Thematic Analysis: From Coding Decisions to Defensible Themes

Reader outcome

This guide helps a qualitative researcher move from raw material to a defensible set of themes while making analytic choices visible. It treats thematic analysis as a flexible family of practices rather than a button in qualitative software or a mechanical sequence that guarantees insight.

What thematic analysis is

Thematic analysis is used to identify, analyse and communicate patterned meaning in qualitative material. Braun and Clarke’s influential account presented it as accessible and theoretically flexible. Flexibility is a strength only when the researcher explains the assumptions and decisions that shape the analysis.

A theme is not simply a frequently mentioned topic. It is a coherent pattern of meaning that helps answer the research question. Frequency can matter, but a less common pattern can be analytically important when it reveals a mechanism, contradiction, institutional condition or marginalised experience.

Decide the analytic orientation

Before coding, state the orientation of the study. Important decisions include:

• whether coding is primarily inductive from the material or guided by existing theory;
• whether the analysis focuses on semantic content or more latent assumptions and meanings;
• whether themes are understood as discovered features, researcher-developed interpretations, or a combination framed by the study’s epistemology;
• whether the purpose is experiential description, critical interpretation, evaluation, design improvement or theory development.

These are not decorative philosophy statements. They affect what counts as evidence, how codes are written and how claims are justified.

Build an auditable workflow

1. Familiarise yourself with the material

Read and re-read transcripts, field notes, documents or open responses. Correct obvious transcription errors without erasing meaningful speech. Write early analytic notes: surprises, tensions, recurring images, contradictions and questions. Familiarisation is not passive reading; it begins interpretation.

2. Generate initial codes

Code segments that are relevant to the research question. A code should capture why a passage matters, not merely repeat its first noun. Compare a descriptive code such as “lack of information” with a more analytic code such as “institutional uncertainty shifts planning risk onto the student.” Both can be useful, but they perform different work.

Code consistently across the dataset while allowing the codebook to evolve. Preserve a change log showing merged, renamed and retired codes. In reflexive forms of thematic analysis, coding does not need to imitate inter-rater reliability procedures; in team or codebook approaches, agreement and adjudication may be relevant. The chosen quality logic must match the declared approach.

3. Develop candidate themes

Group codes into broader patterns that address the research question. Use maps, tables or memos to test relationships between central organising concepts, subthemes and exceptions. A candidate theme needs an argument, not just a container of quotations.

4. Review and refine

Check candidate themes against the coded extracts and the wider dataset. Ask whether the material within a theme is coherent, whether themes are sufficiently distinct, what evidence challenges the interpretation, and whether the set of themes answers the research question. Split, merge or discard candidates when necessary.

5. Define and name themes

Write a short analytic statement for each theme: what pattern does it capture, why does it matter, what is its boundary, and how does it contribute to the overall account? Name the theme so that it communicates the argument. A title such as “Support” is weak; “Support is available only after students demonstrate crisis” conveys a claim that can be examined.

6. Construct the report

The final report should interweave analytic explanation, selected evidence and connections to the research question and literature. Quotations illustrate and support the interpretation; they do not analyse themselves. Include variation and negative cases when they refine the claim.

These phases are recursive. Researchers often return to the data, revise codes and redraw themes while writing. The process should be systematic without pretending to be perfectly linear.

Mini worked example

Research question: “How do international master’s students experience uncertainty during the first semester?”

Possible extracts and initial codes:

• “Every office sent me to another office.” → responsibility is circulated; support pathway is opaque.
• “I stopped asking because I felt I should already know.” → self-silencing; uncertainty becomes personal failure.
• “The student group explained the process in one evening.” → peer networks repair institutional gaps.

A weak topic summary might be “Problems with information.” A stronger candidate theme could be “Students privatise institutional uncertainty until peers make the system legible.” The theme connects the extracts into an interpretive pattern and can be tested across the dataset. It would need contrary cases, boundary conditions and evidence about how the pattern varies.

Using MAXQDA or NVivo

Software can store material, apply and retrieve codes, support memos, visualise relationships and preserve an audit trail. It does not decide what a theme means or whether the interpretation is coherent. A large code-frequency chart is not a substitute for analysis. Export and archive the code system, memos and version history needed to reconstruct decisions.

Demonstrating quality

A defensible report usually makes the following visible:

• alignment between research question, epistemological position and analytic approach;
• a clear account of how material was selected, prepared and coded;
• reflexive consideration of the researcher’s role and assumptions;
• evidence that themes were reviewed against the whole dataset;
• attention to contradictory or deviant material;
• enough evidence for readers to understand the basis of each claim;
• a transparent boundary between participant meaning, researcher interpretation and wider theoretical inference.

Avoid importing quality criteria from another method without explanation. Member reflection, multiple coders, saturation or inter-rater statistics may be useful in some designs but are not universal requirements for every form of thematic analysis.

Common mistakes

• Presenting interview questions as themes.
• Listing topics without an organising analytic idea.
• Treating six phases as a checklist completed once in order.
• Reporting only code frequencies.
• Using many quotations with little interpretation.
• Claiming themes “emerged” while hiding the researcher’s decisions.
• Mixing incompatible forms of thematic analysis without acknowledging the change in assumptions.
• Allowing software outputs to determine substantive conclusions.

Source and verification notes

The conceptual foundation is the primary Braun and Clarke article on thematic analysis in psychology. This guide paraphrases the method and adds an original workflow, example and quality-control framework. Researchers should state the specific form and orientation of thematic analysis they intend to use so that the quality criteria remain coherent.