How to Do Thematic Analysis: From Interview Transcripts to Meaningful Themes

You have completed your interviews.

You now have pages of transcripts containing stories, explanations, experiences, contradictions, pauses, examples, and ideas.

Then comes the difficult question:

What do I actually do with all this text?

For many qualitative studies, one possible answer is thematic analysis.

But thematic analysis is not simply reading transcripts, highlighting repeated words, and turning the most common topics into themes.

Analysis requires you to move from raw qualitative data toward an interpretation of patterns of meaning that help answer your research question.

That involves careful reading, coding, comparison, theme development, refinement, and writing.

The process is also rarely perfectly linear. Researchers often move backwards and forwards between the data, codes, candidate themes, and developing interpretations as their understanding becomes deeper.

The aim is not merely to organise your transcripts.

It is to make a defensible analytical argument about what the data mean.

What is thematic analysis?

Thematic analysis is an approach for identifying, analysing, and interpreting patterns of meaning across qualitative data.

Depending on the study, that data might include:

• interview transcripts;

• focus-group transcripts;

• written responses;

• diaries;

• documents; or

• other forms of qualitative material.

Thematic analysis is flexible, but flexibility does not mean “anything goes.”

You still need to explain:

• what version or approach to thematic analysis you are using;

• how you approached coding;

• how themes were developed;

• how your theoretical or conceptual position influenced the analysis; and

• how the resulting interpretation addresses the research question.

University qualitative-analysis training similarly treats coding, developing themes, analytic interpretation and writing as connected parts of analysis rather than separate mechanical tasks.

Make sure thematic analysis fits your study

Do not use thematic analysis simply because you collected interviews.

Interviews are a data-generation method.

Thematic analysis is an analytical approach.

The two are not automatically tied together.

Other qualitative approaches may analyse interview data differently depending on the methodology and research question.

Before proceeding, ask:

Am I actually interested in identifying and interpreting patterns of meaning across this dataset?

If yes, thematic analysis may be appropriate.

Your choice should remain consistent with your research methodology rather than being selected only because the method sounds familiar.

Start with the research question

Your research question helps determine what is relevant in the dataset.

Suppose your study asks:

How do first-generation university students experience academic belonging during their first year?

Your transcripts may contain discussion of:

• friendships;

• lecturers;

• accommodation;

• family;

• finances;

• commuting;

• assessment;

• clubs;

• food;

• weather;

• employment;

• confidence; and

• dozens of other subjects.

Not everything mentioned automatically belongs in the final analysis.

The analytical question is:

What in these accounts helps me understand students' experiences of academic belonging?

This does not mean ignoring unexpected findings.

An unexpected issue may become analytically important precisely because participants reveal a connection you had not anticipated.

But the research question remains an important anchor.

Understand what a theme is

One of the easiest mistakes is treating every topic as a theme.

Imagine several participants mention:

• lecturers;

• classmates;

• family;

• financial pressure; and

• commuting.

You could create five headings using those words.

But you may have created topic summaries, not meaningful themes.

A theme should capture a pattern of shared meaning organised around a central idea.

For example, extracts concerning lecturers, classmates, and family might contribute to a broader theme such as:

Belonging depended on being recognised as someone who legitimately belonged at university.

Now the theme makes an analytical claim.

It tells the reader something about how belonging operated in participants' accounts, rather than merely reporting who they talked about.

Braun and Clarke's reflexive thematic-analysis guidance makes this distinction particularly important: themes are patterns of shared meaning organised around a central concept, rather than simple domain summaries.

Themes are not determined by frequency alone

Suppose twelve participants mention workload, while only four discuss feeling embarrassed about asking lecturers for help.

Does that automatically make workload the more important theme?

No.

Frequency can sometimes be relevant, but thematic importance depends on the research question and the analytical meaning of the pattern.

Those four accounts might reveal something important about how students negotiate belonging, legitimacy, and help-seeking.

Likewise, a frequently mentioned issue may be largely descriptive and contribute little to the central research question.

Do not reduce qualitative analysis to:

“Most participants mentioned X, therefore X is a theme.”

Ask instead:

What does this pattern help me understand?

Familiarise yourself with the data

Before building themes, spend time becoming familiar with the dataset.

