The Complete Guide to Qualitative Data Analysis

Qualitative data analysis is the disciplined process of turning interviews, observations, documents, images and other non-numerical evidence into defensible interpretations. It is not simply highlighting interesting quotations or arranging comments under headings. A strong analysis demonstrates how the researcher moved from raw material to codes, categories, themes, narratives, explanations or theoretical claims.

Packages 1–3 established the analytical foundation and covered thematic, framework, content, grounded theory, IPA and narrative analysis. Package 4 completes the guide's overview of major qualitative analysis methods by adding discourse analysis, conversation analysis, template analysis, document analysis, case study analysis, ethnographic analysis and qualitative comparative analysis (QCA).

1. What Is Qualitative Data Analysis?

Qualitative data analysis examines meaning, experience, process, language, interaction, identity, context and social practice. The aim is to produce an interpretation that answers the research question while remaining grounded in the evidence.

The material analysed may include:

• interview and focus-group transcripts;
• field notes and observational records;
• diaries, letters and reflective journals;
• policy documents, reports and archival records;
• photographs, videos and visual artefacts;
• websites, social-media discussions and online communities;
• meeting minutes, organisational communications and case files;
• researcher memos and analytical notes.

The central task is not to reduce this material mechanically. The researcher must identify what matters, compare patterns, interpret relationships, examine contradictions, connect evidence to context and develop claims of an appropriate scope.

1.1 Description, analysis and interpretation

These three levels are often confused.

A thesis may contain accurate description but still have weak analysis. For example, reporting that “participants expressed concern about monitoring” merely restates the interviews. Stronger analysis may show that monitoring anxiety differed by professional status, was intensified where implementation lacked consultation, and altered how employees interpreted organisational trust.

1.2 Analysis begins before formal coding

Researchers often imagine analysis beginning when transcripts are imported into NVivo or another software package. In practice, analysis starts earlier. Decisions about what to observe, whom to interview, which questions to ask, what to transcribe and how to record context already shape the evidence available for interpretation.

Analytical thinking should therefore accompany data collection. Researchers can write field notes, record emerging ideas, identify gaps and refine later interviews without prematurely fixing the findings.

2. How Qualitative Analysis Differs from Quantitative Analysis

Qualitative and quantitative analysis both require systematic reasoning, but they answer different kinds of questions and use different forms of evidence.

2.1 Qualitative analysis is not unstructured

Flexibility does not mean absence of method. A defensible analysis explains how evidence was selected, coded, compared and interpreted. It also shows how alternative explanations were considered and how the researcher's assumptions were managed.

2.2 Numbers may appear in qualitative analysis

Researchers may count participants, documents or code occurrences to support orientation. However, frequency alone rarely establishes importance. A theme mentioned by only a few participants may still be analytically significant if it reveals a hidden mechanism, a negative case or the experience of a marginalised group.

3. The Overall Qualitative Analysis Workflow

Different methods use different procedures, but most qualitative analyses involve the following broad movement:

3.1 Stage 1: Prepare the evidence

Collect files, transcribe recordings, anonymise identifying details, preserve contextual information and create a secure, consistent structure. Poor preparation creates errors that later analysis cannot fully correct.

3.2 Stage 2: Familiarise yourself with the dataset

Read, listen and review the material repeatedly. Record early impressions, contradictions, emotional tone, contextual details and possible relationships. Familiarisation prevents the researcher from coding isolated sentences without understanding the whole account.

3.3 Stage 3: Code systematically

Coding labels segments of data that are relevant to the research question. Codes may describe content, identify actions, capture participant language, represent theoretical concepts or mark contradictions. Coding should be systematic but need not be rigidly uniform across every method.

3.4 Stage 4: Develop categories, themes or other analytical structures

Codes are compared and grouped into broader patterns. Depending on the method, these may become themes, categories, narrative structures, discourse repertoires, case explanations, process models or theoretical concepts.

3.5 Stage 5: Interpret the patterns

The researcher asks what the patterns mean, why they occur, how context shapes them, which mechanisms may explain them, and how they relate to theory and previous research.

3.6 Stage 6: Test and refine the analysis

Strong analyses examine negative cases, compare groups or cases, revisit earlier transcripts, seek alternative explanations and check that claims are supported by sufficient evidence.

3.7 Stage 7: Write the findings

Writing is part of analysis rather than a final administrative step. The act of explaining a theme often reveals weak boundaries, unsupported claims or missing relationships. Good qualitative writing integrates evidence and interpretation instead of presenting long quotations followed by minimal commentary.

4. Preparing Qualitative Data for Analysis

Data preparation creates the foundation for credible analysis. Researchers should preserve both the content of the evidence and the context needed to interpret it.

4.1 Create a data inventory

A data inventory records every source included in the study. It may contain:

• a unique source identifier;
• the source type;
• date and location;
• participant or case characteristics;
• consent and anonymisation status;
• transcription status;
• analytical notes;
• storage location and version.

This prevents missing, duplicated or incorrectly labelled material and supports the audit trail.

