Completing the coding process in NVivo is an important step in qualitative research, but it does not automatically mean that our analysis is complete. After coding interviews, focus groups, or open-ended responses, we still need to examine relationships between codes, interpret patterns across the dataset, and develop themes that meaningfully address our research questions.
Moving from NVivo codes to themes requires more than grouping similar labels together. We need to consider the meaning of participants’ experiences, the context of their responses, our analytical approach, and the evidence supporting our interpretations.
A defensible qualitative analysis should demonstrate how we moved from the original data to initial codes, candidate themes, and final interpretations. NVivo can help organise and explore this process, but the researcher remains responsible for making and justifying analytical decisions.
In this guide, we explain a practical workflow for developing, reviewing, and reporting qualitative themes using NVivo while maintaining methodological transparency and researcher involvement.
Choose an Analytical Approach Before Selecting Coding Features
Before we begin developing themes from NVivo codes, we should clarify the qualitative methodology and analytical approach guiding our study. The way we create, organise, review, and interpret codes depends on the approach we have selected.
For example, reflexive thematic analysis focuses on developing patterns of shared meaning through the researcher’s active interpretive engagement with the dataset. In contrast, codebook-oriented approaches may use structured coding frameworks, defined code descriptions, and systematic comparisons across researchers or cases.
These approaches do not use identical assumptions about coding and theme development. In reflexive thematic analysis, researcher subjectivity is recognised as part of the analytical process, and coding agreement is not treated as a universal measure of quality. In codebook-oriented approaches, a documented framework can help guide coding decisions and support a more structured analytical workflow.
NVivo can support different qualitative analysis workflows, but its software features do not determine our methodology. We should select analytical procedures according to our research questions, methodological assumptions, and study design rather than simply following the features available in the software.
Before coding or developing themes, we should be able to explain:
What is the main purpose of our qualitative study?
Which analytical approach are we following?
What assumptions guide our interpretation of the data?
How will we move from coded extracts to themes?
What evidence will we use to support our interpretations?
Our answers provide the methodological foundation for the subsequent analysis.
Step 1: Revisit the Original Data and Become Familiar with Coded Extracts
Once we have coded our qualitative data, we should revisit the original transcripts and become familiar with the context surrounding coded extracts.
This is important because individual codes may capture only part of a participant’s experience or explanation. An extract may appear to represent one idea when viewed separately, but its surrounding conversation may reveal a different or more complex meaning.
We can use NVivo to review coded references, compare extracts, organise code hierarchies, and retrieve relevant sections of transcripts.
During this process, we should consider:
What meaning does each code capture?
Are different codes describing similar experiences or distinct ideas?
Do some codes appear to be connected through a broader concept?
Are there contradictions or unusual experiences within the dataset?
How do these observations relate to our research question?
Have we overlooked important context surrounding the coded extracts?
We should also revisit the complete transcript when necessary rather than relying exclusively on isolated coded segments.
Why Analytical Memos Matter
Memos are particularly useful for documenting why we are combining, separating, redefining, or retaining codes.
For example, we may initially have two codes, such as “Limited Supervisor Feedback” and “Unclear Research Guidance”. After reviewing the extracts, we may identify a broader pattern concerning students’ experiences of inconsistent academic support.
An analytical memo can record:
The codes being examined.
The extracts or observations that prompted the review.
The possible relationship between the codes.
Alternative interpretations.
The reason for retaining, combining, or separating the codes.
Questions that require further examination.
Memos allow us to record analytical decisions rather than relying on memory when we develop or revise themes.
Step 2: Explore Relationships Between Codes and Identify Patterns
After reviewing the coded data, we can begin exploring relationships between codes and identifying possible patterns across the dataset.
At this stage, we should avoid treating every group of related codes as a theme. Instead, we need to examine what the codes collectively reveal about the research question.
For example, codes such as:
Fear of making mistakes
Checking work repeatedly
Seeking supervisor approval
Difficulty making independent decisions
may initially appear to describe separate experiences.
However, when we examine the underlying meaning of the extracts, they may collectively suggest a broader pattern involving uncertainty and dependence on external reassurance.
We should ask whether these codes reflect a shared idea, how the participants describe their experiences, and whether the proposed relationship is supported by the wider dataset.
NVivo can assist with this exploration through:
Reviewing coded references across different nodes.
Examining relationships between codes.
Comparing coded material across cases.
Using matrix coding queries where appropriate.
Organising related codes into parent and child structures.
Reviewing analytical memos alongside coded extracts.
These features help us organise and examine the data. They do not independently establish the meaning of a pattern.
Step 3: Develop Candidate Themes From the Coded Data
Once we have explored relationships between codes, we can begin developing candidate themes.