This may involve:

• transcribing interviews;

• checking transcripts against recordings;

• reading transcripts repeatedly;

• noting initial observations;

• recording questions;

• identifying contradictions;

• noticing potentially important patterns; and

• reflecting on your reactions to the data.

The purpose is not to memorise every sentence.

It is to begin understanding the dataset as a whole before fragmenting it into codes.

In published applications of thematic analysis, familiarisation commonly involves checking transcripts and repeatedly reading the dataset before systematic coding begins.

Do not rush through transcription

If you transcribe your own interviews, transcription can itself contribute to familiarisation.

You hear:

• how participants phrase ideas;

• where they hesitate;

• what they emphasise;

• where their accounts change direction; and

• which issues repeatedly return.

If automated transcription is used, do not assume the output is accurate.

Errors can occur with:

• names;

• accents;

• specialist terminology;

• overlapping speech;

• quiet sections; and

• poor recordings.

Check transcripts to the level required by your analytical approach before treating them as authoritative data.

Article #30's interview-guide discussion already emphasised planning recording and transcription before data collection. Thematic analysis is where the consequences of those decisions become very visible.

Make initial notes—but do not confuse them with final themes

During familiarisation, you may write notes such as:

• “Several students describe avoiding lecturers.”

• “Peer relationships seem important during first semester.”

• “Interesting contradiction: wants independence but also wants guidance.”

• “Commuters describe missing informal interactions.”

• “Financial pressure may affect participation.”

These are useful analytical observations.

But they are not automatically codes or themes.

At this stage, allow yourself to notice possibilities without forcing the dataset into a finished structure too early.

Your understanding should be allowed to change.

What is coding?

Coding involves identifying and labelling features of the data that are relevant or potentially meaningful to your analysis.

Suppose a participant says:

“I used to sit at the back and leave immediately after lectures because everyone else seemed to know what they were doing. I didn't want anyone to realise how lost I felt.”

Possible codes might include:

• hiding uncertainty;

• avoiding peer interaction;

• feeling academically inadequate;

• comparing self with others;

• fear of being exposed; or

• withdrawing from university spaces.

Different researchers may code the same extract differently because coding involves analytical judgment.

The question is not:

“What is the one objectively correct code?”

It is:

“What aspect of this extract matters for my research question and analytical approach?”

Codes are not themes

A code usually captures something interesting or relevant about a particular piece of data.

A theme is a broader pattern of meaning developed across multiple coded extracts.

For example:

Codes:

• hiding confusion;

• reluctance to ask lecturers;

• pretending to understand;

• comparing oneself with confident classmates;

• fear of appearing academically weak.

These codes might eventually contribute to a candidate theme such as:

Belonging required performing academic competence.

The theme does more than collect similar statements.

It proposes an interpretation connecting them.

That movement from data → codes → patterns → themes → analytical argument is one of the central tasks of thematic analysis.

Code enough context to preserve meaning

Suppose a participant says:

“At first I hated group work because I felt like everyone knew more than me. But after our second project, my group started asking for my opinion, and that changed everything.”

If you code only:

“I hated group work”

you lose the transformation contained in the account.

If you code only:

“that changed everything”

you lose what caused the change.

Your coded extract should contain enough surrounding material for you to understand what the participant meant when you return to it later.

Do not fragment transcripts so aggressively that the meaning disappears.

Coding can be semantic or more interpretive

Some coding stays relatively close to what participants explicitly say.

For example, a participant says:

“I couldn't attend evening events because my bus home left at six.”

A relatively semantic code might be:

Transport limits participation.

A more interpretive analysis might consider how the account reflects unequal access to the informal spaces where university belonging develops.

Neither approach is automatically superior.

What matters is that your analytical depth is appropriate to:

• the research question;

• methodology;

• theoretical position; and

• version of thematic analysis being used.

Do not claim a deeply interpretive analysis if your codes and themes merely restate participants' words.

Inductive and deductive orientations

You may also encounter the distinction between inductive and deductive thematic analysis.

An inductive orientation allows coding and theme development to be strongly informed by patterns encountered in the dataset.

A deductive orientation gives greater analytical priority to existing theory, concepts, or research questions.

In practice, the distinction is not always absolute.

Researchers bring prior knowledge and assumptions to analysis even when trying to remain open to unexpected patterns.

Conversely, a theoretically informed analysis can still identify material that does not fit initial expectations.