4.2 Protect the original data

Researchers should preserve untouched master copies and conduct cleaning or formatting on working copies. Audio, images and documents should be stored securely, with access restricted according to ethical approval and data-protection requirements.

4.3 Anonymise without destroying meaning

Names and identifying details may need to be replaced, but over-anonymisation can remove analytically important context. For example, replacing every job role with “employee” may obscure differences between managers, technical specialists and frontline staff.

A practical approach is to retain relevant analytical attributes while removing details that could identify individuals or organisations.

4.4 Preserve contextual metadata

A transcript records speech, but not automatically the setting, interruptions, participant role, interactional atmosphere or events that occurred before and after the interview. Researchers should link transcripts to field notes and relevant metadata.

4.5 Decide what belongs in the analytical corpus

Not every collected item must be analysed in the same way. Researchers should distinguish between:

• primary data directly analysed to answer the research question;
• contextual material used to understand the setting;
• procedural records documenting recruitment or fieldwork;
• analytical records such as memos, coding decisions and diagrams.

5. Transcription and Data Quality

Transcription converts spoken interaction into a written or otherwise analysable record. It is not a neutral clerical process. Decisions about pauses, emphasis, overlap, laughter, incomplete sentences and non-verbal behaviour affect what can be analysed.

5.1 Select the level of transcription required

5.2 Automated transcription

Automated tools can accelerate transcription, but their output must be checked. Errors often occur with accents, specialist terminology, multiple speakers, poor audio and code-switching. An inaccurate transcript can distort the meaning of the evidence.

5.3 Transcription as familiarisation

When researchers transcribe their own interviews, they begin noticing patterns, tone and contradictions. When transcription is outsourced, researchers should still listen to recordings while checking transcripts and write familiarisation notes.

5.4 Translation and multilingual data

Translation adds another interpretive layer. Researchers should explain who translated the material, when translation occurred, how culturally specific terms were handled and whether quotations were checked by bilingual speakers. Translating before analysis may simplify coding but may also lose linguistic nuance. Analysing in the original language and translating selected quotations can preserve meaning but requires appropriate language competence.

6. Organising the Dataset

Good organisation reduces the risk of analytical errors and makes the project easier to audit. A consistent file structure is more important than sophisticated software.

6.1 Naming conventions

File names should be consistent and non-identifying. For example:

INT_P07_2026-06-14_Transcript_v2.docx

This may indicate interview, participant code, date, document type and version. Avoid informal names such as “final transcript latest corrected.”

6.2 Version control

Researchers should know which transcript, codebook or memo is current. Major analytical changes should be documented rather than overwritten silently. This is especially important in team projects.

6.3 Case and attribute tables

A case table summarises relevant participant or document characteristics. It allows researchers to compare themes across roles, sites, periods or demographic dimensions without repeatedly searching individual files.

6.4 Data security and backups

Maintain secure backups, separate identifying information from analytical files where possible, and follow the retention period stated in the ethics application. Cloud storage should comply with institutional requirements.

7. Familiarisation with the Data

Familiarisation means developing a broad and detailed understanding of the dataset before reducing it into codes. It is more than reading each transcript once.

7.1 A practical familiarisation process

1. Read or listen to each source as a whole.
2. Write a short summary of the participant, case or document.
3. Record striking ideas, tensions and contradictions.
4. Note the relationship to the research questions.
5. Identify contextual details that may shape meaning.
6. Record early comparisons with other sources.
7. Separate observation from interpretation in notes.

7.2 Source summaries

A one-page source summary may include:

• the participant's position and relevant background;
• the central story or argument;
• important events or experiences;
• strong or unusual statements;
• contradictions within the account;
• possible relationships to other cases;
• questions for later analysis.

7.3 Do not code too early

Immediate line-by-line coding can fragment the data before the researcher understands the overall account. Some methods require close early coding, but even then researchers should retain a sense of the whole transcript or case.

7.4 Write initial analytical memos

Memos preserve developing ideas that would otherwise be forgotten. An initial memo might ask why junior employees described monitoring as personal distrust while managers described it as organisational accountability. The memo does not need to be correct. It creates a question that can be tested against the data.

8. Reflexivity Before and During Analysis

Reflexivity is the systematic examination of how the researcher's identity, assumptions, theoretical commitments, relationships and decisions shape the research. It does not require pretending to remove all influence. It requires making that influence visible and analytically responsible.

8.1 Positionality

Researchers should consider how their professional role, social position, insider or outsider status, and relationship to participants may affect access, disclosure and interpretation. A manager interviewing employees may receive different accounts from an independent researcher. An insider may understand organisational language but take routine practices for granted.

8.2 Prior assumptions

Before coding begins, researchers can write a pre-analysis memo identifying:

• what they expect to find;
• which theoretical ideas they favour;
• which participants they may find more credible;
• which outcomes they hope the study will support;
• which personal experiences shape their interest.

This memo does not eliminate bias. It makes assumptions available for later scrutiny.

8.3 Reflexivity during interpretation

Researchers should ask:

• Why am I noticing this pattern?
• What evidence would challenge my interpretation?
• Am I treating articulate participants as more truthful?
• Have I ignored inconvenient or minority accounts?
• Am I imposing theoretical language participants did not use?
• How did my interview style shape the data?

