [number] themes was agreed, drawing on the research questions and on codes that emerged from the data, and was revised until no further codes were needed [except for an "other" category]. The framework was then applied to all transcripts. Indexed data were charted into a matrix with cases in rows and themes in columns, retaining references to the transcripts. Patterns within and across cases were interpreted in [analytic memos / team discussions]. [If applicable: An AI-assisted tool supported indexing; each code was checked against its supporting quote and revised by the researchers.]
Conclusion
Framework analysis is a structured, transparent way of comparing cases by theme, strongest in applied and team-based work. Decide early whether you need the matrix, document how your framework changed, and keep interpretation in your own hands.
See a sample analysis with codes, verbatim evidence and the coverage indicator, or try it with your own material: three analyses free, no card needed.