# Thematic Analysis: Codes vs Themes, With a Worked Example

See a thematic analysis example that distinguishes codes from themes and links an interpretation to extracts and a counterexample.

- Canonical URL: https://www.harmate.com/en/blog/thematic-analysis-codes-vs-themes-with-a-worked-example
- Author: Harmate Team
- Published: 2026-10-04
- Updated: 2026-10-22T13:22:34.773755+00:00
- Language: en

## Content

# Thematic Analysis: Codes vs Themes, With a Worked Example

Thematic analysis is not a count of recurring words. It builds an argued answer to a research question from patterns of meaning in a dataset and keeps the relationship between interpretation and source extracts visible. A code labels an idea in a passage; a theme offers a coherent way to explain a pattern that matters across the material.

There are multiple approaches to thematic analysis. This article focuses specifically on Virginia Braun and Victoria Clarke’s reflexive thematic analysis. Their [phases](https://www.thematicanalysis.net/doing-reflexive-ta/) are flexible, recursive guides, not a recipe that guarantees a correct interpretation. Other approaches organize the development and checking of themes differently. Mixing them without saying so—for example, claiming a reflexive approach while requiring an inter-coder agreement score—makes the method harder to understand.

## When does thematic analysis fit the question?

It can be useful when a research question concerns meanings, experiences, or ways of describing a situation across qualitative material such as interviews, open-ended responses, diaries, or documents. It can help examine how people describe a difficulty or what patterns shape their experiences. It does not turn those accounts into a representative measure of a whole population or automatic proof of cause.

Before coding, write down the question the analysis is meant to answer and define the material in scope. “What do participants say?” is too broad. “How do participants describe the conditions that help them reuse a practice after training?” identifies a phenomenon and context without deciding the themes in advance.

An analysis is not better because it uses software, counts more codes, or contains more tables. Its quality depends on the fit among the question, the chosen method, engagement with the data, and the argument presented to the reader.

If your question is less about a pattern across a dataset and more about how meaning is built in an exchange—through turns, repetitions, pronouns, or modality—[discourse analysis](/en/blog/discourse-analysis-read-an-exchange-without-guessing-intentions) offers a different lens, with its own limits. It is distinct from reflexive thematic analysis.

If you are still planning data collection, the [semi-structured interview guide](/en/blog/semi-structured-interview-guide-method-steps-and-15-examples) includes an adaptable guide, neutral follow-up questions, and a worked example of notes to retain for analysis. Plan the questions and preserve their context before coding.

## Spend time with the dataset before naming themes

In reflexive thematic analysis, familiarization is analytic work, not a formality before the “real” analysis. Read and reread the material. Note what surprises you, what you expected to find, the words people use, and the contexts that may change their meaning.

Fictional, simplified example: a training team asks participants a few weeks after a workshop whether they reused a retrospective practice. One person writes, “I tried, but the meeting was already full.” Another writes, “We brought the question into the team check-in because we had twenty minutes.” Do not yet conclude that lack of time is the general cause, or that both people describe the same situation. Keep their wording, response timing, and any context the research design permits.

A familiarization memo can separate three things:

| Note | Fictional example | Status |
|---|---|---|
| Wording | “the meeting was already full” | dataset excerpt |
| First reading | available time may matter | provisional interpretation |
| Question to check | do other accounts describe a comparable context? | analytic direction |

Keeping these apart makes it easier to revise an early reading and prevents an interpretation from being presented as though it came directly from the text.

## A code is not a theme

A code is an analytic label applied to a passage considered relevant to the question. Depending on the chosen approach, it may stay close to the participant’s wording or offer a more interpretive reading. A theme is more developed: in Braun and Clarke’s reflexive approach, it organizes a pattern of meaning around a central idea. The authors distinguish [codes from themes](https://www.thematicanalysis.net/faqs/); a list of topics—“time,” “tools,” “managers”—may summarize what the material mentions without explaining what the pattern means for the question.

In the fictional example, “not enough time” could be an initial code. After rereading, the analyst might explore a provisional theme such as “a practice stays peripheral until work makes room for it.” That theme advances an interpretation by linking different accounts through an organizing idea. Its elegance alone does not validate it; it must be examined against the dataset.

Before developing a theme, ask whether lack of time is the central pattern or one context among others, what changes when a space is available, and whether an informal paired trial qualifies the idea that shared space is necessary.

Codes help you work with passages; they are not natural categories or judgments about people. Themes develop through the relationship among codes, the question, context, and extracts—and by returning to material that does not fit neatly. [This broader method for reading open responses](https://www.harmate.com/en/blog/analyzing-200-open-responses-without-bias-a-method-for-actionable-decisions) takes a different route through corpus, signals, and decisions; it does not define Braun and Clarke’s reflexive approach.

