Framework analysis in qualitative research: a 5-step guide with examples

Framework analysis is a structured qualitative method developed by Jane Ritchie and Liz Spencer in 1994 for applied and policy research. It follows five steps: familiarisation, identifying a thematic framework, indexing, charting, and mapping and interpretation. The defining feature is a matrix where rows represent cases and columns represent themes, enabling systematic cross-case comparison. Tools like Skimle can automate the indexing and charting stages while preserving full transparency back to source data.

If you search for the qualitative method most used inside UK government departments, NHS evaluation teams, and public policy research institutes, you will find the same answer: framework analysis, developed by Jane Ritchie and Liz Spencer at the National Centre for Social Research (NatCen) in 1994.

While thematic analysis is the dominant method in academic research, framework analysis fills a different niche. It was designed explicitly for research with specific questions, predetermined objectives, and real-world time constraints (the conditions that define almost every applied or policy project). Its structured, matrix-based approach makes it transparent and auditable in ways that matter when findings inform public decisions.

This guide covers what framework analysis is, how to apply all five steps, a worked example from a policy context, and a clear comparison with thematic analysis so you can choose the right method for your project. If you work in the public sector, NGO space, or applied health research, this is the method to know.

What is framework analysis?

Framework analysis is a qualitative data analysis method that organises data into a systematic matrix structure, enabling structured comparison across participants and themes. It was first described in Ritchie and Spencer's 1994 chapter "Qualitative data analysis for applied policy research," published in Bryman and Burgess's edited volume Analysing Qualitative Data (Routledge).

The method was developed at NatCen to support government and public sector clients who needed rigorous qualitative findings delivered within defined timescales and to specific research questions. Its adoption spread rapidly through UK government research, the NHS, and applied social research more broadly.

Gale and colleagues (2013), in a widely cited article in BMC Medical Research Methodology, describe the framework method as "becoming an increasingly popular approach to the management and analysis of qualitative data in health research." Their paper has been cited thousands of times, reflecting the method's growing adoption in multi-disciplinary and clinical research settings.

The core output of framework analysis is a framework matrix: a spreadsheet-style structure where each row represents a participant (or case) and each column represents a theme or sub-theme. Cells contain concise, verbatim or summarised extracts from the data. This matrix makes the analytical process auditable and transparent in a way few qualitative methods can match.

What makes framework analysis different?

Three characteristics set framework analysis apart from other qualitative methods:

1. It is both deductive and inductive. The analytical framework combines codes drawn from the research questions (a priori) with codes that emerge from the data. This makes it genuinely hybrid, not purely data-driven and not purely theory-driven.

2. It is cross-sectional. The matrix structure enables comparison across participants on each theme. You can scan down a column to see how every participant discussed a particular issue, rather than only reading case-by-case.

3. It is designed for teams. Because the framework and matrix are explicit and shareable, framework analysis works well in multi-disciplinary teams where different analysts code different subsets of data. The Gale et al. (2013) paper specifically addresses this use case in health research settings.

When should you use framework analysis?

Framework analysis suits projects where several conditions apply:

• Research questions are specified in advance. The method is well-suited to studies with defined objectives rather than purely exploratory work.
• Comparison across cases is needed. The matrix structure is designed to facilitate cross-case analysis, making it valuable when you need to compare how different participants, organisations, or subgroups respond to the same issues.
• Time and resources are constrained. The structured approach is more efficient than iterative open-ended methods when analysis needs to fit within a project timeline.
• The team is multi-disciplinary. Framework analysis can involve team members who are not specialist qualitative researchers, provided an experienced qualitative analyst leads the process.
• Transparency and auditability matter. When findings will inform policy, procurement, or public decisions, the explicit matrix provides a clear evidence trail.

If you are working on stakeholder consultation analysis, public sector policy research, regulatory evaluation, or NHS programme assessment, framework analysis is likely the right choice.

What are the 5 steps of framework analysis?

Framework analysis proceeds through five steps, as described in Ritchie and Spencer (1994) and elaborated by subsequent scholars including Gale et al. (2013) and Goldsmith (2021).

