Framework analysis is a structured, matrix-based method for analysing qualitative data by case and by theme. Researchers become familiar with the data, build and apply an analytical framework, chart concise summaries into a case-by-theme matrix, and interpret patterns, differences, relationships, and exceptions. It is especially useful for applied, comparative, and team-based research.
Introduction
Qualitative interviews, focus groups, documents, and observational notes can produce hundreds of pages of detailed information. Researchers must reduce and organise this material without separating participants’ statements from the contexts in which they were made.
Framework analysis provides a systematic way to do this. Its defining output is usually a matrix in which rows represent cases, columns represent categories or themes, and cells contain concise, traceable summaries of the relevant data.
This guide explains:
- What framework analysis is.
- When it is appropriate.
- How its five traditional stages relate to the seven-stage Framework Method.
- How to construct and refine an analytical framework.
- How to chart data into a matrix.
- How to interpret patterns without reducing qualitative research to counting.
- How to report the method transparently.
- How software and artificial intelligence can support—but not replace—researcher judgement.
Key Takeaways
- Framework analysis organises qualitative data so researchers can examine both individual cases and themes across cases.
- Its defining feature is a matrix containing traceable summaries rather than disconnected quotations.
- The framework may be inductive, deductive, or developed through a combination of both.
- Charting reduces the volume of data but should preserve meaning, context, variation, and links to the original source.
- The matrix supports interpretation; it does not perform the interpretation automatically.
- Transparent documentation, reflexivity, and critical examination of exceptions remain essential.
What Is Framework Analysis?
Framework analysis is a qualitative data-analysis method that uses a systematically developed set of codes and categories to index data and organise it in a matrix. The matrix allows researchers to compare what different cases say about the same issue while retaining a view of each case as a whole.
The method was developed for applied social-policy research in the United Kingdom and was later adopted widely in health, education, evaluation, implementation research, public services, and other applied disciplines (Ritchie & Spencer, 1994; Gale et al., 2013).
It is sometimes described as an inherently comparative form of thematic analysis because it identifies patterns of meaning while making comparison across cases especially explicit (Goldsmith, 2021).
The defining feature: the framework matrix
A typical framework matrix contains:
- Rows: Participants, organisations, documents, sites, focus groups, or other cases.
- Columns: Codes, categories, subcategories, or analytical domains.
- Cells: Concise summaries of what a case contributes to a category.
- Source locators: Transcript page numbers, paragraph numbers, timestamps, document identifiers, or links to coded passages.
Researchers can therefore:
- Read across a row to understand one case in context.
- Read down a column to compare the same issue across cases.
- Compare groups, sites, time periods, or other case characteristics.
- Return from a summary to the original data when greater detail is required.
Important Terminology
| Term | Meaning in framework analysis |
|---|---|
| Code | A label attached to a meaningful segment of qualitative data |
| Category | A broader grouping of related codes |
| Analytical framework | An organised hierarchy of codes and categories used to index the dataset |
| Indexing | Systematically applying the analytical framework to the data |
| Charting | Summarising indexed data in the relevant cells of a framework matrix |
| Case | The unit represented by a row, such as a participant, organisation, document, site, or focus group |
| Theme | A developed pattern, proposition, or explanation produced through interpretation of categories and cases |
| Analytic memo | A written record of interpretations, questions, relationships, and decisions made during analysis |
Codes, categories, and themes should not be treated as interchangeable. Codes classify segments. Categories organise related codes. Themes are usually more interpretive and express a meaningful pattern or explanatory idea developed from the analysis.
Framework Analysis, the Framework Method and Related Terms
Is framework analysis the same as the Framework Method?
The expressions are often used interchangeably. “Framework analysis” is common in applied social and policy research, while “the Framework Method” is especially common in multidisciplinary health research.
Both refer to the same central logic:
- Develop an analytical framework.
- Apply it systematically.
- Chart summarised data into a matrix.
- Compare cases and categories.
- Interpret patterns and relationships.