8.4 Team reflexivity

Multiple coders do not automatically remove subjectivity. Team members may share the same assumptions. Productive team analysis involves discussing differences, tracing why interpretations diverge and refining the codebook or conceptual model.

9. Choosing the Right Qualitative Analysis Method

The analysis method should be chosen according to the research question, methodology, philosophical position, type of data and intended output. Researchers should not select thematic analysis merely because it is familiar or because software tutorials are widely available.

9.1 Start with the intended analytical product

9.2 Check compatibility with methodology

A methodology and an analysis method are related but not identical. A case study may use thematic, framework or narrative analysis, but the analysis must still preserve the case boundaries and contextual explanation. A grounded theory study cannot simply use generic thematic analysis and claim to have generated theory.

9.3 Consider the unit of analysis

The unit may be:

• a participant's whole account;
• a segment of text;
• a story or episode;
• an interactional turn;
• a document;
• a case or organisation;
• a cultural practice;
• a discourse or interpretive repertoire.

Methods that fragment data into small segments may be unsuitable when the whole narrative or interactional sequence is central.

9.4 Consider whether the approach is inductive, deductive or abductive

An inductive analysis develops patterns primarily from the data. A deductive analysis applies concepts from theory or a framework. An abductive analysis moves between data and theory to develop the most plausible explanation. Many doctoral studies use a combination, but the researcher should explain which stage used which logic.

9.5 Consider the expected audience

Applied research for policy or services may benefit from framework analysis because it supports transparent comparison and practical recommendations. A theoretical sociology thesis may require deeper discourse, narrative or grounded theory analysis. Audience should not override methodological fit, but it influences how findings need to be organised.

10. A Decision Framework for Selecting an Analysis Method

1. What is the research question asking? Meaning, process, story, language, interaction, culture, case explanation or practical comparison?
2. What is the methodology? Case study, phenomenology, grounded theory, ethnography, narrative inquiry, qualitative description or another design?
3. What is the unit of analysis? Segments, whole accounts, interactions, cases, documents or practices?
4. What form should the output take? Themes, categories, theory, narrative interpretation, discourse explanation or case model?
5. How structured should the process be? Flexible exploration, matrix-based comparison or predefined framework?
6. How important is sequence? If order and interaction matter, segment-based thematic coding may lose essential meaning.
7. How will theory be used? As a starting framework, a sensitising resource, or an outcome developed from data?
8. What volume and type of data are available? Method choice must remain feasible without sacrificing analytical depth.

10.1 Quick selection table

11. Common Mistakes Before Coding Begins

11.1 Choosing the method after data collection

Researchers sometimes collect interviews and only then decide how to analyse them. This can create mismatch because interview questions, sampling and transcription may not provide the evidence required by the chosen method.

11.2 Treating software as the method

NVivo, ATLAS.ti, MAXQDA and similar programs organise data; they do not decide what constitutes a theme or produce interpretation automatically. The researcher remains responsible for the analytical logic.

11.3 Collecting too much data

Large datasets may appear impressive but can produce superficial analysis. Researchers should align data volume with available time, the complexity of the method and the depth expected.

11.4 Failing to define the analytical unit

Without a clear unit of analysis, researchers may code sentences, participants, documents and events inconsistently. The unit can vary during analysis, but changes should be deliberate.

11.5 Ignoring context

Extracting quotations from their surrounding account may alter meaning. Context includes who spoke, where, in response to what, under which institutional conditions and at what point in the process.

11.6 Confusing interview questions with themes

Organising findings under the interview guide often reproduces data-collection categories rather than generating analysis. Themes should reflect patterns of meaning or explanation, not merely the order in which questions were asked.

11.7 Beginning with too many rigid codes

A detailed predefined codebook may be useful in deductive analysis, but it can prevent researchers from noticing unexpected ideas. Even framework-led studies should retain space for emergent codes.

11.8 Failing to document decisions

Researchers may remember why a theme was renamed or merged during analysis but cannot reconstruct the process months later. Memos, codebook versions and decision logs strengthen transparency.

11.9 Assuming agreement equals validity

High agreement between coders can demonstrate consistency in some designs, but it does not prove that the interpretation is meaningful. Some reflexive approaches treat coding differences as opportunities for discussion rather than errors to eliminate.

11.10 Using quotations as substitutes for analysis

A long quotation may illustrate a point, but the researcher must explain why it matters, how it relates to other evidence and what interpretation it supports.

12. What Examiners Expect from the Analysis Plan

Before examining the findings themselves, an examiner should be able to understand the analytical process from the methodology chapter.

12.1 Questions examiners commonly ask

• Why was this analytical method chosen?
• How does it fit the research question and methodology?
• What exactly was coded or analysed?
• When did analysis begin?
• Was the approach inductive, deductive or abductive?
• How were codes developed and refined?
• How did codes become themes, categories or explanations?
• How were contradictions and negative cases handled?
• How did reflexivity shape the process?
• What role did software play?
• How can a reader trace findings back to the data?

