Thematic Analysis Example: How to Turn Interviews Into Themes

A thematic analysis example is the quickest way to learn this method, so that is what you get here. We will take four short interview quotes and turn them into finished themes you could put straight into a results chapter.

What thematic analysis actually is

Thematic analysis is a way of finding patterns of meaning in text. That text could be interview transcripts, focus group notes, or open ended survey answers.

Most people follow six steps published by two psychologists, Virginia Braun and Victoria Clarke, in 2006. You read the data, label it, group the labels, check the groups, name them, then write them up.

It sounds neat. In real life you will jump back and forth between steps, and that is completely normal.

The data we will use

I wrote these quotes for this guide. They are not from a real study, so please do not cite them.

Research question: How do postgraduate students feel about using AI tools in their research?

P1: "I never cite what it tells me. I always open the PDF and check."

P2: "For the first month I felt like I was cheating. Nobody told us what was allowed."

P3: "I got burned once. It gave me a reference and the paper did not exist. Now I check every single one."

P4: "I do not tell my supervisor how much I use it. It feels safer to show her the finished chapter."

A thematic analysis example, step by step

Step 1. Read everything first

Read all your data before you label anything. Then read it again.

Jot down what you notice. My notes said: everyone talks about checking, two people describe a feeling rather than an action, and two imagine their supervisor's reaction without ever asking her.

That last note turned into my best theme.

Step 2. Label the data with codes

A code is a short label on a meaningful chunk of text. Think of it like a sticky note.

Label generously. You can always merge codes later, but you cannot recover something you never noticed.

Step 3. Group codes into possible themes

Now step back and look for codes that share one idea.

My first groups were: nobody trusts the output, nobody told us the rules, managing the supervisor's opinion, and it saves time.

Step 4. Check the groups honestly

This is where weak themes fall apart.

"It saves time" had only one code behind it, and it was really just a topic. I cut it and mentioned it in a background sentence instead.

"Nobody told us the rules" and "managing the supervisor" were saying the same thing from two angles, so I merged them.

That left two themes, which is a healthy number for a masters project.

Step 5. Give each theme a real name

Theme 1: Verify everything, trust nothing. Students treat AI output as a rough draft with no authority until they have checked the source themselves.

Theme 2: Silence on every side. Students are making up their own rules because neither the university nor their supervisor has said anything.

One word names almost always signal a topic rather than a theme, so use a short phrase instead.

Step 6. Write it up

Here is what a finished theme looks like on the page.

Verify everything, trust nothing

All four students treated AI output as provisional. P1 drew the line at citation: "I never cite what it tells me." For P3, one fabricated reference reset the whole process: "Now I check every single one." A single bad experience mattered more than weeks of useful output, which suggests these students were reacting to the worst case they had personally seen rather than weighing reliability overall.

Notice the shape. Make a claim, show a quote, then explain what it means. If you stop at the quote, you have described your data but not analysed it.

Themes are not topics

This is the mistake that costs the most marks.

"AI tools" is a topic. It just tells your reader what you asked about. "Verify everything, trust nothing" is a theme, because it tells them what you found.

Quick test: if you can name your theme with a single word lifted from your interview guide, it is probably a topic.

Mistakes to avoid

1. Writing that theme "emerged." You built them. Say you constructed or developed them.

2. Claiming saturation. Braun and Clarke argued in 2021 that saturation does not fit reflexive thematic analysis, so justify your sample size by how focused your question is instead.

3. Having eight themes. That usually means you stopped at the coding stage.

4. Stacking three quotes together with no interpretation between them.

Should you use AI to code your data?

It depends what you ask it to do.

Studies published in 2025 and 2026 found that AI tools match human coders reasonably well when applying a codebook you wrote yourself. They are far less reliable when asked to invent codes and themes from scratch, and researchers in those studies concluded these tools can support human analysis but not replace it.

So use AI for transcription, tidying codes, or a first pass against your own codebook. Do not let it name your themes. Always check your institution's rules, and say in your methods what you used and how you checked it.

Where this fits in your dissertation

Thematic analysis sits inside the wider choices you make about your research methodology. It also connects to the theory you chose in your theoretical framework example and to the limits you set in your scope and delimitation section.

FAQs

How many themes should I have?

Usually two to four for a masters dissertation. More than six normally means you stopped too early.

How many interviews do I need?

There is no magic number. Six to twelve is common for small projects. Explain your reasoning rather than quoting a rule.

Can I do thematic analysis on survey answers?

Yes. Open ended survey responses work well, though the answers are shorter, so you usually need more of them.

Do I need software like NVivo?

No. Plenty of published studies were coded in a spreadsheet. Software helps with big datasets, but it does not do the thinking for you.

How can Canyam help with my methods chapter?

You can search across millions of papers and read summaries that show how other studies handled their methods, then open the originals to check the details before you cite them.

Try it yourself

The fastest way to learn is to copy this thematic analysis example with your own material. Take four quotes, code them, group the codes, then name the group in more than one word. If the name fails the topic test, you have learned something useful in an hour instead of three weeks.