Authors may divide the workflow into different numbers of stages, but this does not necessarily indicate a different method.
Framework analysis is not frame analysis
Frame analysis or framing analysis examines how language, communication, media, or social interaction constructs a particular understanding of an issue. It has different intellectual origins and research purposes.
Framework analysis, by contrast, is a structured method for organising and interpreting qualitative datasets.
Framework analysis is not framework synthesis
Framework synthesis adapts matrix-based framework principles to the synthesis of findings from multiple qualitative studies, such as within a systematic review. The unit being analysed is usually a study finding or report rather than a primary interview participant.
When Should Framework Analysis Be Used?
Framework analysis is most suitable when a study requires systematic comparison across cases while preserving the context of individual accounts.
It is particularly appropriate when:
- Research questions are focused and clearly stated.
- Interviews or documents cover broadly comparable topics.
- Researchers need to compare participants, organisations, sites, or groups.
- The project has practical, policy, service, educational, or organisational objectives.
- Some categories are informed by prior theory, policy questions, or an interview guide.
- The research must also remain open to unexpected findings.
- Multiple researchers or stakeholders are involved.
- A transparent audit trail is important.
- The qualitative component forms part of a mixed-methods study.
- Decision-makers require findings organised around clearly defined issues.
Suitable data sources
Framework analysis is used most often with semi-structured interview transcripts, but it can also be adapted to:
- Focus-group transcripts.
- Open-ended survey responses.
- Observation notes.
- Diaries.
- Policy documents.
- Meeting records.
- Case files.
- Online discussions.
- Participant-generated materials.
The data must contain enough comparable material to support a shared matrix structure.
When Is Framework Analysis Less Suitable?
Framework analysis should not be selected merely because it looks organised or resembles a spreadsheet.
Another method may be more appropriate when:
- The research focuses on the sequence and structure of an individual life story.
- The analysis requires close examination of turn-taking and interaction.
- The principal interest is how discourse constructs social realities.
- Each case concerns completely different topics, making common matrix columns artificial.
- The study aims to generate a formal explanatory theory through theoretical sampling.
- The research requires highly idiographic examination of a small number of individual experiences.
- Compressing data into categories would remove the linguistic, narrative, or interactional features that matter most.
For example, narrative analysis may be preferable for life histories, conversation analysis for recorded interactions, discourse analysis for language and power, and interpretative phenomenological analysis for detailed idiographic experience.
Five Stages or Seven Stages?
The traditional framework approach is commonly presented in five stages. Gale et al. (2013) expanded the operational procedure into seven stages.
| Five-stage formulation | Seven-stage operational formulation |
|---|---|
| 1. Familiarisation | 1. Transcription |
| 2. Identifying a thematic framework | 2. Familiarisation |
| 3. Indexing | 3. Coding |
| 4. Charting | 4. Developing a working analytical framework |
| 5. Mapping and interpretation | 5. Applying the analytical framework |
| 6. Charting data into the framework matrix | |
| 7. Interpreting the data |
The two models are compatible. The seven-stage version separates activities that are combined within the traditional five stages.
For teaching and concise reporting, the five-stage model is often sufficient. For project planning and team workflows, the seven-stage model can make responsibilities and outputs clearer.
How to Conduct Framework Analysis Step by Step
Step 1: Become Familiar With the Data
Familiarisation means reading, listening to, and reflecting on the dataset until the researcher understands its scope, diversity, context, and initial areas of significance.
Researchers may:
- Check transcripts against recordings.
- Read transcripts several times.
- Review field notes and contextual records.
- Record initial impressions.
- Note recurring concepts, contradictions, emotional expressions, and unusual cases.
- Write reflexive notes about assumptions and expectations.
- Identify issues that may need further exploration during data collection.
Familiarisation should not be treated as a quick administrative stage. Early exposure to tone, sequence, hesitation, emphasis, and context helps researchers interpret later summaries accurately.
Questions to ask during familiarisation
- What issues appear repeatedly?
- Which views differ substantially?
- What language do participants use?