Step 1: Familiarisation

Before any coding begins, the analyst must become deeply familiar with the data. This involves reading and re-reading transcripts, field notes, and documents; listening to recordings where available; and writing initial reflective notes about emerging impressions.

What you are doing: Building an overview of the breadth and depth of the dataset. Noting initial ideas, but not imposing structure yet.

Practical guidance: For a project with 20-30 interview transcripts, this stage typically takes one to three hours per transcript on first reading. Braun and Clarke's recommendation that researchers allocate more than half of total project time to data analysis applies here too. Research shared on ResearchGate suggests that manual analysis requires approximately eight hours of analyst time for every one-hour interview, across the full process from familiarisation to write-up.

In a policy context: A team evaluating a new housing benefit programme would read all transcripts from interviews with frontline workers, claimants, and local authority staff before forming any view about what themes matter.

During familiarisation, it is useful to keep notes of what strikes you as important, surprising, or recurring. These notes feed into the next step.

Step 2: Identifying a thematic framework

At this step, the analyst constructs the analytical framework: the set of themes and sub-themes that will structure the matrix. This is where framework analysis blends deductive and inductive thinking.

A priori codes come from the research questions, interview guide topics, programme theory, or existing literature. If your study asks about barriers to uptake of a health intervention, "barriers to uptake" is an a priori code.

Emergent codes come from the data itself, discovered during familiarisation. Participants may raise issues not anticipated in the interview guide. These are captured as additional codes in the framework.

What you are doing: Constructing a working index of themes and sub-themes that is grounded in both the research objectives and the data. The framework is not fixed at this stage; it is provisional and will be refined as indexing proceeds.

Practical guidance: Develop the initial framework by working through a representative subset of transcripts (typically five to eight, selected for variety). Involve the whole research team in reviewing and discussing the draft framework before finalising it.

This process is comparable to codebook development in deductive qualitative coding. The difference is that framework analysis does not require you to commit entirely to a pre-set codebook; the framework evolves as you encounter the data. For more detail on building deductive coding structures, see our guide on predefined categories in Skimle.

In a policy context: For the housing benefit evaluation, the a priori framework might include themes drawn directly from the programme logic model: eligibility assessment, application process, payment administration, and claimant experience. Emergent codes might add unexpected themes such as digital exclusion or landlord behaviour that arose frequently in interviews but were not anticipated.

Step 3: Indexing

Indexing is the application of the thematic framework to the full dataset. Every meaningful passage in every transcript is annotated with the relevant code or codes from the framework.

This is the most time-intensive step in manual framework analysis. Each document is worked through systematically, and sections are tagged with the theme numbers or labels from the analytical framework. Multiple codes can be applied to the same passage if it addresses more than one theme.

What you are doing: Systematically coding the entire dataset using the established framework. This creates the coded data that will be charted in the next step.

Practical guidance: Keep the original data intact during indexing; the index (code) is added alongside the text, not instead of it. Use a consistent notation system so that any team member can follow the coding. Some analysts annotate paper transcripts; most modern projects use qualitative software or structured spreadsheets.

For large datasets, this is the step where AI-assisted tools add the most practical value. Skimle's predefined categories feature lets you define a framework drawn from your research questions and apply it systematically across all uploaded documents, with each coded extract traceable back to the source passage. This can compress indexing from days to hours on a dataset of 30 or more interviews.

In a policy context: The housing benefit team would work through all 40 transcripts, marking every passage where a participant discusses application complexity, waiting times, payment errors, or any other theme in the framework. A passage where a claimant describes both a confusing form and a long wait would receive two index codes.

Step 4: Charting

Charting is the construction of the framework matrix. For each theme in the framework, a chart (a column in the matrix) is created. Each participant occupies a row. The analyst transfers the indexed passages into the relevant cells, usually in summarised form but with direct quotes preserved where they are analytically important.

What you are doing: Reorganising the data from a document-by-document format into a thematic, cross-case format. After charting, you can scan a single column to see every participant's contribution to a single theme.

