Methods Guide

Dissertation Coding and Thematic Analysis: From Transcripts to Themes

How to code qualitative data, build a codebook, develop themes that hold up in a viva or defense, and write the analysis up transparently.

Updated October 2026 · 7 min read

Dissertation coding and thematic analysis turn hours of interviews into a small set of well-evidenced themes that answer your research questions. Done well, every theme can be traced back to the data.

This guide follows the process from first reading to final write-up. It uses Braun and Clarke's thematic analysis as the main example, because it is the approach most dissertations cite, and notes where grounded theory or codebook methods work differently.

The Six Phases of Thematic Analysis at a Glance

Braun and Clarke describe six phases: familiarize yourself with the data, generate initial codes, construct initial themes, review those themes, define and name them, and write the report. The phases are recursive, so you will move back and forth rather than march through them once.

A theme is a pattern of shared meaning organized around a central idea, not just a topic that many participants mentioned.

Choose Your Analytic Approach First

"Thematic analysis" covers several approaches with different assumptions. Your methodology chapter must name the one you used and explain why it fits your question and your stance on knowledge.

ApproachHow coding worksInter-rater checks?Good fit when
Reflexive thematic analysis (Braun and Clarke)Organic, evolving codes; researcher subjectivity treated as a resourceGenerally not, as agreement scores conflict with its assumptionsYou want an interpretive account of meaning across a dataset
Codebook thematic analysis (e.g. template or framework analysis)A structured codebook, often partly set in advanceSometimes, depending on the team and designApplied or policy research with predefined questions
Coding reliability approachesFixed codebook applied by several codersYes, often with a statistic such as Cohen's kappaA more positivist stance, larger teams, or deductive questions
Grounded theory (Strauss and Corbin; Charmaz)Open, axial or focused, then theoretical codingNot usually centralYou aim to build a theory of a process

Mixing the language of one approach with the procedures of another is a frequent examiner comment. For example, claiming reflexive thematic analysis while reporting kappa scores suggests confusion about the method.

Phase One: Get Close to the Data

Read every transcript at least once before coding, ideally while listening to the audio. Note first impressions in a separate memo, not in the codes themselves.

If you transcribe yourself, that counts as familiarization. Check transcripts for accuracy and make sure identifying details are removed before anyone else sees them.

Phase Two: Generating Initial Codes

A code is a short label for a feature of the data that is relevant to your question. Code line by line or by meaningful segment, and code inclusively: one extract can carry several codes.

  • Semantic codes capture what participants explicitly said ("worried about cost").
  • Latent codes capture underlying ideas or assumptions ("care framed as a moral duty").
  • In vivo codes use participants' own words ("just getting through the shift").

In grounded theory this stage is called open coding (or initial coding in Charmaz's version). The next step there, axial or focused coding, links codes into categories around conditions, actions and consequences.

Software or by Hand?

NVivo, ATLAS.ti, MAXQDA and similar tools help you organize and retrieve extracts, but they do not analyze anything for you. Hand coding with highlighters or a spreadsheet is perfectly acceptable for a smaller dataset. Choose what lets you keep a clear audit trail.

Building and Maintaining a Codebook

Even in reflexive work, a simple codebook keeps your thinking visible. For codebook and reliability approaches it is essential.

FieldWhat to record
Code nameA short, specific label
DefinitionWhat the code captures, in one or two sentences
Include / excludeBoundaries that separate it from similar codes
Example extractOne typical quotation, with participant ID
Change logWhen and why the code was merged, split or renamed

The change log is what examiners love. It shows that your themes developed through analysis rather than being decided before you started.

Phases Three to Five: From Codes to Themes

Group related codes into candidate themes, then test each one against the coded extracts and the whole dataset. Ask whether the theme has a clear central idea, whether it is distinct from the others, and whether it answers part of your question.

Worked mini-example. In a study of newly qualified nurses, codes included "afraid to ask senior staff", "pretending to know", "relief when a mentor checked in" and "learning from mistakes in private".

An early candidate theme, "Support", was too broad: it was a topic, not an idea. After review it became "Performing competence": new nurses hid uncertainty to appear capable, which delayed their learning. The codes about mentors became a subtheme, "Permission to not know", which showed when the pattern broke.