- Which ideas relate directly to the research questions?
- What unexpected concerns appear?
- Which cases seem unusual?
- How might the researcher’s position influence what is being noticed?
Step 2: Develop the Analytical Framework
The analytical framework is an organised set of codes and categories used to classify the data consistently. It may draw on research questions, interview topics, theory, literature, policy priorities, and concepts emerging from the dataset.
A rigorous framework should be:
- Relevant to the research questions.
- Clear enough for consistent application.
- Flexible enough to accommodate unexpected data.
- Hierarchical where necessary.
- Supported by definitions and decision rules.
- Revised when it does not fit the data.
Deductive sources for framework categories
Categories may be informed by:
- Research objectives.
- Interview-guide sections.
- Existing theories.
- Programme logic models.
- Policy standards.
- Prior empirical literature.
- Implementation frameworks.
Inductive sources for framework categories
Categories may also be developed from:
- Participants’ recurring concerns.
- Unexpected experiences.
- Contradictions.
- In-vivo expressions.
- Relationships observed during open coding.
- Issues overlooked in the original research design.
Example codebook structure
| Code | Definition | Include | Exclude | Example indicator | Version note |
|---|---|---|---|---|---|
| Access barrier | A circumstance that makes the service difficult to obtain | Travel, cost, appointment availability, disability access | Difficulty using the service after access has been obtained | “The nearest clinic is two hours away” | Added in v1 |
| Technical usability | Ease or difficulty of operating the platform | Login, navigation, audio, video, device compatibility | General distrust not connected to operation | “I could not find the link” | Split from “technology” in v2 |
| Trust and privacy | Confidence in clinicians, technology, confidentiality, or data handling | Privacy concerns, identity verification, clinician relationship | Technical malfunction without a trust concern | “I did not know who could see the recording” | Definition refined in v3 |
A codebook may also record:
- Parent and child categories.
- Whether multiple coding is permitted.
- Typical and borderline examples.
- Decisions made during team meetings.
- Date and reason for each revision.
Test the framework before full indexing
Apply the provisional framework to a varied subset of cases. Select material that reflects diversity in participants, sites, experiences, data richness, and data-collection conditions.
Then ask:
- Are important passages uncoded?
- Do multiple codes capture the same concept?
- Are some categories too broad?
- Are distinctions between categories meaningful?
- Do definitions work across different cases?
- Does the framework force data into unsuitable categories?
- Is an “other,” “context,” or provisional category needed?
The framework is normally refined iteratively. It should not be declared final before it has been tested against the dataset.
Step 3: Index the Complete Dataset
Indexing is the systematic application of the agreed analytical framework to all relevant data.
Researchers work through each transcript or document and assign one or more codes to meaningful sections.
Good indexing requires:
- Consistent use of code definitions.
- Attention to explicit and implicit meaning where the methodology permits.
- Inclusion of contradictory and negative evidence.
- Documentation of uncertain coding decisions.
- Revision and recoding when the framework changes.
- Retention of passages that do not fit the expected pattern.
Can one passage receive several codes?
Yes. A passage may address several analytically relevant issues.
For example:
“The video appointment saved me a two-hour journey, but I was uncomfortable discussing my condition while my family could hear me.”
This passage might be indexed under:
- Reduced travel burden.
- Convenience.
- Privacy at home.
- Disclosure constraints.
Multiple coding should be purposeful rather than automatic. Excessive overlap can make the framework difficult to interpret.
Team indexing
In team projects, several researchers may independently code a shared sample and then discuss:
- Differences in interpretation.
- Code definitions.
- Missing categories.
- Boundary cases.
- The influence of disciplinary perspectives.
- Changes required before full indexing.
The objective is not necessarily mechanical agreement. The procedure should match the study’s epistemological position. Discussion can be used to deepen interpretation as well as improve consistency.
Step 4: Chart the Data Into a Framework Matrix
Charting means summarising the indexed material for each case and category in a matrix while retaining enough context and source information to recover the original meaning.