Recent applications of thematic analysis continue to distinguish broadly between data-led inductive and concept-led deductive orientations.

The important thing is to explain your analytical orientation rather than casually claiming:

“The themes emerged completely from the data.”

Themes require analytical work.

Code systematically

Do not carefully code the first three transcripts and then skim the remaining fifteen because you think you already know the themes.

Your analysis needs systematic engagement with the dataset relevant to your chosen approach.

As you code, you may:

• create new codes;

• revise existing codes;

• merge similar codes;

• split overly broad codes;

• rename codes;

• reconsider earlier extracts; and

• notice relationships between codes.

That evolution is normal.

A published University of Sydney study, for example, describes repeatedly returning to interview data to reassess codes and refine themes rather than treating coding as a one-pass exercise.

Keep an audit trail of analytical decisions

As your analysis develops, record important decisions.

You might note:

• why two codes were merged;

• why a candidate theme was abandoned;

• why a surprising pattern became important;

• how your theoretical framework influenced interpretation;

• why a code was renamed;

• why certain extracts were moved between themes; or

• how your understanding changed after rereading transcripts.

This can help you explain the analytical process later and reduce reliance on memory.

Your notes do not need to become a diary of every mouse click.

Record decisions that help make the development of the analysis understandable.

Software can organise data—it cannot do the interpretation for you

Qualitative-analysis software such as NVivo can help researchers:

• store transcripts;

• attach codes;

• retrieve coded extracts;

• organise categories;

• write memos; and

• examine relationships within the coded dataset.

But software does not remove the need for analytical judgment.

A program can retrieve every extract assigned to a code.

It cannot decide for you what those extracts mean in relation to your research question.

University qualitative-analysis teaching similarly separates the use of analytic tools from the interpretive work of developing codes, themes and arguments.

You can also conduct thematic analysis without specialist software, particularly with manageable datasets.

The quality of the analysis depends more on the reasoning than on the brand of software.

Begin looking for relationships among codes

Once you have developed a substantial set of codes, begin asking how they relate.

Suppose your codes include:

• hiding confusion;

• avoiding asking questions;

• fear of judgment;

• comparing self with peers;

• pressure to appear capable;

• relief when others admit difficulty;

• confidence after receiving recognition.

You might notice that several codes concern the need to appear academically legitimate.

That gives you the beginning of a candidate theme.

Other codes may form a different pattern.

Some may not fit anywhere yet.

Do not force every code into a theme simply to make the analysis look tidy.

Develop candidate themes around shared meaning

Once you have coded the dataset, the next task is not simply to turn each code into a theme.

Instead, look across your codes and coded extracts for broader patterns of shared meaning.

Suppose you have codes such as:

• hiding confusion;

• avoiding asking lecturers for help;

• pretending to understand;

• comparing yourself with confident classmates;

• fear of being exposed as academically weak; and

• pressure to appear capable.

These codes may share an underlying idea:

Students felt that belonging required them to appear academically competent.

That could become a candidate theme:

Performing competence to protect belonging

The theme brings several related experiences together around a central organising idea.

This is different from creating a theme called:

Academic experiences

which says very little about the pattern itself.

A theme should make an analytical claim

Compare these possible theme names:

Lecturers

Peer relationships

Challenges

with:

Recognition from others made belonging feel legitimate

Belonging was negotiated through comparison with peers

Asking for help risked exposing academic insecurity

The second group tells the reader something about what the researcher has interpreted in the data.

That is usually a stronger basis for thematic analysis than simply using interview topics as themes.

Your interview guide may organise conversations around lecturers, peers, support, and challenges.

Your final themes do not have to reproduce those headings.

If they do, ask whether you have genuinely analysed the data or simply reorganised your interview guide.

Candidate themes can change

The first themes you develop are provisional.

You may discover that:

• two themes actually express the same central idea;

• one theme contains several unrelated patterns;

• a theme has too little meaningful evidence;

• some extracts fit better elsewhere;

• a supposed theme is merely a topic;

• one theme is actually a subtheme; or

• an important pattern has been overlooked.

This is normal analytical development.

Do not become attached to a candidate theme simply because you have already given it a clever name.

Review themes against coded extracts

Return to the extracts associated with each candidate theme.

Ask:

• Do these extracts belong together meaningfully?

• What central idea connects them?