A theme should represent a meaningful pattern relevant to our research question, rather than simply act as a heading for several related topics.
For example, consider a qualitative study investigating postgraduate students’ experiences of using generative AI during academic research.
The following example is entirely hypothetical and created for demonstration purposes. It does not represent actual participant data or findings from a completed research project.
Research question: How do postgraduate students experience the use of generative AI during academic research?
| Illustrative participant extract | Initial code | Possible analytical direction |
|---|---|---|
| “I use it to understand difficult papers before reading them properly.” | AI-assisted understanding | AI as an initial learning resource |
| “I still verify the references because sometimes the information is incorrect.” | Verification of AI output | Trust requires independent checking |
| “It saves time, but I don’t want to rely on it for my own arguments.” | Efficiency versus independent thinking | Balancing assistance with research autonomy |
| “I ask it to explain concepts, but the final interpretation is mine.” | Retaining interpretive responsibility | Researcher ownership of interpretation |
These codes may support a candidate theme such as:
Using AI for support while retaining research ownership
However, we should not automatically finalise this theme because the codes appear related. We need to revisit the original extracts, examine their context, consider contradictory experiences, and determine whether they form a coherent pattern of shared meaning relevant to our research question.
What Makes a Candidate Theme Meaningful?
Before accepting a candidate theme, we should examine the following:
Relevance: Does the theme address an important aspect of our research question?
Coherence: Do the supporting extracts contribute to a meaningful central idea?
Distinctiveness: Is the theme sufficiently different from other candidate themes?
Evidence: Can we support the interpretation with relevant participant data?
Analytical depth: Does the theme explain something meaningful rather than simply describe a topic?
Methodological alignment: Does the theme reflect the assumptions of our selected analytical approach?
A strong theme should communicate what the data mean in relation to the study, not simply identify what participants discussed.
Step 4: Review, Refine, Merge, or Separate Candidate Themes
Candidate themes should remain provisional until we have examined their supporting evidence and relationship with the complete dataset.
As our analysis develops, we may:
Merge themes that represent closely related patterns of meaning.
Separate themes that contain distinct analytical ideas.
Rename themes to communicate their central meaning more clearly.
Redefine theme boundaries.
Remove candidate themes that lack sufficient analytical support.
Develop subthemes where they provide a meaningful structure.
For example, a researcher investigating academic supervision may initially identify two candidate themes:
Challenges in receiving academic guidance.
Difficulties making independent research decisions.
After reviewing the extracts, the researcher may determine that these themes represent different aspects of a broader pattern involving the development of research autonomy.
Alternatively, the researcher may find that the two patterns are analytically distinct and should remain separate.
The decision should depend on the meaning of the data and the research question, not simply on the convenience of reducing the number of themes.
Candidate Theme Review Table
We can use a table like the following to document theme development:
| Review question | What we should examine |
|---|---|
| Does the theme have a clear central idea? | Whether the theme communicates a meaningful pattern |
| Are the supporting extracts relevant? | Whether the evidence supports the proposed interpretation |
| Does the theme overlap with another? | Whether the themes represent distinct analytical ideas |
| Are contradictory accounts considered? | Whether alternative or divergent experiences have been examined |
| Is the theme relevant to the research question? | Whether it contributes to answering the study’s objectives |
| Is the theme consistent with the methodology? | Whether its development follows the selected analytical approach |
This review helps us explain how candidate themes were evaluated and why particular analytical decisions were made.
Step 5: Use NVivo Matrices, Cases, and Quotations to Explore the Evidence
NVivo provides several features that can help us explore qualitative data, but their outputs require careful interpretation.
1. Matrix Coding Queries
Matrix coding queries can help us compare coded material across cases, participant groups, or other analytical categories.
For example, we may explore how experiences of AI use differ across postgraduate students with varying levels of research experience.
A matrix may reveal where coded material occurs across different groups. We should then revisit the relevant extracts and interpret the observed patterns in relation to the research question.
2. Cases and Attributes
Cases and attributes can support structured comparisons between participants, organisations, or other units of analysis when our research design includes relevant characteristics.
For example, we may compare interview responses by academic level, research experience, or another relevant participant characteristic.
These comparisons should be guided by the research design. We should not introduce unnecessary demographic or participant categories simply because NVivo allows us to organise them.
3. Coded Quotations and References
Coded quotations and references allow us to revisit the actual evidence behind a code or candidate theme.
This is essential for assessing whether our interpretation accurately reflects participants’ accounts.
When selecting quotations for reporting, we should consider:
Whether the extract directly supports the analytical point.
Whether sufficient context is provided.
Whether the quotation represents the participant’s meaning accurately.
Whether contradictory or alternative experiences need to be discussed.