This is not the same as copying every quotation into a table.
A strong matrix cell contains:
- A concise summary written in the researcher’s words.
- Distinctions or changes within the case.
- Important conditions or context.
- A short quotation only when wording is analytically important.
- A source locator.
- A note where evidence is absent, unclear, or contradictory.
What should not happen during charting?
Avoid:
- Copying long, unedited transcript passages.
- Reducing a complex account to a single keyword.
- Removing qualifications such as “sometimes,” “only when,” or “except.”
- Treating silence as evidence of absence without checking whether the topic was discussed.
- Mixing evidence and interpretation without indicating the difference.
- Omitting negative cases because they do not fit the dominant pattern.
Hypothetical framework matrix example
Suppose a study explores how patients experience video consultations. Four illustrative interviews are charted below.
| Case | Access and convenience | Technical usability | Trust and privacy | Conditions for future use |
|---|---|---|---|---|
| P01: Older adult living alone | Avoided bus travel and waiting; convenience depended on daughter arranging appointment. [P01, paras. 14–19] | Repeated login difficulty; could not adjust audio independently. [P01, paras. 22–31] | Trusted known clinician but was unsure whether the call was recorded. [P01, paras. 39–42] | Would reuse with a simple link and telephone support. [P01, paras. 57–60] |
| P02: Rural employee | Saved several hours of travel and reduced time away from work. [P02, paras. 10–16] | Smartphone connection worked; one brief interruption did not affect consultation. [P02, paras. 24–27] | High trust because appointment was with regular clinician. [P02, paras. 34–37] | Preferred video for routine follow-up but not physical symptoms requiring examination. [P02, paras. 51–55] |
| P03: Parent of young children | Easier than arranging childcare, although home interruptions reduced concentration. [P03, paras. 8–18] | Audio delay made turn-taking difficult. [P03, paras. 25–30] | Comfortable with clinician; limited privacy because children were present. [P03, paras. 36–44] | Would reuse for short appointments scheduled around school hours. [P03, paras. 52–58] |
| P04: Person with long-term condition | Convenient for frequent monitoring and prescription review. [P04, paras. 11–17] | Platform was easy after first use. [P04, paras. 20–24] | Concerned about data storage despite trusting clinical staff. [P04, paras. 32–40] | Would reuse after receiving clear information about confidentiality and data retention. [P04, paras. 47–54] |
This is a hypothetical teaching example, not a report of an actual study.
Reading across the matrix
Reading across P01 shows that access, usability, trust, and future willingness are connected. Video care reduced travel, but dependence on a family member and unresolved recording concerns limited independent use.
Reading down the matrix
Reading down “Trust and privacy” reveals that trust in a known clinician does not automatically remove concerns about technology, household privacy, recording, or data storage.
Comparing subgroups
Researchers might compare:
- Rural and urban participants.
- First-time and experienced users.
- Younger and older participants.
- Routine and urgent consultations.
- Participants with and without digital support.
Subgroup comparisons should be planned or justified analytically. Researchers should not search indiscriminately for differences until a convenient story appears.
Step 5: Map and Interpret Patterns
Mapping and interpretation involve examining the matrix to identify patterns, variation, relationships, explanations, typologies, and exceptions that answer the research questions.
Researchers can ask:
- What experiences are shared across cases?
- Where do views differ?
- Which conditions explain those differences?
- Do certain categories occur together?
- Are there sequences or pathways across categories?
- Which cases contradict the developing explanation?
- Are apparent absences caused by interview design?
- What contextual factors shape an experience?
- How do findings relate to theory or prior evidence?
- What practical implications follow from the pattern?
From descriptive category to interpretive theme
A category such as technical usability describes an area of the data.
A more interpretive theme might be:
Digital convenience depends on invisible support work.
This theme connects several categories. Participants may describe video consultation as convenient while relying on relatives, administrators, reliable devices, private space, or technical assistance. The interpretation goes beyond listing technical problems and explains the conditions under which convenience becomes possible.