• Are there extracts that contradict the developing interpretation?

• Is the theme internally coherent?

• Have I included material merely because it mentions the same topic?

• Does the theme help answer the research question?

Suppose a candidate theme is:

Support creates belonging

But some extracts concern emotional support from friends, others concern financial support from family, and others concern technical advice from lecturers.

The word support appears across them, but that does not necessarily mean they form one coherent pattern of meaning.

You may need to rethink what actually connects the extracts.

Review themes against the whole dataset

Theme review should not stop with the extracts already assigned to the theme.

Return to the wider dataset.

You may find:

• relevant extracts you missed;

• examples that complicate the theme;

• participants whose accounts challenge the pattern;

• important differences between contexts; or

• evidence that the theme is too broad.

This helps prevent the analysis from becoming self-confirming.

Once you develop an idea, it is easy to notice only the evidence that supports it.

Returning to the dataset helps you ask whether your interpretation genuinely works across the material relevant to the analysis.

Pay attention to contradiction and variation

Qualitative analysis does not require every participant to say the same thing.

Suppose most participants describe peer relationships as helping them belong, but several describe peer groups as spaces of comparison and exclusion.

Do not automatically discard the minority accounts as exceptions.

The variation may reveal something important.

Perhaps peer relationships are not simply “supportive.”

Perhaps they operate as both a source of connection and a standard against which students judge whether they fit.

That tension may produce a stronger analysis than reporting:

Most participants said peers were important.

Contradictions can deepen themes rather than destroy them.

Consider subthemes carefully

A broad theme may contain meaningful sub-patterns.

For example:

Theme: Performing competence to protect belonging

Possible subthemes might include:

• hiding confusion from peers;

• reluctance to seek academic help; and

• using achievement as evidence of legitimacy.

Subthemes can help organise complexity.

But do not create them merely to make the analysis look sophisticated.

If a subtheme does not add useful structure or analytical distinction, you may not need it.

Define the central organising idea

Before finalising a theme, try completing this sentence:

This theme captures...

For example:

This theme captures how participants concealed academic uncertainty because appearing capable was tied to whether they felt they legitimately belonged at university.

That statement gives the theme a clear centre.

If your definition becomes:

This theme captures lecturers, peers, confidence, support, challenges, workload, adjustment and belonging...

the theme is probably too broad.

A strong theme should have a coherent analytical focus.

Give themes informative names

Theme names help readers understand your argument.

A name such as:

Challenges

is technically possible but analytically weak.

What kind of challenges?

Why do they matter?

A more informative name might be:

Struggling privately to preserve an image of competence

Likewise:

Support

could become:

Recognition from others made university feel like a place participants could belong

You do not need dramatic or poetic theme names.

Clarity matters more.

The name should communicate the central idea without forcing the reader to guess.

Write a short definition for every theme

Before writing the findings chapter, prepare a concise explanation of each theme.

For example:

Theme: Performing competence to protect belonging

This theme captures how participants concealed confusion, avoided asking for help, and compared themselves with apparently confident peers because academic competence became connected to whether they felt entitled to belong at university.

A definition helps you check that:

• the theme has a central idea;

• it differs from other themes;

• you know what belongs within it; and

• you can explain its relevance to the research question.

If two theme definitions sound almost identical, they may need to be combined or differentiated more clearly.

Do not force a predetermined number of themes

There is no universal rule that a dissertation needs:

• three themes;

• five themes;

• six themes; or

• one theme per research objective.

The appropriate number depends on the dataset, research question, analytical approach, and level of complexity.

Too many themes can leave the analysis fragmented and shallow.

Too few can force unrelated patterns together.

Choose enough themes to tell a coherent analytical story without turning every code into a separate section.

Reflexivity matters

The researcher does not approach qualitative data as an empty container.

You bring:

• previous reading;

• theoretical commitments;

• disciplinary training;

• personal experiences;

• assumptions;

• expectations; and

• decisions about what to notice and pursue.

Reflexivity involves examining how those positions and decisions shape the research process and interpretation.

Suppose you entered the study expecting financial pressure to be the main barrier to belonging.

During coding, you may be especially sensitive to financial issues.

Reflexive practice might involve asking:

Am I giving this pattern analytical importance because participants' accounts support it, or because I expected to find it?

