The Complete Guide to Thematic Analysis

This final cumulative Package 6 preserves every section from Packages 1–5 and adds a practical application and defence toolkit: worked analytical templates, reporting models, viva preparation, troubleshooting, audit documents, supervisor-review workflows and a final submission-readiness framework.

This final cumulative Package 5 preserves every section from Packages 1–4 and adds a complete NVivo workflow, advanced queries and visualisation, AI-assisted thematic analysis, prompt engineering, confidentiality and ethics safeguards, audit trails, publication guidance and an integrated end-to-end quality framework.

1. What Is Thematic Analysis?

Thematic analysis examines a qualitative dataset to identify patterns of shared meaning that are relevant to a research question. These patterns are developed through systematic engagement with the data and are expressed as themes. A theme is not simply a frequently mentioned topic. It is an interpretive pattern that organises evidence around a central idea.

1.1 The basic analytical movement

A thematic analysis usually moves from a broad dataset toward increasingly focused interpretation:

This movement is rarely linear. Researchers return repeatedly to the original data, revise codes, reconsider theme boundaries and test whether their developing interpretation genuinely answers the research question.

1.2 Topic summaries are not automatically themes

1.3 What thematic analysis can produce

• a structured account of shared experiences;
• an explanation of how participants understand a process;
• an interpretation of assumptions embedded in accounts;
• a comparison of patterned meanings across groups or contexts;
• a theoretically informed account of how social meanings are produced.

1.4 Strengths

• Applicable across many disciplines and data types.
• Can support experiential, critical, constructionist and pragmatic inquiries.
• Produces accessible findings while allowing substantial interpretive depth.
• Works with relatively small or large qualitative datasets.
• Can be conducted manually or with software such as NVivo.

1.5 Limitations

• Flexibility can lead to methodological vagueness.
• Researchers may combine incompatible procedures without recognising it.
• Theme development can remain descriptive if theoretical engagement is weak.
• Claims about reliability or saturation may be inappropriate for some reflexive approaches.
• Software outputs can be mistaken for analysis.

2. Development of Thematic Analysis

Researchers identified patterns in qualitative material long before thematic analysis was formalised as a named method. Its contemporary visibility is strongly associated with the stepwise guidance developed by Virginia Braun and Victoria Clarke. Their work helped establish thematic analysis as a method in its own right rather than an unnamed procedure used inside other methodologies.

Subsequent methodological debate clarified that “thematic analysis” does not refer to one uniform technique. Different traditions make different assumptions about researcher subjectivity, coding frames, agreement between coders and the status of themes.

2.1 Why this development matters

A thesis that simply states “the data were analysed thematically” leaves major questions unanswered:

• Was the approach reflexive, codebook-based or reliability-oriented?
• Were codes generated inductively, deductively or through a combination?
• Did the researcher treat themes as discovered entities or interpretive constructions?
• Was coding agreement sought, and if so, why?
• What theoretical assumptions guided interpretation?

3. Three Major Families of Thematic Analysis

A useful contemporary distinction is between reflexive thematic analysis, codebook thematic analysis and coding-reliability thematic analysis. These are not rigid boxes, but they help researchers avoid mixing procedures built on contradictory assumptions.

3.1 Reflexive thematic analysis

Reflexive thematic analysis treats researcher subjectivity as an analytical resource that must be examined rather than eliminated. Coding is usually organic and evolves as the researcher develops understanding. Themes are produced through sustained engagement with data, theory and context.

This approach is well suited to studies asking how people make sense of experiences, how meanings are organised, or how social assumptions shape accounts. It can be more experiential or more critical depending on the theoretical framing.

3.2 Codebook thematic analysis

Codebook approaches use a structured coding framework to organise analysis. The codebook may begin from the research questions, prior theory, interview schedule or early data engagement and is revised as analysis proceeds. This is useful for multidisciplinary teams, applied projects, large datasets and studies requiring comparison across predefined domains.

3.3 Coding-reliability approaches

Coding-reliability approaches emphasise consistency between coders. Categories may be defined in advance, coders trained and agreement assessed. This logic is appropriate only when the study’s epistemological position treats coding consistency as meaningful evidence of quality.

4. When Should You Use Thematic Analysis?

Thematic analysis is appropriate when the research question concerns recurring patterns of meaning across a dataset and does not require the distinctive commitments of another methodology.

4.1 Suitable purposes

• understanding shared experiences of a policy, service or transition;
• examining barriers, facilitators and perceived consequences;
• exploring how professionals understand roles or responsibilities;
• identifying patterned meanings across interviews, focus groups or documents;
• comparing perspectives across participant groups;
• developing a theoretically informed account without constructing a full grounded theory.

4.2 Suitable data

• semi-structured and unstructured interviews;
• focus groups;
• open-ended survey responses;
• diaries and reflective accounts;
• policy or organisational documents;
• online discussions and social-media material;
• field notes, where patterned meaning is the analytical focus.

4.3 Worked examples

5. When Thematic Analysis Is Not the Best Choice

Thematic analysis should not be selected merely because it appears easier. Another method may be more suitable when the research question requires a specific analytical object.

5.1 A practical decision rule

Choose thematic analysis when patterned meaning across the dataset is the main object of analysis. Choose another method when experience, discourse, narrative form, interaction, theory generation or case configuration is itself the central analytical object.

6. Research Questions and Study Designs

6.1 Questions that work well

• How do participants experience…?
• How do participants understand…?
• What meanings do participants attach to…?
• What barriers and facilitators shape…?
• How are perceptions similar or different across groups?
• How does a particular context influence participants’ accounts?

6.2 Questions that may be too broad

“What do employees think about digital transformation?” provides little analytical direction. A stronger formulation might be: “How do frontline employees interpret the effects of digital transformation on professional autonomy and service quality?”

6.3 Alignment matrix

7. Philosophy and Theoretical Alignment

Thematic analysis is theoretically flexible, but it is not theoretically empty. Researchers must explain what they believe qualitative accounts can reveal and how they understand the relationship between language, experience and social reality.

7.1 Realist or experiential orientation

Participant accounts are treated as meaningful access to experiences, perceptions and realities, while recognising that accounts are situated and partial.

7.2 Constructionist orientation

Analysis examines how meanings are socially produced through language, culture, institutions and available interpretive resources.

7.3 Critical realist orientation

Accounts are interpreted as shaped by both real conditions and socially mediated understanding. Researchers may move between reported experience, institutional structures and underlying mechanisms.

7.4 Pragmatic orientation

The analysis is organised around the practical problem and research purpose, while still making transparent how meanings and interpretations are handled.

7.5 Semantic and latent analysis

7.6 Inductive, deductive and abductive coding

8. Preparing the Dataset

Good thematic analysis begins before coding. Poor file organisation, inconsistent transcription, missing metadata and uncontrolled document versions can undermine the audit trail.

8.1 Define the dataset

Distinguish between the entire data corpus, the specific dataset selected for analysis and individual data items such as transcripts or documents. Explain inclusion and exclusion decisions.

8.2 Suggested folder structure

8.3 Participant identifiers

Use stable identifiers that allow linkage without exposing identity. For example, “P07_Manager_PublicHospital” may preserve analytically useful attributes, but only where re-identification risk is adequately controlled.

8.4 Metadata

Record attributes relevant to the analysis, such as role, location, experience level, organisational unit or study wave. Do not collect or retain unnecessary identifying data.

8.5 Version control

• Use dated filenames or a controlled version convention.
• Do not overwrite the only copy of an NVivo project.
• Keep a change log for major codebook and theme revisions.
• Store encrypted backups in approved locations.

8.6 Data integrity check

• Every transcript corresponds to the correct recording.
• Speaker labels are consistent.
• Missing sections and inaudible passages are marked.
• Anonymisation has been checked.
• Consent and data-governance conditions permit the intended analysis.
• All files open correctly before import into NVivo.

9. Transcription Decisions

Transcription is an analytical decision, not a neutral clerical task. The required level of detail depends on the research question.

9.1 Quality assurance

Researchers should listen to recordings while checking transcripts, especially when transcription has been outsourced or automated. Automated transcription may save time but can distort accents, technical terms, names, code-switching and emotionally significant passages.