Examine negative and deviant cases
A credible analysis should not report only the most common pattern.
A negative or deviant case may:
- Contradict an emerging explanation.
- Reveal a missing condition.
- Show that a category has been defined too broadly.
- Identify an alternative pathway.
- Limit the transferability of a conclusion.
The researcher should refine the explanation rather than hide the exception.
Do not turn the matrix into survey statistics
Researchers may sometimes describe the distribution of a coded issue, especially in descriptive or mixed-methods work. However, qualitative samples are not usually designed to estimate population prevalence.
A statement such as “13 of 20 participants mentioned privacy” does not prove that 65% of the wider population holds that concern. Frequency also does not establish importance: an issue mentioned once may have major theoretical, ethical, or practical significance.
Is Framework Analysis Inductive or Deductive?
Framework analysis can be inductive, deductive, or hybrid.
Deductive framework analysis
A deductive analysis begins with categories informed by:
- A theory.
- A policy framework.
- Research questions.
- An implementation model.
- An evaluation framework.
- Existing literature.
Researchers still need to check whether these categories fit the data and whether important material falls outside them.
Inductive framework analysis
An inductive analysis develops categories primarily through open engagement with the dataset.
The matrix remains structured, but its columns are built from concepts identified during familiarisation and coding rather than imposed entirely in advance.
Hybrid framework analysis
Many applied studies use a hybrid approach:
- Broad domains come from the research questions or interview guide.
- Subcategories are developed from the data.
- Unexpected issues are added during coding.
- The framework is refined throughout indexing.
- Interpretation connects data-driven categories with prior theory.
A hybrid design is often useful when a project must address specified stakeholder questions while remaining open to participants’ unanticipated concerns.
Framework Analysis Compared With Other Qualitative Methods
| Method | Main analytical aim | Typical structure | Best suited to | Important distinction |
|---|---|---|---|---|
| Framework analysis | Compare and interpret cases and categories in a structured matrix | Cases × categories matrix | Applied studies, evaluations, team projects, focused comparative questions | Matrix charting and traceable cross-case comparison are central |
| Thematic analysis | Develop patterns of shared meaning across a dataset | Flexible coding and theme-development process | Broad range of exploratory or focused questions | Does not require a framework matrix and includes several philosophically different variants |
| Qualitative content analysis | Systematically classify and interpret textual content | Categories, coding rules, and sometimes descriptive counts | Structured description or interpretation of documents, responses, or communications | May place greater emphasis on categorisation, manifest content, or frequency, depending on the approach |
| Grounded theory | Generate an explanatory theory grounded in data | Iterative coding, constant comparison, memoing, and theoretical sampling | Processes or social phenomena requiring theory development | Theory generation and theoretical sampling are defining objectives |
| Narrative analysis | Understand how stories are structured and given meaning | Case-centred narrative interpretation | Life histories, identity, temporality, and storytelling | Sequence and narrative form may be lost in a cross-case matrix |
| Interpretative phenomenological analysis | Explore how individuals make sense of significant experience | Detailed idiographic case analysis | Small, relatively homogeneous samples and lived experience | Prioritises individual experiential depth before cross-case claims |
Framework analysis versus thematic analysis
Framework analysis and thematic analysis overlap because both can involve coding, categories, pattern identification, comparison, and theme development.
The most useful distinction is not that one is deductive and the other inductive. Either may use inductive or deductive reasoning.
Framework analysis is distinguished by its:
- Explicit analytical framework.
- Systematic application across cases.
- Matrix-based charting.
- Strong emphasis on within-case and cross-case comparison.
- Practical orientation toward clearly defined questions.
Thematic analysis is a broader family of approaches and does not require a case-by-category matrix.
Advantages of Framework Analysis
Transparent data organisation
A documented framework and matrix make it easier to show how researchers moved from raw data to findings.
Strong cross-case comparison
Researchers can compare cases systematically without losing sight of how categories relate within each case.
Suitable for multidisciplinary teams
The structure enables researchers, practitioners, service users, and other contributors to examine summarised evidence together.