Name themes so they say something. "Performing competence" tells the reader the argument; "Challenges" does not.

How Much Data Is Enough?

Saturation, the point where new data add no new codes or insights, comes from grounded theory and is widely cited in dissertations. A frequently quoted study by Guest, Bunce and Johnson found most codes appeared within the first twelve interviews of a fairly homogeneous sample, but that finding does not transfer to every study.

Braun and Clarke argue that saturation does not fit reflexive thematic analysis, since meaning is generated by the researcher rather than waiting to be found. Many examiners now accept "information power" or a reasoned justification based on your aims, sample and depth of data instead. Check what your department expects.

Making Your Coding Defensible

  • Audit trail: keep coded transcripts, codebook versions and theme maps.
  • Reflexive memo: record how your background might shape interpretation, and what you did about it.
  • Negative cases: report extracts that do not fit and explain them.
  • Thick description: give enough context for readers to judge transferability.
  • Trustworthiness: many dissertations use Lincoln and Guba's criteria of credibility, transferability, dependability and confirmability.

Member checking (asking participants to review findings) is common but debated. Use it only if it fits your approach, and explain how you handled disagreement.

Writing the Analysis Up

Present each theme with a short definition, an analytic narrative and two or three well-chosen extracts. The narrative should do most of the work; quotations illustrate it.

A thematic map or a table of themes, subthemes and contributing participants helps readers. Keep interpretation in the findings for reflexive work, and save comparison with the literature for the discussion chapter unless your department prefers a combined chapter.

Common Coding Mistakes Examiners Notice

MistakeFix
Themes that mirror the interview questionsLook for patterns that cut across questions; a theme is an idea, not a heading from your guide
Codes that are just topics ("workload")Make codes say something about the data ("workload used to justify skipped breaks")
Counting mentions as proof of importanceArgue importance from meaning and relevance to the question, not frequency alone
Quotations with no analysisFollow each extract with your interpretation of what it shows
No record of how themes changedKeep dated theme maps and a codebook change log from the start

A quick self-test: could another researcher follow your audit trail from a theme back to the exact extracts that support it? If not, tighten the record before you submit.

How DissertationHorse Helps with Coding and Thematic Analysis

Send anonymized transcripts or field notes, your research questions, your chosen approach and your methodology chapter if you have one. A writer experienced in qualitative methods prepares a custom analysis write-up with a codebook, theme definitions and an audit trail, for you to study and use as a reference.

We work only with data you have collected. We never invent participants, quotations or findings. Free revisions within the scope of your original request are included.

Want to See Your Analysis Done Well?

Send anonymized transcripts, your questions and your chosen approach. You get a custom analysis write-up with an audit trail, written to your brief.

Order Thematic Analysis Help

Free revisions · Full refund if late · Written from scratch for your order

Frequently Asked Questions

What is the difference between a code, a category and a theme?

A code labels one feature of the data. A category groups related codes. A theme is a pattern of shared meaning built around a central idea that helps answer your research question.

How many themes should a qualitative dissertation have?

There is no fixed number, but many dissertations report three to six main themes, sometimes with subthemes. Fewer, well-developed themes are better than a long list of thin ones.

Do I need inter-rater reliability for thematic analysis?

It depends on the approach. Coding reliability methods use it, but Braun and Clarke advise against agreement scores in reflexive thematic analysis. Say which approach you use and stay consistent with it.

Should I use NVivo or code by hand?

Either is acceptable. Software helps with large datasets and retrieval, while hand coding suits smaller studies. The software does not analyze the data, so your examiners will still focus on your reasoning.

Is thematic analysis the same as grounded theory?

No. Thematic analysis identifies patterns of meaning across a dataset. Grounded theory aims to build a theory of a process, using constant comparison, theoretical sampling and memo writing.

Can I code data deductively from a theory?

Yes. Deductive or theory-driven coding is legitimate, and many studies combine it with inductive coding. Explain where your initial codes came from and how new ones were added.

How do I show saturation in my methodology chapter?

If you use saturation, describe how you tracked new codes across interviews, for example with a table. If your approach does not use saturation, justify your sample size another way.

Can I code transcripts I have already collected and transcribed?

Yes, as long as you collected them under ethical approval that covers this use, and they are anonymized before sharing.