Practical guidance: The matrix is typically built in a spreadsheet (Excel or Google Sheets). Each cell should contain a concise summary of what the participant said about that theme, along with a reference to the original transcript passage. Cells should be substantive but not over-long; aim for two to five sentences per cell.

The framework matrix for a 30-interview study with 10 themes will have 30 rows and 10 columns. This creates a structured, navigable overview of the entire dataset. The matrix is not the end product; it is the analytical working document from which interpretation proceeds.

In a policy context: The housing benefit matrix would have 40 rows (one per interviewee, including frontline staff, claimants, and local authority officers) and perhaps 12 columns covering each theme and sub-theme. A cell in the "application process" column for a frontline worker might read: "Describes the online form as having too many mandatory fields; claimants often arrive with incomplete documentation. Quote: 'We spend most of our time helping people fill in the same fields they've already filled in somewhere else.' (p.4)"

Step 5: Mapping and interpretation

The final step moves from description to interpretation. Using the completed matrix as the analytical foundation, the researcher identifies patterns, associations, and explanations across cases. This is where the analytical work of framework analysis occurs.

What you are doing: Moving from "what participants said" to "what the data means." Looking for patterns within themes, divergences between subgroups, and relationships between themes. Generating explanatory accounts that answer the research questions.

Practical guidance: Mapping and interpretation involves several distinct analytical activities:

• Range and nature: What is the range of views or experiences on each theme? What are the most common, most extreme, and most unexpected positions?
• Association: Are there relationships between themes? Does one condition appear to lead to another? Do certain subgroups cluster in particular cells?
• Exception and divergence: Who diverges from the overall pattern, and why? Exceptions are often as analytically valuable as the central tendencies.
• Conceptualisation: Constructing higher-order accounts or typologies that capture the patterns found in the data.

In a policy context: The housing benefit team would map associations between the application process theme and the claimant experience theme, perhaps finding that claimants who encountered problems during application reported more negative ongoing experiences. They might identify a typology of three frontline worker responses to system constraints: workarounds, escalation, and resignation. These interpretive findings become the substance of the evaluation report.

What does a framework matrix look like?

The table below shows a simplified example of a framework matrix from a hypothetical policy evaluation of a community mental health programme. The rows are individual participants; the columns are themes.

Reading down the "Access to services" column, a researcher can immediately compare how different participant types experience the same theme. Reading across a single row shows the full picture for one participant. Both views are analytically useful.

How does framework analysis compare with thematic analysis?

The most common question about framework analysis is how it differs from thematic analysis. They share a family resemblance (both identify themes in qualitative data), but they differ in important ways.

The comparison with reflexive thematic analysis (Braun and Clarke's updated approach) is particularly worth noting. Reflexive TA emphasises the researcher's active role in knowledge production and treats themes as interpretations rather than findings waiting to be discovered. Framework analysis sits closer to the realist end of the epistemological spectrum, which suits applied and policy contexts where the goal is to answer specific questions rather than to reflect on the analytical process itself.

For a broader comparison of qualitative analysis approaches, see our guide on content analysis vs thematic analysis and our overview of qualitative content analysis.

When to choose framework analysis over thematic analysis

Choose framework analysis when:

• Your research questions are defined in advance
• You need to compare responses systematically across participants or subgroups
• You are working in a team with members at different levels of qualitative expertise
• Your client or funder expects an auditable evidence trail
• The project involves policy-relevant topics in health, housing, welfare, education, or public services

Choose thematic analysis (or reflexive TA) when:

• Your research is exploratory and the questions will evolve as you engage with the data
• You are working alone or with a small team of experienced qualitative researchers
• The goal is interpretive depth and theoretical development rather than cross-case comparison
• You want the flexibility to revise themes substantially mid-analysis

A worked example: evaluating a public health programme

To make the five steps concrete, consider a team of four researchers evaluating a government-funded programme to reduce childhood obesity in primary schools. The evaluation commissioned 35 semi-structured interviews with headteachers, school nurses, parents, and programme coordinators.

Step 1 (Familiarisation): Each researcher reads approximately nine transcripts in full, writing brief notes about recurring topics, striking quotes, and anything unexpected. Team notes mention that headteachers frequently discuss the challenge of timetabling PE alongside curriculum pressures, something not in the original programme theory.