9.2 Translation

When data are collected in one language and analysed or reported in another, document who translated the material, at what stage, how ambiguous terms were handled and whether original-language extracts were retained for checking.

9.3 Non-verbal features

Record laughter, long pauses, emotional expressions or interruptions only when they may contribute to interpretation. Excessive notation can make transcripts unreadable without adding analytical value.

10. Familiarisation

Familiarisation is the sustained process of becoming analytically acquainted with the breadth, depth and texture of the dataset. It starts during data generation and continues throughout analysis.

10.1 A practical familiarisation cycle

1. Listen to the recording or review the original source.
2. Read the transcript in full without immediately formalising codes.
3. Write brief notes about initial impressions, tensions and surprises.
4. Record questions rather than premature conclusions.
5. Compare the item with earlier data while preserving its distinctive context.
6. Return to the research question and note potentially relevant patterns.

10.2 Familiarisation memo template

10.3 Worked example

10.4 What familiarisation is not

• reading transcripts once;
• highlighting only statements that confirm expectations;
• writing polished themes before systematic coding;
• relying solely on word-frequency outputs;
• delegating all reading to software or AI.

11. Reflexivity Before Coding

Reflexivity is the disciplined examination of how the researcher’s position, assumptions, decisions and relationships shape knowledge production. It does not require the researcher to eliminate subjectivity; it requires subjectivity to be recognised and worked with transparently.

11.1 Sources of researcher influence

• disciplinary training and preferred theories;
• professional experience;
• insider or outsider status;
• personal investment in the topic;
• expectations developed from the literature;
• relationships with participants or organisations;
• funding, access and institutional constraints.

11.2 Positionality statement prompts

• What is my relationship to the setting and participants?
• Which experiences make some interpretations more visible to me?
• Which assumptions do I bring about the phenomenon?
• Where might power differences have shaped data generation?
• Which interpretations do I find uncomfortable or attractive?
• How will I document analytical decisions and changes?

11.3 Reflexive journal structure

11.4 Bracketing

Researchers sometimes use the term “bracketing” to describe efforts to identify and temporarily hold assumptions in view. In thematic analysis, complete removal of prior understanding is rarely plausible. A more defensible goal is transparent reflexive engagement.

12. Overview of the Six-Phase Process

The six phases commonly associated with reflexive thematic analysis provide a recursive analytical process rather than a rigid checklist.

12.1 Recursive movement

A researcher may begin writing a theme and discover that its central idea is unclear, requiring renewed coding and comparison. This is not analytical failure. It is normal qualitative reasoning when changes are documented and justified.

12.2 Where NVivo fits

NVivo can assist with storage, coding, retrieval, comparisons, memos, queries and visualisation. It does not decide what matters, define a theme’s central organising concept or determine whether an interpretation is credible. Package 5 will provide a complete NVivo workflow.

13. Building a Defensible Thematic Analysis Plan

Before coding, write a concise analysis protocol. This prevents later confusion and creates a basis for transparent methodological reporting.

13.1 Analysis-plan template

13.2 Example analysis statement

13.3 Team analysis

Team involvement should match the chosen approach. In reflexive thematic analysis, team members may independently engage with selected data and discuss different readings to deepen interpretation. The purpose is not necessarily to force consensus. In codebook or reliability-oriented designs, calibration and coding consistency may have a different role.

14. Common Beginner Mistakes

1. Calling interview topics themes. Themes need a central organising concept.
2. Assuming repetition equals importance. Frequency can inform analysis but does not determine meaning.
3. Choosing thematic analysis because it seems easy. The method must fit the question.
4. Mixing incompatible variants. Reliability testing should not be added automatically to a reflexive design.
5. Claiming purely inductive analysis. Researchers always bring prior knowledge and assumptions.
6. Skipping familiarisation. Coding without understanding the dataset fragments context.
7. Using the interview guide as the final codebook. This often reproduces questions rather than generating analysis.
8. Letting NVivo create the analysis. Software retrieves and organises; the researcher interprets.
9. Treating every code as equally important. Analytical relevance is not the same as data volume.
10. Writing themes too early. Premature closure limits discovery and comparison.
11. Ignoring contradictions. Deviant and negative cases sharpen interpretation.
12. Over-anonymising data. Removing all context can weaken analysis.
13. Under-anonymising data. Detailed combinations of attributes may reveal identity.
14. Using quotations as substitutes for analysis. Extracts require interpretive commentary.
15. Confusing codes and themes. Codes capture features; themes organise broader patterns of meaning.
16. Reporting only procedure. A methods chapter must explain analytical logic, not just steps.
17. Using “bias” only as a threat. Researcher influence should be examined reflexively.
18. Claiming saturation without definition. The relevance of saturation depends on the approach.
19. Seeking certainty where interpretation is required. Quality comes from coherent, grounded reasoning, not mechanical proof.
20. Failing to preserve an audit trail. Major changes in codes and themes should be documented.
21. Ignoring theoretical fit. Semantic, latent, experiential and critical claims require different justifications.
22. Overloading themes. A theme containing unrelated ideas lacks internal coherence.
23. Producing too many themes. A fragmented chapter often reflects insufficient synthesis.
24. Producing one theme for every research question. Theme structure should emerge from the analytical account, not a formatting formula.
25. Using AI-generated codes without verification. Confidentiality, hallucination and loss of context must be controlled.

15. What Examiners Expect Before Coding Begins

Examiners assess whether the analytical design is coherent, transparent and capable of answering the research question.

15.1 Weak methodological wording

15.2 Stronger methodological wording

15.3 Questions you should be able to answer in a viva

• Why did you choose thematic analysis instead of grounded theory or content analysis?
• How do you define a theme?
• What makes your analysis reflexive?
• How did your position influence interpretation?
• How did you move from data extracts to themes?
• What did NVivo contribute, and what did it not do?
• How did you handle contradictions and minority accounts?
• Why are your claims warranted by the evidence?

16. Package 1 Preparation Checklist

• My research question genuinely requires analysis of patterned meaning.
• I can explain why thematic analysis is more suitable than competing approaches.
• I have identified the family of thematic analysis I am using.
• My epistemological and theoretical assumptions are explicit.
• I have decided whether coding will be inductive, deductive, abductive or hybrid.
• I have decided whether analysis will be semantic, latent or move between both.
• I have defined the data corpus and analytical dataset.
• Transcription and translation decisions are documented.
• Files are anonymised, organised and backed up.
• Relevant participant or document attributes are preserved securely.
• I have completed initial familiarisation memos.
• I have begun a reflexive journal and assumptions log.
• The role of collaborators is consistent with the chosen analytical approach.
• The role of NVivo is described as supportive rather than determinative.
• I have a written analysis protocol before detailed coding begins.

17. Coding Foundations: What Coding Actually Does

Coding is the process of assigning concise analytical labels to segments of qualitative data that are relevant to the research question. A code can identify an action, perception, emotion, process, assumption, contradiction, contextual condition or possible explanation. Coding reduces a large dataset into manageable analytical units without replacing the need to retain context.

17.1 Description, interpretation and abstraction

A strong analysis may move between these levels. Early coding often stays closer to participants’ words, while later coding becomes more interpretive as the researcher compares cases and develops an explanation.

17.2 What makes a useful code?

• It is relevant to the research question.
• It captures one coherent idea.
• Its label is specific enough to guide later retrieval.
• It can be distinguished from neighbouring codes.
• It remains connected to the original extract and context.
• It supports comparison across the dataset.

17.3 Data-driven and concept-driven coding

18. Choosing the Unit of Coding

The unit of coding is the segment of data to which a code is attached. It may be a word, phrase, sentence, paragraph, conversational turn, document section or longer passage. The correct unit is the smallest segment that preserves enough context for the intended interpretation.

18.1 Common units

18.2 Semantic completeness

Do not code only the striking sentence if the preceding or following lines are required to understand it. Include enough text to preserve who is speaking, what event is being discussed and whether the statement is serious, ironic, hypothetical or retrospective.

18.3 Overlapping coding

The same extract may receive multiple codes when it legitimately expresses several analytically relevant ideas. Overlapping coding is not duplication if each code captures a distinct aspect of meaning.