12.2 Weak and stronger methodological reporting

12.3 Package 1 analytical readiness checklist

12.4 Worked PhD example

Research question: How do public-sector employees interpret the introduction of AI-supported performance monitoring?

Potential approach: Reflexive thematic analysis may be appropriate if the purpose is to identify patterns of shared meaning across employee accounts. Framework analysis may be preferable if the study must compare departments, grades and implementation stages systematically. Discourse analysis would be more suitable if the research asks how employees and managers construct concepts such as fairness, accountability and trust through language.

Preparation: Interviews are transcribed using clean verbatim conventions, with pauses retained where they materially affect meaning. Participant attributes include role, department, length of service and exposure to the system. Each transcript receives a whole-case summary before coding.

Reflexivity: The researcher records prior concerns about surveillance technologies and actively examines accounts in which monitoring is experienced positively.

Analytical test: The final themes must explain more than whether employees liked or disliked monitoring. They should show how interpretations were shaped by professional autonomy, managerial communication, perceived procedural fairness and previous organisational experience.

13. Thematic Analysis

Thematic analysis identifies and interprets patterns of meaning across a qualitative dataset. It is widely used because it can answer questions about experiences, beliefs, practices, barriers, identities and social processes without requiring the researcher to adopt a single substantive theory.

Its flexibility is also its main risk. Researchers sometimes call any topic summary “thematic analysis.” A rigorous study must specify which form of thematic analysis it uses, how themes were developed and what role the researcher played in interpretation.

13.1 What thematic analysis produces

The output is a set of themes that capture patterned meaning relevant to the research question. A theme is not merely a frequently mentioned topic. It should organise evidence around a central interpretive idea.

13.2 Major forms of thematic analysis

These approaches should not be combined casually. For example, reflexive thematic analysis does not normally treat inter-coder agreement as proof of quality because interpretation is understood as situated rather than mechanically reproducible.

13.3 A practical thematic analysis process

1. Familiarise: Read the full dataset and write early analytical notes.
2. Code: Label relevant segments systematically across the corpus.
3. Develop candidate themes: Group codes around broader patterns of meaning.
4. Review themes: Test whether each theme works within individual extracts and across the dataset.
5. Define themes: Clarify the central organising concept, scope and boundaries.
6. Write analytically: Integrate interpretation, evidence, context and links to the research question.

13.4 Semantic and latent analysis

Semantic analysis focuses on explicit meaning: what participants directly say. Latent analysis examines assumptions, ideas or social meanings underlying those statements. A study may use both, but it should explain when it moves beyond explicit content.

13.5 Inductive and deductive thematic analysis

Inductive coding is primarily guided by the data, while deductive coding is shaped by theory, prior research or an analytical framework. Many doctoral studies use an abductive process: they begin openly, compare emerging patterns with theory, and return to the data to refine the explanation.

13.6 Worked example

In a study of AI-supported performance monitoring, initial codes might include fear of misinterpretation, managerial reassurance, gaming the metric, visibility of effort and loss of discretion. These could contribute to a broader theme such as “Measurement as a redefinition of trustworthy work.” The theme would examine how monitoring changes not only behaviour but also employees’ understanding of what the organisation recognises as legitimate performance.

13.7 Common mistakes

• Using interview questions as themes.
• Naming themes with one-word topics.
• Reporting only the most common opinions.
• Failing to distinguish codes from themes.
• Including themes that overlap heavily.
• Presenting quotations without interpretation.
• Claiming themes “emerged” without explaining analytical decisions.

14. Framework Analysis

Framework analysis is a systematic, matrix-based method developed for applied qualitative research. It is particularly useful when a study has focused questions, specific information needs and a requirement to compare participants, groups, organisations or cases transparently.

The method retains a clear connection between summarised findings and the original data while enabling structured comparison across a large dataset.

14.1 What framework analysis produces

The central analytical tool is a framework matrix. Rows usually represent cases or participants, while columns represent categories or analytical issues. Each cell contains a concise summary linked back to the source data.

14.2 Typical stages

1. Familiarisation: Review data and note recurrent and important issues.
2. Develop the analytical framework: Create categories from research questions, theory and emergent data.
3. Index or code the data: Apply the framework systematically.
4. Chart into the matrix: Summarise data by case and category while preserving references.
5. Map and interpret: Compare cases, identify patterns, explain differences and develop conclusions.

14.3 Strengths of framework analysis

• Supports transparent comparisons across cases and groups.
• Handles both deductive and inductive categories.
• Is well suited to multidisciplinary research teams.
• Creates a visible audit trail from interpretation to source material.
• Works effectively for policy, health, public-sector and evaluation studies.

14.4 Risks and limitations

The matrix can encourage over-compression. Rich accounts may be reduced to short summaries, and researchers may mistake charting for analysis. The interpretive stage must move beyond filling cells to explain why patterns occur, how categories interact and which contextual conditions shape differences.