Supports deductive and inductive reasoning
Predefined questions can be addressed while unexpected findings are incorporated.
Useful for large qualitative datasets
A matrix provides an overview that is difficult to obtain from transcripts or coded retrievals alone.
Relevant to applied decisions
Findings can be organised around programme components, implementation barriers, stakeholder groups, service experiences, or policy questions.
Compatible with mixed methods
Cases and categories can be connected carefully with quantitative case attributes, survey findings, or implementation measures.
Limitations of Framework Analysis
Charting can remove nuance
Poor summaries may strip away tone, ambiguity, sequence, emotion, or social context.
It is time-consuming
Coding, framework refinement, charting, checking, interpretation, and team discussion require substantial time.
Structure can create false confidence
A complete-looking matrix does not guarantee a strong interpretation.
Predefined categories can dominate
Researchers may overlook unexpected issues when they treat an initial framework as fixed.
Highly heterogeneous data may not fit
A common matrix becomes artificial when cases do not address comparable topics.
Researchers may overquantify the findings
The spreadsheet format can encourage unjustified counting or claims of statistical representativeness.
Interpretation still requires expertise
Software and matrices organise evidence. They do not determine what the evidence means.
Ensuring Quality and Rigor
Quality in framework analysis depends on more than following the five stages.
Match the method to the research question
Explain why case-by-category comparison is appropriate for the study’s objectives and data.
Maintain an audit trail
Record:
- Framework versions.
- Code definitions.
- Coding decisions.
- Reasons for merging or splitting categories.
- Meeting notes.
- Analytic memos.
- Changes to matrix structure.
- Interpretive propositions.
- Negative cases.
- Links between findings and source data.
Practise reflexivity
Researchers should examine how their:
- Professional roles.
- Disciplinary backgrounds.
- Personal experiences.
- Theoretical commitments.
- Relationships with participants.
- Organisational interests.
may influence coding and interpretation.
A reflexive journal can record assumptions, emotional responses, uncertainties, and changes in understanding.
Preserve traceability
Every substantial matrix summary should be traceable to a transcript, document, timestamp, or coded source.
Before reporting a finding, researchers should return to:
- The matrix cells.
- The associated coded passages.
- The broader case context.
- Relevant negative or contradictory evidence.
Use team discussion purposefully
Team meetings can:
- Challenge assumptions.
- Clarify categories.
- Identify alternative interpretations.
- Examine deviant cases.
- Include disciplinary and stakeholder perspectives.
- Prevent one analyst’s view from dominating unnoticed.
Numerical intercoder agreement is not a universal requirement. Its appropriateness depends on the study’s methodological and epistemological assumptions.
Report sufficient evidence
Present:
- A clear account of framework development.
- Illustrative quotations.
- Variation across cases.
- Exceptions and qualifications.
- The relationship between description and interpretation.
- Limitations of the sample and analysis.
For interview and focus-group studies, researchers should also consult reporting guidance such as COREQ.
Framework Analysis in Team-Based Research
A practical division of responsibilities may include:
| Activity | Possible responsibility |
|---|---|
| Data familiarisation | All analysts review a diverse sample |
| Initial coding | Two or more analysts code selected cases |
| Framework development | Whole analysis team |
| Framework approval | Qualitative lead with documented team input |
| Full indexing | Distributed among trained analysts |
| Charting | Analysts assigned cases or categories |
| Quality checks | Cross-review of selected matrix cells |
| Interpretation | Collaborative workshops plus individual memoing |
| Final writing | Lead author with challenge and review from the team |
An experienced qualitative researcher should oversee the design, framework logic, reflexive practice, interpretation, and reporting, particularly where some team members are new to qualitative analysis.
How to Report Framework Analysis in a Dissertation or Article
A methods section should explain more than “the data were analysed using framework analysis.”
Information to report
Include:
- Why framework analysis was selected.
- Whether the approach was inductive, deductive, or hybrid.
- The unit of analysis.