19. Coding Techniques for Thematic Analysis

19.1 Descriptive coding

Descriptive codes summarise the explicit subject of a passage. They are useful for initial organisation but may remain too broad for final analysis.

19.2 In Vivo coding

In Vivo coding uses participants’ own words as labels. It preserves distinctive language and can reveal locally important expressions, identities or metaphors.

In Vivo codes should not be used merely because a phrase sounds memorable. Their analytical value lies in how they represent shared meanings or cultural language.

19.3 Process coding

Process codes use gerunds to capture action and change, such as negotiating authority, avoiding disclosure, reframing failure or learning through improvisation. They are especially useful for research questions about implementation, adaptation and decision-making.

19.4 Emotion coding

Emotion codes identify feelings expressed, implied or attributed in the data. The researcher should distinguish participant-reported emotion from the researcher’s interpretation.

19.5 Values coding

Values coding captures beliefs, attitudes and principles that shape participants’ judgements. Examples include fairness as equal treatment, loyalty to professional standards and efficiency valued over participation.

19.6 Structural coding

Structural codes mark passages relating to research-question domains or interview topics. They are useful for navigating large datasets, but usually need to be supplemented by more interpretive codes.

19.7 Evaluation coding

Evaluation codes capture judgements of effectiveness, value, legitimacy or quality, such as training viewed as performative or policy considered operationally unrealistic.

19.8 Causation coding

Causation codes identify participant explanations of why events occurred. They should be framed as attributed explanations rather than proven causal facts.

19.9 Versus coding

Versus codes capture conflict or tension: speed versus consultation, standardisation versus professional discretion, or visibility versus psychological safety.

19.10 Magnitude coding

Magnitude coding records intensity, frequency or direction alongside a code. For example, resistance may be coded as low, moderate or high when such distinctions are meaningful and consistently applied.

19.11 Simultaneous coding

Simultaneous coding applies more than one code to the same passage. It is appropriate when an extract connects process, emotion and context—for example, concealing uncertainty, fear of judgement and hierarchical team culture.

19.12 Pattern coding

Pattern codes are later-cycle codes that synthesise several initial codes into a more explanatory category. For example, learning informally, peer troubleshooting and avoiding official support may be brought together under shadow capability building.

20. A Practical First-Cycle Coding Workflow

1. Select one information-rich transcript rather than automatically starting with Participant 1.
2. Read the transcript without coding to recover its overall account.
3. Write a brief case memo summarising context, key tensions and first impressions.
4. Code systematically from beginning to end.
5. Keep code labels provisional and specific.
6. Record uncertainties and emerging interpretations in memos.
7. Code a contrasting transcript next.
8. Compare new extracts with existing codes.
9. Create, split, rename or merge codes only when the distinction is analytically meaningful.
10. Review uncoded passages before closing the transcript.

20.1 How much should be coded?

In an inclusive reflexive analysis, researchers may code broadly across the dataset during initial phases. In a tightly deductive project, only material relevant to predefined analytical questions may be coded. The choice should be stated explicitly.

20.2 Line-by-line versus selective coding

20.3 Coding questions to ask

• What is happening in this passage?
• What does the participant appear to be trying to achieve?
• What assumption makes this account sensible?
• What condition enables or constrains the action?
• What is being valued, feared, resisted or normalised?
• How does this passage relate to other cases?
• What alternative interpretation is possible?

21. Building a Useful Codebook

A codebook is a structured record of code names, definitions, inclusion rules, exclusions and examples. It is essential in codebook and reliability-oriented designs and can also be useful as a flexible documentation tool in reflexive thematic analysis, provided it is not treated as permanently fixed.

21.1 Recommended fields

21.2 Example codebook entry

21.3 Parent and child codes

Hierarchies can improve organisation, but they should not be mistaken for themes. A parent code such as responses to organisational change might contain child codes for complying strategically, withdrawing effort and reframing change as opportunity.

21.4 Naming rules

• Prefer precise phrases over vague nouns.
• Use parallel grammatical forms within a family of codes.
• Avoid code names that contain several unrelated ideas joined by “and.”
• Do not create synonyms unless they capture meaningful distinctions.
• Use participant language when it adds analytical value.

22. Refining, Merging and Splitting Codes

22.1 When to merge

Merge codes when they consistently capture the same underlying meaning and their separation does not improve analysis. Review all extracts before merging to avoid erasing differences.

22.2 When to split

Split a code when it contains distinct actions, contexts or meanings that behave differently across the dataset. For example, resistance may need to be separated into public challenge, procedural delay and private disengagement.

22.3 When to retire a code

A code may be retired if it is irrelevant to the research question, too broad to guide interpretation, duplicated elsewhere or supported only by a misread extract. Retired codes should remain visible in the audit trail.

22.4 Code density

A very large number of codes is not automatically evidence of rigour. Excessive fragmentation can make synthesis difficult. Conversely, a very small number of broad codes may reproduce interview topics rather than analyse meaning.

23. Worked Coding Example

23.1 Possible later pattern

Across several cases, the codes efficiency producing hidden labour, political suppression of feedback and shadow system restoring control might contribute to a candidate theme such as “Employees absorbing the contradictions of official transformation.”

23.2 What not to do

24. Coding Qualitative Data in NVivo

NVivo can support storage, coding, retrieval, memoing and comparison. It does not decide which passages matter or what a theme means.

24.1 Recommended project setup

1. Create a project with a clear version name and date.
2. Import cleaned transcripts into a consistent folder structure.
3. Create cases for participants, organisations or documents where comparison is required.
4. Add attributes such as role, location or cohort only when ethically appropriate.
5. Create a memo folder for case memos, code memos and reflexive notes.
6. Begin with a provisional code structure rather than hundreds of empty nodes.

24.2 Coding a passage

1. Select the smallest contextually complete passage.
2. Code to an existing node or create a new provisional node.
3. Add a memo when the code requires explanation.
4. Use annotations for local comments tied to a specific phrase.
5. Review coded context periodically to ensure extracts have not been detached from meaning.

24.3 Nodes are not themes

NVivo uses “nodes” as containers for coded material. A node may represent an initial code, category, case or theme, depending on how the project is designed. Do not assume that a folder of nodes constitutes a finished thematic structure.

24.4 Useful NVivo functions during coding

• Highlight coding: visually inspect coded and uncoded sections.
• Coding stripes: view overlapping codes and density.
• Node summaries: record definitions and analytical notes.
• Memos and links: connect interpretation to evidence.
• Cases and classifications: support comparison by participant attributes.
• Text search: locate candidate passages, but never replace reading.
• Word frequency: assist familiarisation, not generate themes.

24.5 Version control

Keep dated backups before major restructuring. Record when nodes are merged, renamed or moved. For collaborative projects, agree how project files will be combined and avoid simultaneous editing of uncontrolled copies.

25. Team Coding and Coding Consistency

The purpose of team coding depends on the analytical tradition. In reflexive thematic analysis, multiple researchers can enrich interpretation by discussing different readings. In codebook analysis, they may refine shared definitions. In reliability-oriented analysis, agreement statistics may be part of the design.

25.1 Calibration exercise

Where a shared codebook is required, team members can code a small, varied sample, compare applications, revise definitions and repeat the exercise. The sample should include ambiguous passages rather than only obvious examples.

25.2 Avoiding false consensus

Do not erase analytically meaningful disagreement merely to produce a high agreement figure. First determine whether disagreement reflects unclear codes, different theoretical assumptions, missing context or genuinely plausible alternative interpretations.

26. Quality Checks During Coding

• Every code has a clear relationship to the research question.
• Code labels distinguish description from interpretation.
• Coded extracts retain enough context to preserve meaning.
• Contradictory and minority accounts are coded rather than ignored.
• Code definitions evolve transparently.
• Memos explain why important codes matter.
• Codes are compared across participants and contexts.
• The researcher revisits entire transcripts, not only retrieved extracts.
• Software outputs are treated as prompts for thinking, not findings.
• Reflexive notes document assumptions affecting coding decisions.

26.1 Negative cases

Actively code data that challenges an emerging interpretation. Negative cases may reveal boundary conditions, subgroup differences or a need to revise the developing account.