Step 2 (Identifying a thematic framework): The team meets to construct the initial framework. A priori themes drawn from the evaluation questions include: programme delivery, teacher and staff engagement, parent and family involvement, and measurable pupil outcomes. An emergent theme, curriculum competition, is added based on the familiarisation notes. The framework is piloted on five transcripts and refined.

Step 3 (Indexing): Each researcher indexes approximately nine transcripts using the agreed framework. Every passage about timetabling challenges is coded under curriculum competition; every passage about family meal planning under parent and family involvement. The team holds a brief weekly meeting to discuss borderline coding decisions and maintain consistency.

Step 4 (Charting): A shared matrix is built in Excel with 35 rows (one per participant) and eight columns (one per theme). Each cell is completed with a summary and the page reference to the original transcript. The "curriculum competition" column reveals that all headteachers mention timetabling constraints, but only two express them as a barrier; the others describe adaptations they have made.

Step 5 (Mapping and interpretation): The team identifies a key association between parent engagement and programme sustainability: schools where parents actively participate in programme activities show stronger maintenance of programme activities over time. The "curriculum competition" theme maps onto a typology of school responses: integrators (who embed physical activity across subjects), separators (who protect discrete PE slots), and resistors (who report the programme adding to an already pressured schedule). These findings directly address the evaluation questions and are presented in the report with cell-level evidence for each claim.

How do AI tools fit into a framework analysis workflow?

Framework analysis has always been compatible with structured, systematic tooling. The matrix itself was originally built in paper templates before researchers moved to spreadsheets. AI-assisted qualitative analysis tools represent the next iteration of the same logic: systematic management of large datasets with full traceability.

The stages where AI tools add most value in a framework analysis workflow are:

Indexing (Step 3): This is the most time-intensive manual stage. Applying a thematic framework to 30 or 40 transcripts manually may take two to four weeks for a team of three. Skimle's predefined categories feature allows you to define the themes in your analytical framework and apply them systematically across all uploaded documents. The output maps directly to the matrix structure you need for charting.

Charting (Step 4): Skimle's table view supports a matrix-style view of coded data, with each insight traceable back to the original passage. This removes the manual labour of transferring coded extracts into separate spreadsheet cells.

Familiarisation and first-pass identification (Steps 1 and 2): AI can help analysts quickly identify which topics recur across a large corpus, surfacing candidate themes for the a priori framework before detailed reading begins.

Critically, AI tools should support rather than replace the interpretive work at Step 5. Mapping and interpretation requires the analyst to make judgements about what patterns mean, why exceptions occur, and how themes relate. That remains human work.

If you are working in a public sector or policy research context, or if you are an academic researcher conducting evaluation research, Skimle's structured approach is designed for exactly this kind of framework-based analysis.

What are the strengths and limitations of framework analysis?

Strengths

Transparency and auditability: The framework matrix provides an explicit audit trail. Anyone reviewing the research can trace a finding back to the cell summaries, and from there to the original transcripts. This is particularly important for research that informs policy decisions or is subject to external scrutiny.

Suitability for team research: Framework analysis was designed for multi-disciplinary teams. The explicit framework and shared matrix mean that analysts at different levels of experience can contribute to the indexing stage, with an experienced researcher leading the interpretive stages.

Efficiency for time-constrained projects: Because the framework is established before full indexing begins, the analysis proceeds more efficiently than iterative open-ended methods. Researchers do not need to develop their analytical approach as they go.

Cross-case comparison: The matrix structure enables systematic comparison in ways that case-by-case qualitative analysis does not. This is valuable for programme evaluation, stakeholder consultation analysis, and policy research where comparison across participant types or subgroups matters.

Compatibility with mixed methods: Framework analysis sits naturally alongside quantitative components in a mixed-methods study, because its matrix output can be quantified (for example, counting the proportion of participants who mention a particular issue) without losing the qualitative depth of the original data. For more on mixed-methods approaches, see our guide on qualitative evidence synthesis.