The goal is not to eliminate yourself from the analysis.

In reflexive thematic analysis, the researcher's interpretive role is part of knowledge production rather than something that can simply be removed. Braun and Clarke explicitly distinguish this from approaches that treat researcher subjectivity primarily as a source of bias to be eliminated.

Be consistent with the version of thematic analysis you claim to use

“Thematic analysis” describes a family of approaches rather than one universally identical procedure.

This matters because practices that make sense under one approach may conflict with the assumptions of another.

For example, if you explicitly claim to use Braun and Clarke's reflexive thematic analysis, be careful about automatically importing practices such as:

• treating themes as objectively waiting inside the data to be discovered;

• requiring multiple coders to produce identical codes;

• using inter-rater reliability as proof that coding is correct; or

• assuming consensus coding removes researcher subjectivity.

Braun and Clarke's guidance specifically cautions against mixing incompatible assumptions from different forms of thematic analysis.

The lesson is broader than any one framework:

Name your analytical approach and use procedures that make sense within it.

Do not assemble a methodology from unrelated techniques simply because each appears in another dissertation.

Use quotations as evidence—not decoration

Participant quotations can help readers see how your interpretation connects to the data.

But a findings section should not become:

Quote → quote → quote → quote

with little analysis.

Likewise, do not write a long interpretation and attach one quotation at the end merely as decoration.

A stronger pattern is:

Analytical claim → relevant extract → interpretation → connection to the broader theme

For example:

Several participants described concealing uncertainty because asking for help risked revealing that they felt academically out of place.

Then provide an appropriate extract.

Afterward, explain what the extract contributes to the interpretation.

The quotation provides evidence.

Your analysis explains why that evidence matters.

Choose quotations carefully

A useful quotation should help illustrate or develop the analytical point.

It does not necessarily need to be:

• the longest;

• most emotional;

• funniest; or

• most dramatic.

Ask:

• Does this extract clearly support the analytical claim?

• Does it add something beyond what I have already paraphrased?

• Is enough context included to understand it?

• Have identifying details been handled appropriately?

• Am I relying too heavily on the same participant?

• Does the quotation represent the theme appropriately?

Also remember that reporting quotations can create confidentiality risks, particularly in small or identifiable populations.

Follow your ethical and institutional requirements.

Do not confuse the findings with the transcript

Your findings chapter should not reproduce interviews in chronological order.

Participant 1 said this.

Participant 2 said that.

Participant 3 disagreed.

That structure usually keeps the analysis centred on individual interviews rather than patterns across the dataset.

Thematic findings are generally organised around the analytical themes you developed.

Individual participants and extracts then contribute evidence to those themes.

Move from description to interpretation

Consider:

Eight participants said they were reluctant to ask lecturers questions.

That describes the dataset.

Now consider:

Reluctance to seek help was often tied to participants' fear that asking basic questions would expose them as less academically capable than their peers. Help-seeking therefore became entangled with whether they felt entitled to occupy the identity of a successful university student.

That is interpretive.

It tells the reader what the pattern may mean.

Good qualitative analysis often contains both description and interpretation, but the findings should do more than count or summarise responses.

A worked example: transcript → codes → candidate theme

Suppose your research question is:

How do first-generation university students experience academic belonging during their first year?

A participant says:

“Everyone in my tutorial seemed confident. Even when I didn't understand something, I kept quiet because I thought maybe I was the only one who didn't get it. When another student finally asked the same question I had, I remember feeling relieved because I realised it wasn't just me.”

Step 1: Notice what is happening

The extract contains several potentially relevant ideas:

• comparison with peers;

• perceived confidence of others;

• hiding confusion;

• fear of being uniquely inadequate;

• reluctance to ask questions; and

• relief when uncertainty becomes shared.

Step 2: Generate codes

Possible codes might include:

• comparing self with confident peers;

• hiding academic uncertainty;

• fear of being the only one struggling;

• silence as self-protection;

• normalisation of difficulty; and

• relief through shared uncertainty.

These are possibilities, not the one universally correct coding solution.

Step 3: Compare with other coded extracts

Suppose other participants describe:

• pretending to understand lectures;

• avoiding office hours;

• feeling embarrassed when asking “basic” questions;

• becoming more confident after hearing peers admit difficulty; and

• feeling accepted when lecturers normalise mistakes.