26.2 Context checks

At intervals, open coded extracts in their original source. Ask whether the interpretation still makes sense in relation to the participant’s wider account.

27. Common Coding Mistakes

1. Coding only keywords rather than meanings.
2. Using broad topic labels for the entire dataset.
3. Creating a new code for every sentence.
4. Using one code for several unrelated meanings.
5. Treating the interview schedule as the final code structure.
6. Ignoring data that does not fit expectations.
7. Deleting early codes without documenting why.
8. Merging codes based only on similar wording.
9. Creating deep NVivo hierarchies before understanding the data.
10. Assuming a frequently coded node is automatically important.
11. Interpreting participant claims as objective causal facts.
12. Applying emotion labels not supported by the extract.
13. Using In Vivo phrases because they sound vivid but lack analytical relevance.
14. Failing to distinguish cases from thematic nodes in NVivo.
15. Trying to achieve intercoder reliability in a reflexive design without justification.
16. Exporting coding reports instead of developing an analytical narrative.
17. Allowing AI or autocoding to determine the code structure.
18. Failing to return to uncoded context.
19. Overlooking silence, hesitation, contradiction or change over time.
20. Moving to themes before codes have been compared and refined.

28. Package 2 Coding Checklist

• I can explain what a code represents in my chosen analytical approach.
• I have defined an appropriate unit of coding.
• I know when overlapping coding is justified.
• I have documented whether coding is inductive, deductive, hybrid or abductive.
• My code names are specific and analytically useful.
• I distinguish descriptive, interpretive and conceptual codes.
• I maintain code definitions, inclusion rules and exclusions where needed.
• I write code and case memos during analysis.
• I compare codes across cases rather than coding each transcript in isolation.
• I preserve contradictory and minority accounts.
• I revisit original context after reviewing coded extracts.
• I can justify any use of multiple coders.
• I have a controlled NVivo backup and versioning process.
• I understand that nodes are containers, not automatically themes.
• I am ready to begin developing candidate patterns without treating code clusters as finished themes.

29. Moving from Codes to Themes

Theme development is the stage at which the analysis moves beyond a list of labelled extracts toward an organised explanation of patterned meaning. The researcher asks not only which codes occur together, but what broader idea makes their relationship analytically significant.

This phase is interpretive and recursive. Candidate themes are assembled, tested, dismantled, combined, separated and rewritten. Some codes will become central to a theme, some will provide contextual detail, some will move into another theme and some will not appear in the final account.

29.1 The analytical shift

29.2 Three common starting points

• Code-cluster review: examine groups of related codes and ask what shared meaning links them.
• Research-question review: inspect how different codes contribute to explaining each part of the research problem.
• Conceptual review: examine tensions, mechanisms, assumptions or processes that appear across multiple codes and cases.

29.3 What happens to unused codes?

Not every code must be promoted into a theme. Some codes remain useful as contextual information, exceptions, subthemes or supporting details. Others may be analytically weak, too isolated or outside the final scope. Retaining a record of excluded codes strengthens the audit trail.

30. Categories, Topics and Themes

Researchers often mistake categories for themes. Categories organise material by subject; themes interpret patterned meaning. A category can be a necessary intermediate step, but the analysis should not stop there if the study claims to offer interpretive thematic analysis.

30.1 Topic summary versus patterned meaning

30.2 A diagnostic test

Complete the sentence: “This theme shows that…” If the answer merely repeats the topic—“this theme shows that participants discussed training”—the theme is underdeveloped. A stronger answer offers an interpretive claim—“this theme shows that organisations transferred responsibility for capability development from the institution to individual employees.”

31. Generating Candidate Themes

Candidate themes are provisional analytical structures. They are hypotheses about how patterned meaning may be organised, not final findings. At this stage, researchers should be expansive enough to explore possibilities but disciplined enough to maintain a relationship with the research question.

31.1 Step-by-step process

1. Export or review the complete code list and code memos.
2. Identify codes that appear conceptually related rather than merely linguistically similar.
3. Lay codes out visually using paper, cards, a spreadsheet, NVivo maps or a whiteboard.
4. Create provisional clusters and write a one-sentence explanation for each.
5. Identify the central idea connecting the codes.
6. Search for extracts that do not fit the emerging cluster.
7. Compare the cluster across cases, contexts and participant groups.
8. Decide whether the cluster is a candidate theme, subtheme, contextual category or unsupported idea.
9. Return to the full dataset and test the emerging interpretation.

31.2 Code-clustering worksheet

31.3 Avoiding mechanical clustering

Codes should not be grouped solely because they share vocabulary. For example, lack of trust in management and management trusting experienced staff both contain the word “trust” but may contribute to different patterns. Conceptual relationship, context and function matter more than wording.

31.4 Candidate-theme memo

• Provisional title.
• Central organising concept.
• Relationship to the research question.
• Key codes and strongest extracts.
• Cases or groups represented.
• Contradictory evidence.
• Possible overlap with other themes.
• What remains unclear.

32. Finding the Central Organising Concept

The central organising concept is the core idea that gives a theme coherence. It explains why the codes belong together and what the theme contributes to the overall analytical story.

32.1 Weak and strong central concepts

32.2 Questions for identifying the core

• What is the theme fundamentally about?
• What shared meaning links the extracts?
• What does the theme reveal that a descriptive category would not?
• What mechanism, tension, assumption or process is operating?
• How does the theme help answer the research question?
• What would be lost if this theme were removed from the analysis?

33. Building Thematic Maps

A thematic map visually represents relationships between candidate themes, subthemes, codes and the overall research question. It is a thinking tool rather than decorative evidence of rigour.

33.1 Types of thematic map

33.2 Example relational structure

33.3 Questions to ask of the map

• Does each theme have a distinct role?
• Are some themes actually causes, contexts or consequences of others?
• Is one theme so broad that it contains the entire analysis?
• Are two themes making essentially the same claim?
• Does the map answer the research question or merely reproduce the interview schedule?
• Can the relationships be explained in prose?

33.4 NVivo and mapping

NVivo project maps and concept maps can help visualise relationships, but the researcher should export or record successive versions. The value lies in documenting how the analytical structure changed, not in producing a visually complex final diagram.

34. Reviewing and Developing Themes

Theme review usually operates at two levels: internal coherence within each theme and external coherence across the whole dataset.

34.1 Level one: review coded extracts

Read all extracts assigned to a candidate theme together. Ask whether they form a meaningful pattern. Remove extracts that do not fit, split the theme if it contains different ideas, or abandon it if no coherent concept remains.

34.2 Level two: review against the full dataset

Return to complete transcripts or documents. Check whether the candidate theme accurately represents the broader data context, whether relevant material was missed and whether the interpretation overstates a small number of vivid extracts.

34.3 Theme-review matrix

34.4 Merge, split, promote or demote

• Merge: when two candidate themes express the same central concept.
• Split: when one theme contains multiple coherent but distinct patterns.
• Promote: when a subtheme proves central to the overall explanation.
• Demote: when a theme is better treated as context, dimension or subtheme.
• Remove: when the pattern lacks relevance, coherence or evidence.

35. Establishing Theme Boundaries

A theme boundary defines what belongs inside the theme and what does not. Clear boundaries reduce repetition and make the final findings chapter easier to structure.

35.1 Boundary statement template

35.2 Boundary problems

35.3 How many themes?

There is no universally correct number. The final structure should be proportionate to the dataset, research question and publication format. A doctoral findings chapter may contain three to six substantial themes, sometimes with subthemes, but coherence matters more than a numerical target.

36. Negative, Deviant and Minority Cases

Negative or deviant cases challenge an emerging pattern. Minority cases may represent a smaller but meaningful variation. They should not be treated as inconvenient data to be hidden; they can refine the scope and credibility of themes.

36.1 Four analytical uses

• Boundary clarification: identify conditions under which a theme does not apply.
• Subgroup differentiation: reveal role, context or demographic variation.
• Mechanism refinement: show that the proposed explanation is incomplete.
• Alternative theme development: support a distinct patterned meaning.