Whether participant confidentiality is protected.
What NVivo Outputs Cannot Establish on Their Own
A matrix, chart, code count, or cluster visualisation does not independently establish the meaning or significance of a theme.
We should use these outputs as analytical aids, then interpret them in relation to the research question, dataset, context, and methodological approach.
For example, a matrix may show that a particular code appears more frequently among one participant group. This observation may provide a useful starting point for further examination, but it does not automatically explain why the difference exists or what it means.
The researcher must provide that interpretation using the available qualitative evidence.
Step 6: Maintain Reflexivity and an Analytical Audit Trail
Qualitative analysis involves researcher interpretation. We should therefore consider how our assumptions, academic background, experiences, and theoretical perspectives influence coding and theme development.
Reflexivity helps us examine these influences and make our analytical decisions more transparent.
We can maintain an audit trail that records:
Changes to codes and code definitions.
Development and revision of candidate themes.
Decisions to merge or separate codes.
Changes to theme boundaries and names.
Analytical memo development.
Consideration of contradictory evidence.
Reasons behind important revisions.
Methodological decisions that influenced interpretation.
The specific role of reflexivity and documentation depends on our selected analytical approach.
For example, in reflexive thematic analysis, reflexivity is central to how researchers understand their involvement in knowledge production. The purpose is not to eliminate subjectivity or demonstrate that all researchers would produce identical codes.
In codebook-oriented approaches, documenting code definitions and changes can help explain how the coding framework was developed and applied.
In both cases, we should be able to communicate how our analytical decisions relate to our methodology.
Limitations of AI-Assisted Features
If we use AI-assisted features for summarisation, coding suggestions, or pattern exploration, we should evaluate their outputs rather than accept them automatically.
AI-generated summaries may overlook:
Context surrounding participant statements.
Contradictory evidence.
Nuanced or implicit meanings.
Differences between participants’ experiences.
Theoretical assumptions influencing interpretation.
Specific language used by participants.
We should also consider confidentiality, data protection, institutional policies, and whether our dataset can be processed using the relevant AI features.
AI-assisted outputs should not replace our methodological reasoning or serve as independent evidence for qualitative findings. Any use of these features should be documented where required by the research institution or publication guidelines.
Step 7: Present Qualitative Findings Without Treating Frequency as Importance
When reporting qualitative findings, we should focus on what our themes explain about the research question rather than simply how frequently individual codes appear.
A frequently occurring code may describe a common experience, but a less frequent observation may reveal a meaningful contradiction, an important exception, or a deeper insight.
Therefore, we should not assume that the number of coded references determines the significance of a theme.
We can present our qualitative findings using a combination of:
Clearly defined themes and subthemes.
Analytical explanations of what each theme represents.
Relevant participant quotations.
Contextual interpretation of the quotations.
Relationships between themes and research questions.
Contradictory or divergent experiences where relevant.
For example, rather than reporting that a theme contains 45 coded references, we should explain what the theme represents, how participants’ experiences illustrate it, and why it matters in relation to our research objectives.
Example of Reporting a Qualitative Theme
Consider the hypothetical theme:
Using AI for support while retaining research ownership
A weak report might simply state that participants frequently used AI to understand research concepts.
A more analytical report would explain that participants described AI as a resource for simplifying complex academic material while maintaining responsibility for verifying information and developing their own interpretations.
Relevant participant quotations could then illustrate how this pattern appeared across the dataset, including any differences or contradictory experiences.
This approach connects the theme with the evidence and explains its relevance to the research question.
A Worked Memo Excerpt: Demonstrating Theme Development
The following analytical memo is fictional and provided only to demonstrate how a researcher might document theme development.
Memo title: Reviewing AI-assisted research practices
Date: Illustrative
Analytical focus: Relationships between AI-assisted understanding, output verification, independent thinking, and researcher responsibility.
Memo excerpt:
“Several coded extracts appear to describe AI as a supportive resource rather than a replacement for the researcher’s judgement. Participants’ accounts of using AI to simplify complex concepts are connected with their descriptions of verifying information and retaining responsibility for their interpretations.
However, these experiences should not be treated as identical. Some participants may prioritise efficiency, while others may be more concerned about accuracy or academic independence.
I should review the relevant extracts together, examine whether these differences represent distinct patterns, and determine whether the candidate theme captures a coherent meaning across the dataset. I should also review accounts that challenge the idea of AI as a supportive resource before finalising the theme.”
This memo demonstrates how we can document an analytical idea, identify relationships between codes, consider alternative interpretations, and record the next steps required before finalising a theme.
Model-Aware Reporting Checklist
Before presenting our NVivo-based qualitative findings, we should review the following checklist:
Have we selected and clearly described our qualitative analytical approach?