- Who conducted the analysis.
- How familiarisation was completed.
- How the initial framework was developed.
- How the framework was tested and revised.
- How indexing was performed.
- How matrices were structured.
- What each matrix cell contained.
- How source traceability was preserved.
- How cases and categories were compared.
- How reflexivity and team discussions were documented.
- Whether and how software or AI was used.
- How disagreements or alternative interpretations were handled.
- Which reporting guideline was consulted.
Example methods wording
Interview transcripts were analysed using the Framework Method. Following familiarisation, two researchers independently coded a purposively varied subset of transcripts. An initial analytical framework was developed from the research questions, interview guide, and inductively identified concepts. The framework was discussed, defined, and revised before being applied to the complete dataset. Indexed data were summarised in case-by-category matrices, with transcript locators retained for each summary. The team compared data within and across cases, examined negative cases, and developed interpretive themes through analytic memoing and discussion.
This wording must be adapted to describe what the researchers actually did.
Structuring the findings
A findings section may be organised by interpretive themes rather than simply reproducing matrix columns.
For each theme:
- State the central interpretation.
- Explain the pattern.
- Show relevant variation.
- Compare cases or groups.
- Present carefully selected quotations.
- Discuss negative or deviant cases.
- Clarify the conditions or context shaping the pattern.
The full matrix normally belongs in secure project files rather than the published article because it may contain confidential or potentially identifiable information.
How Framework Analysis Is Used in Modern Research
Framework analysis remains especially useful in research that must produce transparent and practically relevant findings.
Health and care research
Researchers use it to compare patient, caregiver, clinician, and service-manager perspectives while maintaining links between categories within each account.
Education and health-professions education
It can examine teaching practices, learner experiences, professional identity, curriculum implementation, and institutional barriers.
Policy and programme evaluation
An initial framework may reflect programme components or evaluation questions, while inductive categories reveal unintended outcomes and contextual influences.
Implementation science
Researchers can use established implementation frameworks as deductive starting points while adding locally specific barriers, mechanisms, and adaptations.
Mixed-methods research
Matrix cases can be connected with selected quantitative characteristics. For example, researchers may compare qualitative experiences across performance groups without claiming that the qualitative sample supports population estimates.
Rapid qualitative analysis
Matrix-based summaries can support time-sensitive analysis. However, shorter timelines should not remove necessary familiarisation, verification, reflexivity, or contextual interpretation.
Software for Framework Analysis
Framework analysis can be conducted manually, with spreadsheets, or through computer-assisted qualitative data-analysis software.
| Tool type | Useful functions | Important caution |
|---|---|---|
| Excel or another spreadsheet | Framework matrices, filtering, case attributes, versioned summaries | Protect confidential data, restrict access, and prevent accidental overwriting |
| NVivo | Coding, retrieval, case organisation, framework matrices, links to sources | Software structure should not dictate the analytical framework |
| ATLAS.ti | Coding, retrieval, memoing, networks, and matrix-related comparison | Visual outputs still require researcher interpretation |
| MAXQDA | Coding, summaries, case comparisons, mixed-methods displays, and exports | Automated or AI-supported functions require validation |
| QualCoder | Open-source coding, retrieval, categories, and memoing | Teams need clear procedures for file and version management |
Software does not analyse qualitative data by itself. It stores, retrieves, displays, and helps organise material according to decisions made by researchers.
Artificial Intelligence and Framework Analysis
Artificial intelligence can support selected tasks in framework analysis, but it should not be treated as an autonomous qualitative researcher.
Potentially useful AI-supported tasks
Subject to ethical approval and data-security requirements, AI may assist with:
- Draft transcription.
- Removal or detection of obvious identifiers.
- Preliminary descriptive summaries.
- Suggestions for candidate codes.
- Application of a researcher-defined codebook.
- Detection of passages that do not fit current categories.
- Comparison of summaries with source passages.
- Identification of possible negative cases.
- Formatting or restructuring a matrix.
- Searching a large dataset for relevant passages.