36.2 Worked example

36.3 Do not quantify away exceptions

A statement such as “most participants agreed” may conceal important differences. Frequency can be reported where useful, but the analytical task is to explain why experiences vary and what the variation reveals.

37. Cross-Case and Within-Case Comparison

Thematic analysis identifies patterns across a dataset, but those patterns must remain grounded in cases. Cross-case comparison without within-case understanding can fragment participant accounts and remove context.

37.1 Within-case analysis

• Write a short case summary before or during coding.
• Record the participant’s overall position, context and internal contradictions.
• Note which themes are strongly, weakly or differently represented.
• Preserve temporal movement where relevant.

37.2 Cross-case analysis

37.3 Matrix analysis

A matrix can place themes in columns and cases or groups in rows. Cells contain concise summaries and evidence references. This helps identify similarities, differences and missing data without reducing analysis to counting.

37.4 Caution about subgroup claims

Do not claim that groups differ merely because one code appears more often. Consider sample composition, interview length, question routing, role relevance and whether the difference reflects meaning rather than opportunity to discuss the issue.

38. Themes, Subthemes and Analytical Hierarchy

Subthemes identify important dimensions of a broader central concept. They should clarify the structure of a theme rather than function as a storage system for every code.

38.1 Example hierarchy

38.2 When a subtheme is justified

• It expresses a distinct dimension of the parent theme.
• It has sufficient evidence and analytical importance.
• Its relationship to the central concept is clear.
• It improves the reader’s understanding of variation or process.

38.3 Flat versus hierarchical structures

A flat set of themes may be preferable when each theme makes a distinct contribution. A hierarchical structure is useful when several patterns operate under one broader concept. Avoid hierarchies that are more complex than the underlying argument.

39. Defining and Naming Final Themes

Defining a theme requires writing a concise analytical account of its scope, central concept, evidence base, relationship to other themes and contribution to the research question. Naming should communicate the insight rather than merely label the topic.

39.1 Theme-definition template

39.2 Naming principles

• Use a concise phrase that signals the analytical idea.
• Avoid one-word labels such as “leadership” or “barriers.”
• Avoid titles so literary that the meaning becomes obscure.
• Consider a vivid phrase followed by an explanatory subtitle.
• Ensure names are distinct from one another.
• Do not overstate causality or universality.

39.3 Naming examples

39.4 Final theme summary

Before writing the findings chapter, prepare a one-page summary for each theme containing its definition, subthemes, strongest extracts, negative cases, theoretical links and the exact claim it supports.

40. Worked Example: From Codes to a Final Theme

40.1 Research question

How do public-sector employees experience the implementation of a new digital case-management system?

40.2 Initial codes

• duplicate entry;
• after-hours data cleaning;
• avoiding criticism of director’s project;
• peer-created spreadsheet;
• informal troubleshooting;
• fear of appearing resistant;
• promised time savings;
• protecting clients from system delays.

40.3 Early category

Implementation problems and employee responses. This category organises the data but does not yet offer an interpretive claim.

40.4 Candidate themes

• Hidden labour of implementation.
• Political silence around digital failure.
• Informal systems maintaining service continuity.

40.5 Theme review

Review showed that the three candidate themes were tightly related. Hidden labour and informal systems were not separate experiences; both represented employees taking responsibility for resolving contradictions produced by official implementation. Political silence explained why this burden remained invisible.

40.6 Final theme

40.7 Possible subthemes

• Efficiency producing invisible labour: promised time savings generated duplication and after-hours work.
• Silence protecting strategic projects: staff withheld criticism to avoid being labelled resistant.
• Shadow systems preserving public service: informal tools and peer networks maintained continuity.

40.8 Evidence against overclaiming

Some experienced employees valued the new system’s audit trail. The final account should therefore distinguish between the burden of implementation and the longer-term value of traceability rather than portraying all effects as negative.

41. Package 3 Theme-Development Checklist

• I have distinguished codes, categories, topics and themes.
• Each candidate theme has a clear central organising concept.
• Themes answer the research question rather than reproduce interview headings.
• I have written candidate-theme memos.
• I have reviewed every theme at the level of coded extracts.
• I have returned to complete transcripts and the full dataset.
• Theme boundaries specify what is included and excluded.
• Overlapping themes have been merged or differentiated.
• Broad themes have been split where necessary.
• Weak patterns have been demoted, reframed or removed.
• Negative and minority cases have been examined.
• Cross-case differences are interpreted rather than merely counted.
• Subthemes add analytical structure and are not disguised code folders.
• A thematic map explains relationships among themes.
• Theme names communicate analytical insight.
• Each final theme has a concise written definition.
• I can explain how the themes changed during analysis.
• I have retained an audit trail of merges, splits and exclusions.
• The final structure forms a coherent overall analytical narrative.
• I am ready to write findings using themes as arguments rather than headings for quotations.

42. Planning the Thematic Findings Chapter

A thematic findings chapter should do more than display themes. It should present a coherent analytical argument showing what patterned meanings were developed, how they answer the research question and why they matter. The chapter must guide the reader from the overall analytical structure into each theme, while preserving sufficient evidence, context and variation.

42.1 Start with the chapter's analytical purpose

Before drafting, write one sentence explaining what the chapter as a whole demonstrates. This sentence should not merely say that the chapter “presents the findings.” It should identify the larger insight that connects the themes.

42.2 Recommended opening structure

1. Restate the analytical purpose and relevant research question.
2. Briefly identify the thematic-analysis approach used.
3. Introduce the final themes and explain how they relate.
4. Provide a thematic map or concise overview table where useful.
5. Explain any conventions used for participant identifiers, edited quotations or translations.
6. Signal how the chapter is organised.

42.3 Findings chapter architecture

43. Structuring Individual Theme Sections

Each theme section should establish the theme's central organising concept, demonstrate its pattern across the dataset and interpret its significance. A reader should understand what the theme means before encountering a long series of quotations.

43.1 A practical theme-section template

1. Theme claim: State the central analytical insight.
2. Definition and boundaries: Explain what the theme captures and what it does not.
3. Pattern: Describe how the theme appeared across the dataset.
4. Evidence: Present selected extracts from relevant cases.
5. Interpretation: Explain how the evidence supports the theme.
6. Variation: Address subgroup differences, contradictions or conditions.
7. Mini-synthesis: Close by stating how the theme answers the research question and connects to the next theme.

43.2 Topic sentence versus theme claim

43.3 Subthemes

Use subthemes only when they identify distinct dimensions of the central theme. A subtheme should deepen, qualify or explain the main theme rather than merely divide quotations into convenient topics.

44. Writing an Analytical Narrative

The analytical narrative is the researcher's explanation of what the evidence means. It connects quotations, cases and themes into an argument. Without this narrative, a findings chapter becomes a catalogue of participant comments.

44.1 The evidence–interpretation–significance sequence

44.2 Analytical verbs

Use verbs that communicate interpretation: constructs, normalises, legitimises, protects, obscures, reframes, redistributes, constrains, enables, signals, intensifies, negotiates, distances and reconciles. Avoid overstating claims by using causal verbs where the design supports only participant perceptions.

44.3 Keeping the participant and researcher voices distinct

Signal whether a statement is a participant's view, a pattern developed across accounts or the researcher's theoretical interpretation. Phrases such as “participants commonly framed,” “the pattern suggests,” and “this was interpreted as” help maintain transparency.

44.4 Avoiding repetition

Do not paraphrase a quotation and then repeat it in nearly identical words. Use commentary to reveal something the quotation cannot establish by itself: context, comparison, contradiction, mechanism or theoretical significance.

45. Selecting Strong Quotations

Quotations provide evidence and preserve participants' voices, but they should be selected analytically rather than decoratively. The strongest extract is not always the most dramatic one; it is the extract that clearly supports the claim while preserving context.

45.1 Selection criteria

• Relevance: The extract directly supports the analytical point.
• Clarity: The reader can understand it without excessive explanation.
• Context: Enough surrounding material is retained to avoid distortion.
• Typicality or strategic contrast: The quotation either illustrates a broader pattern or a meaningful exception.
• Economy: It is no longer than necessary.
• Ethical safety: It does not expose identity through distinctive details.
• Range: Evidence is not dominated by one especially articulate participant.