Are our coding decisions consistent with our methodology?
Have we reviewed coded extracts in their original context?
Do our themes represent meaningful patterns rather than collections of topics?
Have we reviewed candidate themes for coherence and distinction?
Have we considered contradictory or divergent evidence?
Have we documented important analytical decisions through memos or an audit trail?
Are our interpretations supported by participant data?
Have we avoided treating code frequency as a direct measure of theme importance?
Have we appropriately documented software and AI-assisted features, where used?
Are our findings aligned with our research questions and objectives?
A defensible qualitative analysis is not defined by the number of codes, themes, charts, or NVivo outputs we generate. It depends on how clearly we explain our analytical approach, interpret the evidence, and demonstrate the connection between participant data and our findings.
How Research10X Supports NVivo-Based Qualitative Analysis
Moving from coded qualitative data to defensible themes can be challenging, particularly when our dataset contains complex participant experiences, overlapping codes, contrasting perspectives, or methodological requirements.
At Research10X, we understand that qualitative analysis involves more than organising transcripts or generating coding outputs. Researchers also need to interpret their data, justify their analytical decisions, and present findings that align with their research objectives.
Through our NVivo qualitative data analysis services, we support researchers with:
Qualitative data organisation and coding.
Codebook development and refinement.
Thematic analysis and candidate theme development.
Pattern exploration and interpretation of coded extracts.
NVivo project organisation and coding workflows.
Qualitative findings interpretation and reporting support.
Methodological alignment and analytical documentation.
We also provide research methodology consulting services to support decisions concerning qualitative research design, analytical approach selection, and methodological justification.
Our focus is on helping researchers understand the analytical reasoning behind their findings rather than simply producing software outputs.
Develop Meaningful Themes From Your NVivo Analysis
Moving from NVivo codes to themes requires careful interpretation, methodological consistency, and a clear connection between participant data and research questions.
Researchers should review coded extracts in context, explore relationships between codes, develop provisional themes, examine contradictory evidence, and document important analytical decisions.
NVivo can make qualitative data organisation and exploration more manageable, but meaningful themes still depend on the researcher’s analytical judgement and the methodological approach guiding the study.
At Research10X, we encourage a research-focused approach to qualitative analysis, where software supports the analytical process and the interpretation remains grounded in the data, research objectives, and methodological assumptions.
Frequently Asked Questions
1. How do we develop themes from codes in NVivo?
2. Does NVivo automatically generate themes from codes?
We can use NVivo to organise codes, explore patterns, run queries, create visualisations, and access AI-assisted features where available. However, we do not treat software-generated suggestions or groupings as final analytical themes without methodological review and interpretation.
3. What is the difference between a code and a theme?
We use codes to label meaningful segments of qualitative data. We develop themes to represent broader patterns of meaning in relation to our research question. A theme should communicate an analytical idea rather than simply group several related codes under a common heading.
4. Should we use code frequency to determine themes?
We do not rely on code frequency alone to determine whether a theme is important. We consider its relevance to the research question, contextual meaning, patterns across the dataset, contradictory evidence, and the analytical approach guiding our study.
5. Can we merge or split codes during thematic analysis?
Yes. We may merge codes when they capture closely related meanings, split codes when they represent distinct concepts, or revise code definitions as our analysis develops. We document important changes and ensure that our decisions remain consistent with the selected methodology.
6. Can we use NVivo for reflexive thematic analysis?
Yes. We can use NVivo to organise qualitative data, manage coded extracts, review transcripts, and support memoing during reflexive thematic analysis. However, we ensure that our use of coding structures, code comparisons, and software features remains consistent with the assumptions and principles of reflexive thematic analysis.
7. How do we report NVivo thematic analysis in a thesis?
We explain our analytical approach, familiarisation process, coding procedures, theme development, review process, and interpretation of findings. We present themes using analytical explanations and relevant participant quotations while maintaining transparency about our use of NVivo and any AI-assisted features.
8. How many codes should we have before developing themes?
We do not follow a fixed number of codes before beginning theme development. The appropriate number depends on our research question, dataset, analytical approach, and the level of detail required. We focus on the meaning and relevance of the codes rather than trying to reach a predetermined count.
9. What is an analytical memo in NVivo?
We use analytical memos to record our observations, interpretations, questions, and decisions during qualitative analysis. They help us document relationships between codes, explain changes to candidate themes, consider alternative interpretations, and maintain a transparent record of our analytical reasoning.
10. What makes a qualitative theme defensible?
We consider a theme defensible when its central idea is clear, its interpretation is relevant to the research question, and its supporting evidence is examined in context. We also explain how the theme was developed, consider contradictory evidence, and demonstrate alignment with our selected analytical approach.