14.5 Worked example

A multi-department study of digital transformation may use departments as cases and framework categories such as leadership support, participation, skills, workload, trust and adaptation. Comparison might reveal that resistance is not primarily caused by low digital competence but by whether employees believe implementation decisions can still be influenced.

14.6 Examiner expectations

A strong account explains how the framework was developed, whether categories were deductive or inductive, how summaries were produced, how the matrix retained links to context and how charted data were transformed into interpretation.

15. Qualitative Content Analysis

Qualitative content analysis systematically classifies textual, visual or documentary material into categories in order to interpret meaning, emphasis and patterns. It is especially useful when the researcher needs a structured analysis of a clearly defined corpus such as policy documents, reports, media texts, open-ended survey responses or interview transcripts.

15.1 What qualitative content analysis produces

The output is an organised category system. Categories may remain relatively descriptive or support deeper interpretation, depending on the research purpose and approach.

15.2 Units of analysis and coding units

The unit of analysis is the larger item being studied, such as an interview, newspaper article or policy document. The coding unit is the segment assigned to a category, such as a phrase, sentence, paragraph or visual feature. Researchers should define both clearly.

15.3 A practical content-analysis process

1. Define the corpus and inclusion criteria.
2. Choose the unit of analysis and coding unit.
3. Develop initial categories inductively, deductively or through a combined strategy.
4. Create category definitions, inclusion rules and examples.
5. Apply categories systematically across the corpus.
6. Review overlaps, gaps and ambiguous cases.
7. Compare patterns across source types, periods, groups or contexts.
8. Interpret what the category pattern reveals about the research problem.

15.4 Manifest and latent content

Manifest content refers to visible or explicit content. Latent content concerns underlying meaning. For example, a policy may explicitly mention “efficiency” repeatedly, while a latent analysis may interpret how efficiency is constructed as more legitimate than employee discretion.

15.5 Frequency and meaning

Counts can support qualitative content analysis, but they do not replace interpretation. Frequency may show prominence, while absences, contradictions and the context in which a category appears may be equally important.

15.6 Worked example

A researcher analysing national AI strategies may code references to innovation, risk, skills, public trust, accountability and economic competitiveness. The analysis may show not only which categories are frequent but how responsibility for AI risk is shifted between governments, firms and individuals.

15.7 Common mistakes

• Using vague categories that overlap.
• Failing to define the corpus.
• Mixing units of analysis inconsistently.
• Treating word counts as self-explanatory findings.
• Ignoring visual, structural or contextual features of documents.
• Calling a simple document summary “content analysis.”

16. Comparing Thematic, Framework and Content Analysis

17. Grounded Theory Analysis

Grounded theory analysis is used when the purpose of a study is to develop an explanatory account or theory of a social process. It is not simply a technique for coding interview transcripts. The design links data collection, coding, comparison, memo writing and theoretical sampling so that emerging explanations are repeatedly tested and refined.

17.1 Core analytical logic

The central logic is constant comparison. Each incident is compared with other incidents, each code with other codes, and each developing category with new evidence. The purpose is to identify similarities, differences, conditions, consequences and relationships rather than merely to group quotations by topic.

• Simultaneous data collection and analysis: early analysis shapes later recruitment and questioning.
• Constant comparison: evidence is compared within and across participants, settings and stages.
• Memo writing: analytical ideas are recorded as they develop.
• Theoretical sampling: later participants or cases are selected because they can clarify an emerging category.
• Theoretical integration: categories are connected into an explanatory model.

17.2 Initial or open coding

Initial coding breaks data into analytically meaningful actions, events or ideas. Codes should remain close enough to the evidence to preserve what participants are doing, experiencing or attempting to achieve. Gerunds—such as “protecting professional identity” or “negotiating managerial pressure”—often help the researcher focus on process rather than static topics.

17.3 Focused and axial coding

Focused coding identifies the most analytically powerful or frequently significant initial codes and uses them to examine larger portions of the dataset. In traditions that use axial coding, the researcher then explores the conditions that give rise to a category, the context in which it occurs, the actions participants take and the consequences that follow.

17.4 Selective coding and theoretical integration

Later analysis identifies a central category that explains how the major categories relate. The final theory should do more than list themes. It should offer a coherent account of the process, show variation, identify conditions and explain outcomes.

Illustrative progression: initial codes such as “testing the new system privately,” “seeking peer reassurance” and “avoiding visible mistakes” may develop into the focused category managing exposure during technological change. This category may then connect with others to explain how employees move from guarded experimentation to confident adoption.

17.5 Theoretical sampling and saturation

Theoretical sampling is not the same as purposive sampling at the beginning of a study. It occurs after categories start to develop. The researcher deliberately seeks people, incidents or settings that can test gaps, boundaries or contradictions in the emerging explanation. Recruitment stops when additional evidence no longer adds important properties or relationships to the developing theory.

17.6 Common grounded theory errors

• Using the label “grounded theory” for any inductive coding exercise.
• Collecting the entire sample before analysis begins.
• Producing themes without developing categories or an explanatory model.
• Ignoring theoretical sampling and memo writing.
• Mixing procedures from different grounded theory traditions without explanation.