Tasks that require human interpretive authority
Researchers must remain responsible for:
- Deciding what counts as meaningful.
- Understanding cultural and situational context.
- Evaluating irony, ambiguity, silence, and emotion.
- Considering the researcher-participant relationship.
- Revising the analytical framework.
- Resolving contradictory interpretations.
- Developing themes and explanations.
- Assessing ethical implications.
- Determining what can legitimately be concluded.
Research comparing human and generative-AI thematic analysis has found faster automated processing but weaker coding agreement and reduced sensitivity to nuanced, interpretive themes (Prescott et al., 2024). These findings support hybrid assistance rather than unverified replacement of human analysis.
Ethical and practical safeguards
Before submitting qualitative data to an AI system:
- Check the consent form and ethics approval.
- Determine whether the tool stores or reuses submitted data.
- Follow institutional and funder policies.
- Use an approved secure environment.
- Remove direct and indirect identifiers where appropriate.
- Consider whether anonymised combinations could still identify a participant.
- Record the tool, model or version, date, prompts, and outputs.
- Verify every AI-generated code and summary against the source.
- Document rejected as well as accepted suggestions.
- Disclose material AI use in the methods section.
Confidential interview transcripts should not be uploaded to a public AI service merely because the interface is convenient.
Common Mistakes in Framework Analysis
Treating the interview guide as the final framework
Interview questions may provide useful starting domains, but participants’ answers frequently cross questions or introduce unanticipated issues.
Correction: Test and refine the framework against diverse cases.
Confusing codes with themes
A list of topics is not automatically a thematic interpretation.
Correction: Explain the pattern, relationship, condition, or meaning connecting categories.
Copying quotations instead of charting
Long extracts create an unmanageable matrix and do not demonstrate synthesis.
Correction: Write concise summaries, retain source locators, and preserve strategically important wording.
Removing context
A summary such as “prefers video” may hide that the preference applies only to routine appointments with a known clinician.
Correction: Preserve conditions, qualifications, and contradictions.
Ignoring empty cells
An empty cell may mean that the topic was not discussed, not applicable, forgotten, or genuinely absent.
Correction: Distinguish these possibilities explicitly.
Fixing the framework too early
A rigid framework can suppress unexpected data.
Correction: Allow revision and document each version.
Reporting frequencies as prevalence
A matrix makes counting easy but does not transform purposive qualitative sampling into representative measurement.
Correction: Use frequency cautiously and prioritise meaning, context, and variation.
Assuming consensus proves quality
Researchers can agree because they share the same blind spot.
Correction: Encourage alternative readings, reflexive discussion, stakeholder challenge, and negative-case analysis.
Allowing software or AI to dictate categories
Automatically generated codes may reflect surface vocabulary rather than the conceptual meaning required by the research question.
Correction: Treat automated outputs as provisional suggestions and verify them against complete cases.
Practical Framework Analysis Checklist
Before finalising the analysis, confirm that:
- The method fits the research questions.
- Cases address sufficiently comparable issues.
- The unit of analysis is explicit.
- The analytical framework has definitions and decision rules.
- Both expected and unexpected data have been considered.
- Framework revisions are documented.
- The complete relevant dataset has been indexed.
- Matrix summaries preserve context and source links.
- Comparisons have been made both within and across cases.
- Negative cases and contradictions have been examined.
- Interpretations are distinguishable from descriptive summaries.
- Reflexive notes and team decisions have been retained.
- Software and AI use has been disclosed.
- Confidential data have been protected.
- Findings are supported by matrices, source passages, and appropriate quotations.
- Limitations are stated without exaggerated claims of generalisability.
Conclusion
Framework analysis is a systematic yet adaptable method for organising and interpreting qualitative data. Its analytical value comes from combining a clearly documented framework with a traceable case-by-category matrix. Used carefully, it supports comparison, teamwork, transparency, and practical findings while retaining the context of individual accounts. Its structure does not remove the need for reflexivity, methodological judgement, and close engagement with the original data.