### Worked example: from extracts to a candidate theme

The following three fictional statements are not enough to establish a general finding; they show how to make the interpretive trail visible.

| Fictional extract | Working code | Possible contribution to a theme | What qualifies it |
|---|---|---|---|
| “the meeting was already full” | no space in the meeting | The practice may remain secondary when no time is set aside | We do not know whether it was tried elsewhere |
| “we brought the question into the team check-in because we had twenty minutes” | shared time available | A shared space may make reflection possible | One account does not show that this ritual is necessary |
| “I tried it with a colleague before the meeting” | informal paired trial | The practice may find space outside the planned ritual | This qualifies a theme focused only on meetings |

A candidate theme might be: **“the practice advances when a concrete space is made for it, whether collective or informal.”** This links the extracts around an idea without presenting scheduled team time as a universal condition. Return to the full dataset, look for accounts that do not fit, and state the contexts to which the theme may apply.

## Develop and revise themes recursively

Braun and Clarke describe [six phases](https://www.thematicanalysis.net/doing-reflexive-ta/) to support and teach reflexive thematic analysis: familiarizing yourself with the data, generating initial codes, constructing candidate themes, reviewing themes, defining and naming themes, and producing the analysis. They emphasize that the phases are recursive: an analyst may revisit earlier work, combine or split themes, or abandon one. They are not mandatory boxes whose completion guarantees quality.

A candidate theme should answer the question, not just gather sentences that use similar words. Check it at two levels: do the extracts collected under the theme form a sufficiently coherent idea? And does that idea still make sense when you return to the dataset as a whole?

Deliberately revisit discrepant accounts. Suppose several participants describe having no space to practice, while another says they reused the method in an informal exchange. Do not remove that account because it complicates the theme. It may reveal a condition—such as an opportunity to try the method with a peer—that sharpens the interpretation. It may also show that the theme needs to be limited to particular contexts.

Reflexivity includes making the analyst’s role visible: decisions, expectations, relationship to the field, and the choices that shaped the themes. In this approach, the analyst’s subjectivity is not a bias to be neutralized or measured through coder agreement; it is a resource for interpretation, provided it is made explicit. Discussion among colleagues or co-analysts may deepen an interpretation, but its purpose is not to converge on a single “correct” reading.

## Keep a record of decisions without fixing meaning

Keep a light revision trail: code versions, why something changed, extracts that support or challenge a theme, and questions that remain open. This lets readers follow the reasoning without pretending that choices were mechanical.

| Revision | Why? | Check to retain |
|---|---|---|
| “not enough time” becomes “no recognized time slot” | accounts distinguish available duration from the status given to the practice | reread extracts where a trial happened outside a meeting |
| provisional theme splits in two | some passages concern authorization, others scheduling | compare contexts without reducing cases to a frequency |
| discrepant extract retained | it shows another condition for applying the practice | state the scope of the final theme |

Do not set a universal frequency threshold at which a pattern becomes a theme. Frequency may be useful for some questions and designs, but it does not determine analytic importance by itself. A rare pattern may illuminate an exception or limit; a very common word may explain little.

## Present an interpretation readers can examine

A useful report connects the question, analytic approach, theme-development process, contextualized extracts, and interpretation. Quotes do not speak for themselves: they illustrate an argument and remain limited to what the dataset supports. For selecting and contextualizing quotations without distorting them, see this [guide to verbatim quotations](https://www.harmate.com/en/blog/verbatim-how-to-use-exact-quotes-without-distorting-their-meaning). If these themes feed into a project review, this [lessons-learned guide and follow-up template](/en/blog/project-lessons-learned-method-questions-and-an-adaptable-template) shows how to connect insights to checkable actions. Protect participants and omit details that could identify a response.

Also state what the analysis does not establish: it does not prove a theme is representative, that every participant shares the same experience, or that a training session caused a change. Claims about change or effect require a design suited to that question and evidence beyond a collection of accounts.

For a discussion of how AI can support reading without deciding for the analyst, see [the limits of AI-assisted questionnaire analysis](https://www.harmate.com/en/blog/ai-survey-analysis-dont-confuse-a-summary-with-a-decision).

## A tool can support traceability, not choose meaning

When a team collects or imports questionnaire responses, a tool can help organize the dataset, suggest themes alongside source wording (starting points to check against the data, not finished themes), and prepare a report. Deciding what those materials mean still depends on the question, context, and human judgment. Harmate can support [collection and documented analysis](/en/product/open-questions); it does not replace the method choice or the analyst’s interpretation.

### Method sources

- [Doing Reflexive TA — Virginia Braun and Victoria Clarke](https://www.thematicanalysis.net/doing-reflexive-ta/): phases, recursive process, and interpretation in their approach.
- [Using thematic analysis in psychology — Braun and Clarke, 2006](https://doi.org/10.1191/1478088706qp063oa): foundational methods paper presenting thematic analysis as a flexible method.
- [Practical thematic analysis — BMJ, 2023](https://www.bmj.com/content/381/bmj-2022-074256): a different applied method guide; it should not be conflated with Braun and Clarke’s reflexive approach.