18. Interpretative Phenomenological Analysis (IPA)

Interpretative phenomenological analysis examines how people make sense of significant lived experiences. It is phenomenological because it attends closely to experience, hermeneutic because analysis involves interpretation, and idiographic because each case is examined in depth before patterns are considered across cases.

18.1 The double hermeneutic

Participants are trying to make sense of their experience, while the researcher is trying to make sense of the participants’ sense-making. The analysis therefore does not claim to reproduce experience directly. It develops a careful, evidence-based interpretation of how experience is understood and expressed.

18.2 Idiographic commitment

IPA begins with detailed case-by-case analysis. The researcher should resist moving too quickly to cross-case themes. Each transcript is treated as a whole account with its own language, tensions and context. Only after a case has been analysed in depth does the researcher move to the next case.

18.3 A practical IPA process

1. Read and re-read: become immersed in the participant’s account.
2. Make exploratory notes: record descriptive, linguistic and conceptual observations.
3. Develop experiential statements or emergent themes: condense important elements while preserving complexity.
4. Connect themes within the case: examine relationships, contrasts, chronology and context.
5. Move to the next case: bracket, as far as possible, ideas from the previous case.
6. Identify patterns across cases: examine convergence, divergence and variation.
7. Write an interpretative account: integrate analytical commentary with carefully selected extracts.

18.4 Worked example

A study explores how first-generation doctoral students experience academic belonging. One participant repeatedly describes “performing confidence” in seminars while privately fearing exposure. Linguistic notes identify hesitation, humour and shifts between “I” and “we.” The interpretation develops around the tension between visible competence and hidden uncertainty. Across cases, the researcher may identify a broader experiential pattern such as belonging as a performance that must continually be sustained, while still showing how that pattern differs between participants.

18.5 What IPA should not become

• A broad thematic analysis of a large and diverse sample.
• A summary of what participants said without interpretation.
• A search for universal psychological laws.
• A method in which cross-case themes erase individual cases.
• A collection of quotations presented without close linguistic or conceptual analysis.

19. Narrative Analysis

Narrative analysis investigates how people organise experience into stories. It examines not only what is said but how events are ordered, how characters and identities are constructed, how turning points are presented, and how cultural narratives shape personal accounts.

19.1 Units of narrative analysis

The unit may be a complete life story, a story about a particular event, a sequence within an interview, a written account, an organisational history or a collection of narratives circulating in a community. The researcher must define what counts as a narrative and avoid fragmenting stories into decontextualised codes unless the chosen narrative approach permits it.

19.2 Major narrative approaches

19.3 A practical analytical process

1. Identify complete stories or narrative episodes.
2. Preserve chronology, context and the relationship between events.
3. Examine plot, characters, turning points, tensions and resolution.
4. Analyse how the narrator positions self and others.
5. Consider audience, interview context and wider cultural narratives.
6. Compare stories without erasing important differences.
7. Develop an interpretation that links narrative form and content to the research question.

19.4 Worked example

Entrepreneurs describing business failure may construct very different narratives from similar events. One may tell a redemption story in which failure becomes evidence of resilience; another may present a contamination narrative in which one decision permanently damages identity and relationships. Narrative analysis examines how these plots shape the meaning of failure rather than treating all references to “failure” as one theme.

19.5 Restorying and its risks

Restorying reorganises fragmented interview material into a coherent chronological account. It can clarify a participant’s trajectory, but it also gives the researcher considerable interpretative power. Any reorganisation should therefore be transparent, faithful to the original account and sensitive to ambiguity, contradiction and silence.

19.6 Common narrative-analysis errors

• Breaking stories into codes until chronology and plot disappear.
• Treating any interview response as a narrative.
• Analysing content but ignoring structure, audience and performance.
• Presenting a polished story that removes contradiction.
• Generalising across cases without preserving distinctive narratives.

20. Comparing Grounded Theory, IPA and Narrative Analysis

21. Method-Selection Decision Framework

22. Common Mistakes and Examiner Checklist

22.1 Common mistakes

• Choosing a prestigious method without matching it to the research question.
• Using “grounded theory,” “phenomenology” or “narrative” as decorative labels.
• Combining incompatible procedures without explaining the methodological logic.
• Describing coding steps while failing to show interpretation.
• Ignoring contradictory cases because they disrupt a clean account.
• Allowing software nodes or code counts to determine the final findings.
• Using the same generic analytical process regardless of method.

22.2 Examiner checklist

• Is the analytical method clearly aligned with the research question and methodology?
• Does the thesis explain the unit of analysis?
• Are the analytical stages described accurately and transparently?
• Can the reader trace claims back to evidence?
• Does the analysis preserve context, variation and negative cases?
• Is interpretation distinguishable from description?
• Are methodological limitations acknowledged?
• Does the final output match what the chosen method is supposed to produce?

23. Discourse Analysis

Discourse analysis examines how language constructs social reality rather than treating speech or text as a transparent report of facts. It asks how particular descriptions become credible, how identities and categories are produced, which assumptions are normalised, and how language supports or challenges relations of power.