Limitations

Risk of premature closure: Constructing the framework early in the process can constrain what the analysis finds. If the a priori codes dominate, emergent and unexpected findings may be underweighted. Mitigation: treat emergent codes as genuinely important and review the framework mid-analysis.

Matrix can oversimplify complexity: Compressing a rich participant account into a cell summary inevitably loses nuance. Cell summaries should be read alongside the full original text, not as replacements for it.

Requires methodological expertise to lead: While the method can involve team members with limited qualitative experience in the indexing stage, mapping and interpretation require an experienced qualitative researcher. Framework analysis in unskilled hands can produce descriptive summaries rather than genuine analytical insight.

Not suited to purely exploratory questions: If you genuinely do not know what you are looking for, an inductive approach such as grounded theory or interpretive phenomenological analysis may be more appropriate.

Frequently asked questions

What is framework analysis in qualitative research?

Framework analysis is a structured qualitative method developed by Jane Ritchie and Liz Spencer (1994) for applied and policy research. It organises qualitative data into a matrix where rows represent participants and columns represent themes, enabling systematic cross-case comparison. Its five steps are: familiarisation, identifying a thematic framework, indexing, charting, and mapping and interpretation.

How does framework analysis differ from thematic analysis?

Both methods identify themes in qualitative data, but they differ in structure, purpose, and setting. Framework analysis is more structured and matrix-based, designed for research with pre-specified questions and team-based projects in applied and policy settings. Thematic analysis (especially reflexive TA) is more flexible and interpretive, suited to exploratory academic research conducted by individual researchers or small teams. Framework analysis uses a hybrid deductive-inductive approach; thematic analysis more commonly starts inductively.

What is the Ritchie and Spencer framework analysis method?

The Ritchie and Spencer framework analysis method refers to the five-step approach first published in their 1994 chapter "Qualitative data analysis for applied policy research," in Bryman and Burgess's Analysing Qualitative Data (Routledge). The method was developed at the National Centre for Social Research (NatCen) for UK government and policy clients. It is characterised by its use of an analytical framework combining a priori and emergent codes, and a charting matrix that enables cross-case comparison.

Can framework analysis be used in health research?

Yes. Gale and colleagues (2013), in their widely cited article in BMC Medical Research Methodology, describe how framework analysis can be adapted for multi-disciplinary health research teams that include clinicians, patients, and lay members. The method's structured approach and explicit framework make it particularly suitable for NHS evaluation research, health services research, and clinical trial qualitative substudies.

Can I use software to support framework analysis?

Yes. While framework analysis was originally carried out with paper templates and spreadsheets, qualitative data analysis software (NVivo, MAXQDA, ATLAS.ti) and more recent AI-assisted tools can support all five stages. AI tools are particularly useful at the indexing stage, where systematic application of a framework to large datasets is time-consuming manually. Skimle's predefined categories feature is designed to support this kind of structured, framework-based analysis with full traceability back to the source data.

Ready to bring structure and speed to your framework analysis? Try Skimle for free and apply your analytical framework across all your qualitative data, with every coded extract traceable back to the original source.

Want to go deeper on qualitative methods? Read our guides on thematic analysis, how to code qualitative data, and analysing stakeholder consultation responses. For public sector teams, our public sector and policy use case page shows how Skimle fits into your existing workflow. If you are working on large-scale government consultations, see our post on AI analysis of public comments.

About the authors

Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile

Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile

Sources

• Ritchie, J., & Spencer, L. (1994). Qualitative data analysis for applied policy research. In A. Bryman & R. G. Burgess (Eds.), Analysing Qualitative Data (pp. 173-194). Routledge.
• 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, 117.
• Goldsmith, L. J. (2021). Using framework analysis in applied qualitative research. The Qualitative Report, 26(6), 2061-2076.
• Gale, N. K., Using the framework method for the analysis of qualitative data in multi-disciplinary health research. University of Birmingham research repository.
• National Centre for Social Research (NatCen). Qualitative Data Analysis.
• Wardle, H. Framework: an introduction. Harvard Qualitative Methods Project.
• ResearchGate: How much research time should one budget per one-hour interview? (community discussion)