Now a broader pattern begins to appear.

Step 4: Develop a candidate theme

You might initially call it:

Academic insecurity

But that still describes a broad topic.

After further analysis, you might develop:

Performing competence to protect belonging

The central idea is that participants sometimes concealed uncertainty because appearing academically capable was connected to feeling legitimate within the university.

Step 5: Review the theme

Return to the coded extracts and wider dataset.

Ask:

• Does this interpretation work across the relevant accounts?

• What variation exists?

• Are there extracts that challenge it?

• Is the theme distinct from other candidate themes?

• Does it illuminate the research question?

Step 6: Define the theme

You might define it as:

This theme captures how participants managed visible signs of uncertainty because academic competence became connected to whether they felt they legitimately belonged at university.

Step 7: Write the analytical story

The final findings section would not simply list the codes.

It would explain how participants' efforts to appear competent shaped help-seeking, peer comparison, and their sense of legitimacy as university students, supported by carefully selected extracts.

That is the movement from transcript to interpretation.

Common thematic-analysis mistakes

Treating interview topics as themes

Your interview-guide headings are not automatically your findings.

Calling every code a theme

Themes operate at a broader level of patterned meaning.

Choosing themes because words occur frequently

Frequency alone does not determine analytical importance.

Coding only what confirms expectations

Unexpected and contradictory evidence deserves attention.

Creating themes too early

Familiarise yourself with and code the dataset before forcing it into a finished thematic structure.

Using themes that are too broad

A theme called Student experiences may contain almost the entire dataset while explaining very little.

Using themes that are too thin

A theme supported by one isolated comment may not represent a meaningful pattern across the dataset, depending on the analytical purpose.

Treating software as the analyst

Software can organise coded material; interpretation remains analytical work.

Using quotations instead of analysis

Participants' words are evidence, not a replacement for your explanation.

Claiming themes simply “emerged”

This wording can hide the researcher's active role in coding, interpreting, organising, and refining patterns.

Mixing incompatible analytical approaches

Be clear about which form of thematic analysis you are using and why its procedures fit your study.

Ignoring reflexivity

Your analytical decisions and assumptions influence what you notice and how you interpret it.

Failing to return to the whole dataset

Candidate themes need to make sense beyond the handful of extracts that first inspired them.

A thematic-analysis checklist

Before finalising your analysis, ask:

1. Does thematic analysis fit my research question and methodology?

2. Have I clearly identified which approach to thematic analysis I am using?

3. Have I familiarised myself sufficiently with the whole dataset?

4. Have transcripts been checked to the level required by my analysis?

5. Have I coded the relevant dataset systematically?

6. Do my codes capture features that matter to the research question?

7. Have I preserved enough context around coded extracts?

8. Have I considered both expected and unexpected patterns?

9. Have I examined contradictions and variation?

10. Are my themes more than topic headings?

11. Does each theme have a clear central organising idea?

12. Are the extracts within each theme meaningfully connected?

13. Have I reviewed candidate themes against the wider dataset?

14. Are the themes sufficiently distinct from one another?

15. Have I avoided forcing a predetermined number of themes?

16. Can I clearly define each theme?

17. Do theme names communicate the analytical idea?

18. Have I reflected on my role in producing the interpretation?

19. Are quotations being used as evidence rather than substitutes for analysis?

20. Does the final thematic story actually answer the research question?

If several answers are no, the analysis probably needs further development.

Final takeaway

Thematic analysis is not a shortcut from transcripts to headings.

It is an analytical process in which you move repeatedly between:

data → codes → patterns → candidate themes → refined themes → interpretation → written argument

Begin by becoming deeply familiar with the dataset.

Code material that matters to your research question.

Look for relationships and patterns across those codes.

Develop candidate themes around shared meaning rather than repeated words alone.

Then challenge those themes.

Return to the extracts.

Return to the whole dataset.

Look for variation and contradiction.

Define what each theme actually says.

Use participant quotations to support your interpretation rather than replace it.

And remain aware that themes are produced through analytical engagement with the data—not simply discovered waiting inside transcripts.

The most useful question is therefore not:

“What topics did my participants mention most often?”

It is:

“What patterns of meaning across these accounts help me answer my research question?”

When your themes can answer that question clearly and convincingly, you have moved beyond organising qualitative data and into meaningful analysis.