23.1 Major traditions

23.2 A practical analytical process

1. Define the corpus, context and social issue being examined.
2. Identify recurrent ways of describing people, problems and solutions.
3. Examine vocabulary, metaphors, contrasts, categories and absences.
4. Analyse the subject positions made available to speakers and audiences.
5. Consider what actions the discourse enables, legitimises or prevents.
6. Connect textual patterns to institutional, historical or political conditions.
7. Test the interpretation against contradictory texts and alternative readings.

23.3 Worked example

A study of university AI policies may identify competing discourses of innovation, academic integrity and student vulnerability. The analysis would not simply count these terms. It would examine how each discourse constructs students, teachers and technology differently—for example, students as potential offenders, as learners requiring protection, or as future professionals who must acquire AI competence.

23.4 Common mistakes

• Treating discourse analysis as ordinary thematic coding of language-related topics.
• Claiming that a text reveals the author's private beliefs.
• Ignoring historical and institutional context.
• Using concepts such as power or ideology without demonstrating them in the data.
• Selecting isolated quotations without analysing the broader discursive pattern.

24. Conversation Analysis

Conversation analysis studies how social interaction is organised moment by moment. It treats order, timing, turn-taking and repair as analytically meaningful. The method normally relies on naturally occurring interaction and highly detailed recordings rather than conventional research interviews alone.

24.1 Core concepts

24.2 Transcription requirements

Conversation analysis often uses Jefferson-style notation to capture pauses, overlap, emphasis, cut-offs, stretched sounds, pace and intonation. The transcript is therefore an analytical representation of interaction rather than a simple written record.

24.3 Worked example

In clinical consultations, a patient may respond to a treatment recommendation with “I suppose I could try it.” Conversation analysis would examine the delay, hedging, intonation and the clinician's next turn to determine how reluctance is displayed and managed. The finding concerns the interactional accomplishment of consent, not merely the topic of treatment preference.

24.4 Common mistakes

• Using ordinary clean-verbatim transcripts for fine-grained interactional claims.
• Removing pauses or overlaps that are central to interpretation.
• Analysing interview opinions rather than naturally occurring action.
• Explaining behaviour through hidden motives instead of observable sequence.

25. Template Analysis

Template analysis is a flexible form of thematic coding that organises codes into a hierarchical template. Researchers may begin with a small set of a priori codes derived from the research questions or theory, while remaining open to substantial revision as the data are analysed.

25.1 Building the template

1. Read a purposive subset of the data.
2. Create initial descriptive and interpretive codes.
3. Organise codes into broad themes and nested subthemes.
4. Apply the initial template to additional cases.
5. Revise definitions, levels and relationships.
6. Remove redundant codes and add unanticipated categories.
7. Finalise the template when it captures the dataset adequately.

25.2 Illustrative hierarchy

• 1. Experiencing digital change
  • 1.1 Anticipating disruption
  • 1.2 Learning through experimentation
  • 1.3 Protecting professional identity
• 2. Organisational conditions
  • 2.1 Managerial communication
  • 2.2 Peer support
  • 2.3 Availability of discretion

25.3 Advantages and limitations

26. Document Analysis

Document analysis examines written, visual or digital records as evidence of events, policy, institutional practice, representation and meaning. Documents should not be treated as neutral containers of information. They are produced by particular actors, for particular audiences, under specific organisational and historical conditions.

26.1 Types of documents

• laws, policies, strategies and official guidance;
• annual reports, minutes, memoranda and internal communications;
• newspapers, websites, advertisements and social-media material;
• diaries, letters, autobiographies and personal archives;
• photographs, forms, templates and administrative records;
• historical archives and institutional records.

26.2 Evaluating documentary evidence

26.3 Practical workflow

1. Define the documentary corpus and inclusion dates.
2. Record provenance, author, audience, purpose and institutional setting.
3. Assess authenticity, omissions and conditions of access.
4. Select an analytical approach: content, thematic, discourse, narrative or historical analysis.
5. Compare documents across actors, time periods and organisational levels.
6. Triangulate documentary claims with other evidence where appropriate.

26.4 Worked example

A study of public-sector digital transformation may compare strategy documents, staff guidance and meeting minutes. Strategy texts may present implementation as inevitable progress, while minutes reveal unresolved resource and workload concerns. The discrepancy becomes an analytical finding about the difference between public justification and internal practice.

27. Case Study Analysis

Case study analysis explains a bounded case—such as an organisation, programme, community, event or implementation process—in its real-world context. A case study is a research design rather than a single coding technique. It may combine interviews, observations, documents and quantitative records.

27.1 Preserve the integrity of the case

Researchers should normally complete within-case analysis before pooling material across cases. This preserves chronology, context, actors and causal relationships that may disappear when all data are coded together.

27.2 Major analytical strategies

27.3 Worked example

A multiple-case study compares AI implementation in three public agencies. Within-case analysis shows that each agency faced similar technical constraints but responded differently. Cross-case comparison reveals that employee participation affected whether problems were interpreted as temporary implementation difficulties or evidence that the system was illegitimate.