45.2 Representative does not mean statistically representative

In qualitative reporting, a “representative” quotation usually means that it illustrates a pattern found across relevant parts of the dataset. Do not imply statistical prevalence unless the study design and reporting support that claim.

45.3 Editing quotations

• Use ellipses only where removal does not alter meaning.
• Use square brackets for brief clarification.
• Correct minor speech disfluencies only under a stated convention.
• Retain distinctive grammar where it contributes to meaning, while avoiding presentation that caricatures participants.
• For translated data, explain whether quotations were translated before or after selection.

45.4 Short and block quotations

Integrate short extracts into sentences. Use block quotations for longer passages only when sequence, nuance or interaction is analytically necessary. Long blocks should be followed by substantial interpretation.

46. Integrating Quotations With Interpretation

A quotation should be introduced, presented and analysed. Avoid beginning a paragraph with an unexplained extract or ending immediately after a quotation without stating what the reader should notice.

46.1 The quotation sandwich

1. Set up: Introduce the analytical point and relevant context.
2. Evidence: Present the quotation.
3. Interpret: Explain specific language, assumptions, tensions or consequences.
4. Connect: Relate the extract to the wider pattern, another case or the research question.

46.2 Multiple quotations in one paragraph

Use several extracts only when the comparison adds analytical value. They may demonstrate recurrence, contrast roles, reveal changes over time or expose a contradiction. Do not stack quotations that make the same point without additional interpretation.

46.3 Quantifying language

47. Writing Cross-Theme Synthesis

Strong findings chapters show how themes form a larger analytical account. Themes may operate sequentially, hierarchically, cyclically, conditionally or in tension with one another.

47.1 Common relationships

47.2 Transitional paragraphs

End a theme by explaining what it establishes and why the next theme is necessary. This creates continuity instead of presenting isolated mini-essays.

48. Reporting Variation, Contradictions and Negative Cases

Thematic analysis identifies patterns, but those patterns should not erase difference. Reporting variation strengthens the analysis by showing its boundaries and conditions.

48.1 Forms of variation

• differences by role, location, career stage or organisational context;
• minority accounts that challenge the dominant pattern;
• changes within one participant's account;
• contradictions between stated beliefs and described actions;
• cases where an expected process did not occur;
• different meanings attached to the same event.

48.2 How to write a negative case

Do not append it as an afterthought. Explain whether it changes the theme's definition, reveals a boundary condition, represents a distinct subgroup or remains an unresolved exception.

48.3 Avoiding false homogeneity

Phrases such as “participants believed” can conceal substantial differences. Use qualified claims and identify the scope of each pattern.

49. Using Tables, Theme Maps and Figures

Visual displays can improve clarity but should not replace interpretation. Every table or figure must have an analytical purpose and be discussed in the text.

49.1 Useful displays

49.2 Example theme overview

50. Findings Chapter Versus Discussion Chapter

Universities and disciplines vary in whether findings and discussion are integrated. The important issue is analytical coherence, not a universal template.

50.1 What belongs in findings?

Theme definitions, evidence, within-dataset comparison and interpretation needed to establish the findings.

50.2 What belongs in discussion?

Extended comparison with prior research, theoretical contribution, explanation of mechanisms, implications, limitations and the study's wider significance.

51. Linking Findings to Literature and Theory

51.1 Three legitimate relationships with literature

• Confirmation: The finding supports or extends an established pattern.
• Contradiction: The finding challenges expectations or reveals a different context.
• Reframing: The finding changes how an existing concept should be understood.

51.2 Avoiding literature dumping

Do not follow every quotation with several citations. First establish the finding from the dataset, then use literature to clarify what is similar, different or newly explained.

51.3 Linking to a theoretical framework

Theory should sharpen interpretation, not force every extract into predetermined categories. Explain what the theory makes visible, where the data exceed it and whether the findings require modification of the framework.

51.4 Contribution language

Use proportionate claims: extends, refines, complicates, contextualises, connects, reveals a boundary condition or offers an alternative explanation. Avoid claiming that one qualitative study “proves” a universal theory.

52. Demonstrating Quality and Trustworthiness in the Written Chapter

Quality is demonstrated through the coherence of the argument, transparency of evidence and reflexive handling of interpretation. It cannot be secured by adding a checklist after analysis.

52.1 Evidence of quality

• clear alignment between research questions and themes;
• theme definitions with identifiable central concepts;
• quotations drawn from a suitable range of cases;
• explicit treatment of variation and contradictions;
• proportionate claims grounded in the dataset;
• transparent distinction between participant accounts and researcher interpretation;
• coherent links among themes;
• reflexive acknowledgement of how analysis was produced;
• an audit trail supporting major coding and theme decisions.

52.2 Thick description

Provide enough contextual detail for readers to judge transferability without exposing participants. Thick description concerns analytically relevant context, not indiscriminate detail.

52.3 Reflexivity in reporting

Where relevant, explain how the researcher's access, relationships, disciplinary assumptions or position influenced which meanings became visible and how competing interpretations were considered.

53. Common Findings-Writing Mistakes

1. Opening each theme with a quotation rather than an analytical claim.
2. Using themes as headings but writing only topic summaries.
3. Presenting quotations without interpretation.
4. Paraphrasing quotations without adding analysis.
5. Stacking several similar quotations to imply rigour.
6. Allowing one articulate participant to dominate the chapter.
7. Reporting frequencies as if qualitative prevalence proves importance.
8. Claiming that all participants shared a view when variation existed.
9. Ignoring negative cases or placing them in a final limitations paragraph.
10. Using subthemes that do not support a central concept.
11. Repeating the methods chapter instead of presenting findings.
12. Introducing new themes in the discussion that were not established in findings.
13. Forcing every theme to correspond to one research question.
14. Following the interview schedule rather than the analytical argument.
15. Using overly long quotations to reduce the need for explanation.
16. Removing so much context that quotations become ambiguous.
17. Including identifying detail in vivid extracts.
18. Overloading the chapter with literature before establishing the findings.
19. Claiming theoretical confirmation without explaining the relationship.
20. Ending theme sections without a synthesis or transition.

54. What Examiners Expect From Thematic Findings

54.1 Viva questions

• Why did you present the themes in this order?
• How did you decide which quotations to include?
• What evidence challenges your strongest theme?
• How do your themes differ from interview topics?
• What is the central organising concept of each theme?
• How does your interpretation go beyond participants' explicit statements?
• How did the theoretical framework influence, but not determine, your analysis?
• What is the most important cross-theme insight?

55. Worked Example: From Extracts to a Findings Paragraph

55.1 Weak descriptive version

55.2 Stronger analytical version

55.3 Why the stronger version works

• It begins with an analytical claim.
• The quotation is introduced with context.
• The commentary explains the significance of duplication.
• Evidence from other cases is synthesised without unnecessary quotation stacking.
• The paragraph returns to the theme's central organising concept.

55.4 Theme conclusion example

56. Package 4 Findings-Writing Checklist

• The chapter has a clear overall analytical argument.
• The order of themes is justified by analytical logic.
• Each theme begins with a central claim and definition.
• Subthemes deepen rather than fragment the main theme.
• Every quotation has a clear analytical purpose.
• Quotations retain enough context to preserve meaning.
• Participant evidence is not dominated by one case.
• Interpretation goes beyond paraphrase.
• Claims distinguish participant views from researcher interpretation.
• Variation and negative cases are integrated into theme sections.
• Cross-theme relationships are explained.
• Tables and figures support rather than replace analysis.
• The role of literature is appropriate to the chapter structure.
• Theory is used to sharpen, not force, interpretation.
• Claims about prevalence are proportionate.
• Ethical and anonymity risks in quotations have been checked.
• Theme conclusions explicitly answer the research question.
• The chapter conclusion synthesises rather than merely repeats.
• I can defend quotation selection and theme ordering in a viva.
• The findings provide a clear foundation for the discussion chapter.

57. The Complete NVivo Workflow for Thematic Analysis

NVivo is a computer-assisted qualitative data analysis system. Its value lies in organising complex datasets, preserving links between interpretations and source material, supporting systematic retrieval and documenting analytical decisions. It does not interpret data autonomously and should never be described as having “generated” themes.