27.4 Common mistakes

• Calling a small sample a case study without defining case boundaries.
• Combining all cases into themes before understanding each case.
• Confusing the case with the unit of data collection.
• Claiming causal explanation without examining rival explanations.

28. Ethnographic Analysis

Ethnographic analysis interprets cultural practices, meanings and social organisation through sustained engagement with a field setting. It draws on observation, participation, fieldnotes, interviews, artefacts and documents to explain how members understand and enact their world.

28.1 From fieldnotes to cultural interpretation

1. Expand fieldnotes promptly after observation.
2. Separate observed action, participant explanation and researcher interpretation.
3. Code recurring practices, interactions, language, spaces and artefacts.
4. Compare formal rules with everyday behaviour.
5. Identify cultural domains, categories, contrasts and tacit norms.
6. Develop thick description linking action to context and meaning.
7. Examine deviant cases, boundary moments and changes over time.

28.2 Insider and outsider perspectives

Ethnographers move between emic understanding—how members describe their world—and etic interpretation—how the researcher explains wider patterns. Strong analysis preserves local meanings while developing concepts that extend beyond participants' own vocabulary.

28.3 Worked example

In an ethnography of a hospital technology team, formal procedures may emphasise documentation and accountability, while observation reveals that urgent problems are solved through informal relationships and improvised messaging groups. Thick description can show how trust and reputation function as an unofficial coordination system.

29. Qualitative Comparative Analysis (QCA)

Qualitative Comparative Analysis is a set-theoretic method for examining how combinations of conditions are associated with an outcome across a modest number of cases. Despite its name, QCA is not a conventional qualitative coding method. It translates case knowledge into calibrated sets and identifies configurations that may be necessary or sufficient for an outcome.

29.1 Main forms

29.2 Necessary and sufficient conditions

A necessary condition is present whenever the outcome occurs, although it may also appear without the outcome. A sufficient condition or configuration reliably produces the outcome, even though other pathways may also produce it. QCA therefore supports equifinality: more than one combination of conditions can lead to the same result.

29.3 Practical workflow

1. Define cases, outcome and theoretically relevant conditions.
2. Develop transparent calibration rules.
3. Construct a data matrix and truth table.
4. Assess consistency and coverage.
5. Resolve contradictory configurations using case knowledge.
6. Minimise configurations into solution pathways.
7. Return to cases to interpret mechanisms and limitations.

29.4 Worked example

A study of twenty public organisations may examine which combinations of leadership support, employee participation, data quality and training are associated with successful AI adoption. QCA may show that strong leadership is not sufficient alone, but adoption occurs through either leadership plus participation or high data quality plus intensive training.

29.5 Common mistakes

• Selecting conditions without theory or strong case knowledge.
• Using arbitrary calibration thresholds.
• Interpreting configurations as universal causal laws.
• Ignoring contradictory cases and limited diversity.
• Reporting software output without substantive case interpretation.

30. Complete Qualitative Method-Selection Matrix

30.1 Software should follow method

NVivo, ATLAS.ti, MAXQDA, Dedoose and spreadsheet tools can support coding, retrieval and comparison. Specialist packages such as fsQCA or R support set-theoretic analysis. Software choice should follow the analytical method and dataset rather than determine them.

31. Examiner's Decision Framework

A convincing methodology chapter should allow an examiner to reconstruct why the chosen method was necessary and how it produced the final claims.

1. State the required form of knowledge. Does the study seek themes, experience, stories, discourse, interaction, culture, case explanation, theory or configurations?
2. Identify the unit of analysis. Clarify whether the analysis concerns segments, cases, narratives, turns, documents, practices or calibrated sets.
3. Show methodological compatibility. Explain how the analysis fits the research design, philosophy and data-generation process.
4. Describe the distinctive analytical logic. Name the procedures that make the method different from generic coding.
5. Demonstrate traceability. Show how evidence moved into codes, categories, interpretations and conclusions.
6. Address researcher judgement. Explain reflexivity, team decisions, disagreements and alternative interpretations.
7. Test boundaries and contradictions. Include negative cases, rival explanations and contextual variation.
8. Match claims to design. Avoid population-wide or causal claims that the evidence cannot support.

31.1 Method-justification template

31.2 Final examiner checklist

• Is the method named precisely rather than generically?
• Is the choice driven by the research question?
• Are the data and transcription suitable for the method?
• Is the unit of analysis explicit?
• Are analytical stages method-specific and reproducible in principle?
• Are evidence, interpretation and theory clearly connected?
• Are contradictions and limitations addressed?
• Does the output match what the method promises to produce?

Conclusion

Package 4 completes the guide's overview of major qualitative analysis methods. Discourse analysis explains how language constructs reality and power; conversation analysis examines the sequential organisation of interaction; template analysis provides flexible hierarchical coding; document, case study and ethnographic analysis preserve context in different ways; and QCA examines configurational pathways across cases.

No single method is inherently superior. The strongest choice is the one that produces the form of knowledge required by the research question while remaining compatible with the study's philosophy, design, data and unit of analysis. A defensible thesis makes this logic explicit and shows a traceable movement from evidence to interpretation.