57.1 What NVivo can do well

• store and organise transcripts, documents, field notes, audio, video and images;
• connect participant cases with ethically appropriate attributes;
• code passages to one or several nodes;
• retrieve all extracts linked to a code or theme;
• support within-case and cross-case comparison;
• record memos, annotations, links and project decisions;
• run text, coding, matrix and comparison queries;
• create maps and charts that assist analytical thinking;
• export structured evidence for writing and audit purposes.

57.2 What NVivo cannot establish

• whether a code is conceptually useful;
• whether a theme has a coherent central organising concept;
• whether an interpretation fits the study's epistemology;
• whether a quotation is ethically safe to publish;
• whether prevalence is analytically significant;
• whether a theoretical claim is warranted.

57.3 A defensible NVivo sequence

58. NVivo Project Setup and Data Architecture

58.1 Create a controlled project

1. Use a descriptive project name including version and date.
2. Store the working file in an approved secure location.
3. Create a separate backup folder and a backup schedule.
4. Record software version, operating system and team access arrangements.
5. Keep a plain-language project log outside NVivo so the project remains interpretable if software access changes.

58.2 Recommended source structure

58.3 Cases and classifications

A case represents an analytical entity, such as a participant or organisation. A classification describes the type of entity and permits attributes to be assigned. Code each participant's full contribution to the relevant case before relying on attribute-based comparisons.

58.4 Import checks

• All sources open correctly.
• Transcript names match participant identifiers.
• Speaker labels and paragraph structure are consistent.
• Case coding is complete.
• Attributes have valid and consistent values.
• Annotations and formatting have not exposed identities.
• The first backup has been created before coding begins.

59. Advanced Coding Workflow in NVivo

59.1 Code with context

Use coding stripes and highlighted coding to inspect how passages have been coded, but periodically open extracts in their full source. A retrieved sentence can appear to support a theme while its surrounding passage qualifies or contradicts that interpretation.

59.2 Node descriptions and memos

Every important node should have a concise definition. Complex or evolving nodes should also have a linked memo explaining analytical significance, boundaries, alternatives and changes over time.

59.3 Restructuring nodes

Before merging nodes, export or review their coded extracts. Record the old names, new name, rationale and date. Before splitting a node, define the distinction and recode all relevant extracts rather than only recent material.

59.4 Coding comparison

A coding comparison query may be appropriate in reliability-oriented or structured codebook studies. In reflexive thematic analysis it should not be used as proof that one interpretation is objectively correct. Divergent coding may be more useful as material for reflexive discussion.

59.5 Autocoding

NVivo may autocode by speaker, paragraph style, question or structured heading. This is useful for organisation. Autocoding by sentiment or pattern should be treated as exploratory and checked manually. It should not replace interpretive coding.

60. Queries and Comparisons in NVivo

60.1 Text search queries

Text search can identify explicit terminology, variations and possible missing passages. It cannot locate meanings expressed through different vocabulary. A search for “trust” will miss statements such as “I stopped checking her work.”

60.2 Word-frequency queries

Word-frequency outputs can support familiarisation and vocabulary exploration. Remove stop words, review stemming choices and return every important term to context. Frequency does not determine thematic importance.

60.3 Coding queries

Coding queries retrieve intersections, unions and exclusions among nodes and cases. They are useful for testing analytical propositions.

60.4 Matrix coding queries

A matrix can compare codes or themes across roles, sites, demographic categories or study waves. Interpret cells by reading the underlying extracts. Cell counts are navigational indicators, not statistical findings.

60.5 Compound and proximity queries

Where available, proximity queries help locate terms or codes occurring near one another. They can surface candidate passages but should not be interpreted without reading the source.

60.6 Query log

Save consequential queries with a clear name, date, purpose, parameters and brief interpretation. This allows the analytical path to be reviewed and repeated.

61. Maps, Charts and Visualisation

Visualisation should clarify relationships and stimulate analysis, not decorate the thesis. Every published figure requires a clear purpose, readable labels and explanation in the text.

61.1 Useful visual forms

• Project maps: show relationships among themes, subthemes, conditions and outcomes.
• Concept maps: develop theoretical or process explanations.
• Hierarchy charts: inspect code structure and volume cautiously.
• Comparison diagrams: explore overlap among sources, nodes or cases.
• Charts: display coding distributions for navigation and internal checking.

61.2 Theme maps

A strong thematic map communicates conceptual relationships rather than merely displaying a node hierarchy. Use directional arrows only where the relationship is genuinely directional. Label connections such as “enables,” “constrains,” “legitimises” or “results in.”

61.3 Export discipline

• remove confidential identifiers;
• use consistent terminology with the written chapter;
• include a figure title and explanatory note;
• retain an editable master version;
• verify readability in print and on mobile screens.

62. Memos, Audit Trails and Backups

62.1 Five memo types

62.2 Minimum audit-trail fields

• date and analytical stage;
• decision or question;
• evidence considered;
• alternative interpretations;
• action taken;
• effect on codes, themes or claims;
• link to relevant NVivo item or exported evidence.

62.3 Backup protocol

1. Create a dated backup before imports, major merges and theme restructuring.
2. Keep at least one backup outside the primary device.
3. Do not rely solely on automatic cloud synchronisation.
4. Periodically test that a backup opens correctly.
5. Retain the software version required to open archived projects.

63. AI-Assisted Thematic Analysis

Generative AI can assist with clerical, exploratory and critical-reflection tasks, but it introduces risks of confidentiality breach, hallucination, decontextualisation, automation bias and undisclosed analytical substitution. The researcher remains accountable for every code, theme and published claim.

63.1 Lower-risk uses

• improving non-sensitive code labels after the researcher supplies definitions;
• generating questions to challenge a candidate theme;
• suggesting alternative interpretations of anonymised short extracts;
• checking whether a theme definition is internally coherent;
• creating a draft audit-trail template;
• editing researcher-written prose for clarity without adding claims;
• identifying possible overlaps in a researcher-generated code list.

63.2 Higher-risk uses

• uploading identifiable transcripts to an unapproved public model;
• asking AI to produce final themes from the entire dataset;
• accepting summaries without checking source data;
• allowing AI to invent representative quotations;
• using AI sentiment or frequency as a proxy for interpretation;
• concealing material AI involvement from supervisors, journals or ethics bodies.

63.3 Human-verification protocol

1. Define the task and why AI is being used.
2. Confirm ethics, consent, institutional policy and data-processing arrangements.
3. Minimise and anonymise the input.
4. Record model, date, settings and prompt.
5. Treat output as a suggestion, not evidence.
6. Verify every suggestion against the original data.
7. Document accepted, modified and rejected suggestions.
8. Disclose substantive use where required.

64. Prompt Engineering for Thematic Analysis

Good prompts define the analytical role, boundaries, epistemological assumptions, required output and verification rule. They should ask AI to support researcher judgement rather than replace it.

64.1 Prompt structure

64.2 Prompt: challenge a candidate theme

64.3 Prompt: refine a theme name

64.4 Prompt: examiner simulation

64.5 Prompt: prose editing

65. Confidentiality, Ethics and Governance

65.1 Questions before using AI

• Does participant consent permit this form of processing?
• Does the ethics approval cover external AI services?
• Where will data be stored and processed?
• Will prompts or outputs be retained for model training?
• Is a data-processing agreement required?
• Can the task be completed with synthetic or highly minimised input?
• Does the university, funder or journal require disclosure?

65.2 Data minimisation

Remove names, locations, rare roles, dates and combinations of attributes that could permit re-identification. Anonymisation is contextual: a job title may identify a participant even when the name is removed.

65.3 Disclosure statement template

65.4 Authorship and accountability

AI systems cannot take responsibility for accuracy, ethics or interpretation and should not be listed as authors. Researchers must follow the current policies of their institution, funder, publisher and disciplinary body.

66. End-to-End Integrated Thematic Analysis Workflow

66.1 Claim-evidence table

Before submission, create a table linking each major claim to supporting extracts, relevant cases, contrary evidence, theory and limitations. This exposes unsupported generalisations before an examiner does.

66.2 Stop rules

Analysis is sufficiently developed when themes are coherent, distinct, well evidenced, relevant to the question and capable of sustaining a persuasive written argument. Stop rules should be based on analytical adequacy, not software counts or a desire to eliminate all ambiguity.

67. Publishing Thematic Analysis

67.1 Methods reporting

• name and justify the thematic-analysis approach;
• state philosophical and theoretical assumptions;
• describe dataset constitution and transcription;
• explain familiarisation, coding and theme development;
• clarify the role of multiple researchers;
• state what NVivo contributed;
• disclose substantive AI assistance;
• describe reflexivity and quality procedures.

67.2 Findings reporting

Journal word limits require selective evidence. Present the strongest themes, define each clearly, integrate short quotations and avoid replacing analysis with extensive procedural detail. Supplementary files may contain codebooks or extended tables when ethically and editorially appropriate.

67.3 Common reviewer concerns

67.4 Reproducibility versus transparency

Interpretive qualitative analysis is not made rigorous by promising that another researcher would produce identical themes. A stronger goal is methodological transparency: readers can understand how interpretations were produced, how alternatives were handled and why the claims are credible.

68. Final Integrated Quality Framework

68.1 Red-flag statements

• “NVivo identified the themes.”
• “The most frequent code became the main theme.”
• “Two coders agreed, therefore the analysis was unbiased.”
• “AI confirmed the findings.”
• “Themes emerged without researcher influence.”
• “All participants agreed,” when contrary accounts exist.

68.2 Stronger alternatives

• “NVivo supported coding, retrieval and comparison.”
• “Themes were developed through iterative interpretation of coded material and the full dataset.”
• “Analytical dialogue was used to challenge assumptions and explore alternatives.”
• “AI-generated suggestions were treated as prompts and verified against the data.”
• “Theme development was documented reflexively.”

69. Package 5 Final Checklist

• The full content from Packages 1–4 remains intact.
• My NVivo project has a clear folder, case and classification structure.
• Nodes have definitions and important nodes have analytical memos.
• Queries are saved with their purpose and interpretation.
• Matrix counts are never treated as statistical results.
• Visualisations communicate genuine conceptual relationships.
• Major node and theme changes are recorded in an audit trail.
• Backups are dated, separate and tested.
• AI use is ethically approved and institutionally permitted.
• No identifiable data are uploaded to unapproved systems.
• AI outputs are verified against original data.
• No AI-generated quotation or empirical claim is presented as evidence.
• Substantive AI assistance is disclosed where required.
• Each major finding can be traced through a claim-evidence table.
• Negative cases and contextual variation are integrated.
• Methods reporting names the thematic-analysis tradition clearly.
• The roles of NVivo, collaborators and AI are accurately distinguished.
• The final account demonstrates coherence, depth, grounding and reflexivity.
• My claims are proportionate to the dataset and design.
• I can defend the complete analytical workflow in a viva or peer review.

71. Practical Thematic-Analysis Toolkit

This section turns the earlier methodological guidance into reusable working tools. The templates are not intended to mechanise interpretation. Their purpose is to make analytical reasoning visible, traceable and easier to challenge before submission.

71.1 Minimum project file set

72. Analysis Planning Matrix

Complete this matrix before intensive coding. It forces alignment between the research question, analytical tradition and practical workflow.

72.1 Alignment test

Read the completed matrix horizontally. If the research question asks how meaning is constructed but the workflow focuses only on counting explicit topics, the design is misaligned. Resolve such contradictions before coding.

73. Codebook and Coding-Register Template

74. Theme-Development Template

74.1 Theme rejection record

Record rejected themes and the reason for rejection. Typical reasons include insufficient coherence, duplication, weak evidence, excessive breadth, topic-like structure or lack of relevance to the research question. This record demonstrates that alternatives were considered.

75. Claim-Evidence Matrix

75.1 Evidence density

A strong claim is usually supported by more than one vivid quotation. Check the underlying coded material, case distribution and contextual variation. One memorable extract can illustrate a claim, but it should not silently carry the entire argument.

76. Methods-Section Reporting Template

A defensible methods account normally answers the following questions in prose:

1. Why was thematic analysis suitable for the research question?
2. Which thematic-analysis tradition was followed?
3. What philosophical and theoretical assumptions shaped interpretation?
4. How was the dataset generated, selected, transcribed and prepared?
5. How did familiarisation occur?
6. How were codes developed and revised?
7. How were candidate themes generated, reviewed, defined and named?
8. How were reflexivity, contradictions and negative cases handled?
9. What roles did collaborators, NVivo and AI play?
10. What ethical and confidentiality safeguards were used?

76.1 Adaptable reporting model

“Data were analysed using reflexive thematic analysis. Analysis was primarily inductive and moved between semantic and latent levels. Following repeated familiarisation, the researcher generated initial codes across the dataset, recorded reflexive and analytical memos, and iteratively developed candidate themes around central organising concepts. Themes were reviewed against both coded extracts and the complete dataset, with attention to variation and disconfirming cases. NVivo supported data management, retrieval and comparison; interpretive decisions remained the responsibility of the researcher.”

77. Findings-Section Template

77.1 Opening the chapter

Begin by restating the analytical focus, introducing the final thematic structure and explaining how the chapter is organised. Avoid repeating the full methods process.

77.2 Reusable theme structure

1. Theme proposition: state the central claim.
2. Explanation: define the pattern and its significance.
3. Evidence: integrate carefully selected extracts.
4. Interpretation: explain how the evidence supports the claim.
5. Variation: show differences, conditions and negative cases.
6. Synthesis: connect the section to the wider argument.

77.3 Quotation commentary test

After every quotation, ask: what does this extract demonstrate, why does it matter, and how does it connect to the theme? If the surrounding prose merely repeats the participant’s words, the section is descriptive rather than analytical.

78. Viva and Examiner Defence

78.1 Ninety-second defence

Prepare a concise explanation covering the research question, chosen thematic-analysis tradition, coding orientation, theme-development logic, quality procedures and principal contribution. This helps reveal inconsistencies before the viva.

79. Troubleshooting Guide

80. Supervisor-Review Workflow

Do not send only a list of themes. Provide a focused review pack containing:

• the research question and analytical approach;
• a one-page thematic map;
• definitions and central concepts for each theme;
• one strong and one challenging extract per theme;
• the claim-evidence matrix;
• a short list of unresolved decisions;
• a sample written theme section.

80.1 Useful review questions

1. Which theme appears least coherent, and why?
2. Where do claims exceed the evidence?
3. Which boundaries are unclear?
4. What alternative interpretation should be considered?
5. Is the analytical contribution visible?
6. Which sections remain descriptive?

Specific questions produce more useful feedback than asking whether the themes “look right.”

81. Final Submission-Readiness Audit

81.1 Final red-team review

Read the work as a sceptical examiner. Mark every sentence that asserts “participants,” “the data,” “the theme” or “the findings” and ask whether the wording is proportionate. Then identify the weakest theme and try to disprove it using the dataset. Revise before submission.

• Braun, V. and Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101.
• Braun, V. and Clarke, V. (2012). Thematic analysis. In H. Cooper et al. (eds.), APA Handbook of Research Methods in Psychology.
• Braun, V. and Clarke, V. (2019). Reflecting on reflexive thematic analysis. Qualitative Research in Sport, Exercise and Health, 11(4), 589–597.
• Braun, V. and Clarke, V. (2021). Thematic Analysis: A Practical Guide. London: SAGE.
• Boyatzis, R. E. (1998). Transforming Qualitative Information: Thematic Analysis and Code Development. SAGE.
• Guest, G., MacQueen, K. M. and Namey, E. E. (2012). Applied Thematic Analysis. SAGE.
• King, N. and Brooks, J. (2018). Template Analysis for Business and Management Students. SAGE.
• Nowell, L. S., Norris, J. M., White, D. E. and Moules, N. J. (2017). Thematic analysis: Striving to meet the trustworthiness criteria. International Journal of Qualitative Methods, 16, 1–13.
• Terry, G., Hayfield, N., Clarke, V. and Braun, V. (2017). Thematic analysis. In C. Willig and W. Stainton Rogers (eds.), The SAGE Handbook of Qualitative Research in Psychology.