Ask five researchers how many types of thematic analysis there are, and you’ll probably get different answers. Four. Three. Six. Seven, if they’re feeling generous, and you’re like:
See, they’re not wrong. They’re counting different things, and thematic analysis isn’t just one method. It’s a family of approaches, and the differences come down to a handful of questions:
- Do your codes come from a theory or from the data?
- Do you stick to what people say, or dig a little deeper?
- Is your coding process structured, and how much do you shape the final themes?
Once you see these questions, the “types of thematic analysis” cease to be confusing and start becoming choices you can make. Much better than guesswork.
In this guide, we’ll walk through all seven approaches, how they fit together, and how to pick the right one for your study. We’ll also look at an example that takes a single quote from raw data to finished theme, and the most common mistakes people make. Let’s get right into it!
So, what is thematic analysis?
Thematic analysis is a well-known method for qualitative data analysis. You identify, analyze, and report patterns of meaning (called themes) across data like interview transcripts, survey responses, focus group notes, etc.
This happens through coding. You read through your data, tag meaningful segments with short descriptive labels (the codes), and then group related codes into broader themes.
Sometimes, this is called thematic coding, and it’s the engine behind every approach in this guide. What changes from one approach to the next is how you code and what you take a theme to be.
Thematic analysis in qualitative research is popular for good reason: it’s flexible, works across data types and research questions, and is user-friendly for neophytes in qualitative methods.
The 7 types of thematic analysis
These seven types of thematic analysis fall into two groups:
- Orientations describe what you’re coding for (inductive vs. deductive, semantic vs. latent)
- Approaches describe how structured your process is, and what role you play as the researcher (reflexive, codebook, coding reliability)
Let’s look at them in order.
I: Where do your codes come from?
#1: Inductive thematic analysis
In inductive analysis, themes come from the data itself, not from a preset premise. You start with the raw material and let the patterns show up, without forcing them into an existing framework. It’s the right call for exploratory work, when you don’t yet know what matters.
Example: In a client satisfaction study, an inductive analysis might surface themes like “ease of use” or “price satisfaction” without you defining those categories in advance.
#2: Deductive thematic analysis
Deductive analysis flips this around. You start from an existing theory or framework and apply its concepts to your data, usually to test a hypothesis or see how well a model fits. Most of your codes are set before you open the first transcript.
Example: A UX researcher checks whether user feedback lines up with a predefined set of usability principles.
II: How deep do you read?
#3: Semantic thematic analysis
Semantic analysis stays at the surface. It focuses on what’s explicitly stated and takes participants’ words at face value. It’s simple, easy to trace back to the data, and a good fit when the meaning is plain.
Example: Customer reviews sorted into themes like “fast shipping” or “quality issues.”
#4: Latent thematic analysis
Latent analysis goes underneath. It looks for implied meanings, assumptions, and ideas that shape what people say. This takes more interpretation because you’re reading for what’s behind the words.
Example: Employee comments about workplace perks might point to a latent theme of “feeling undervalued.”
How the orientations combine
These two questions are independent, i.e., you can mix and match. An analysis can be inductive and semantic, inductive and latent, deductive and semantic, or deductive and latent. An inductive-semantic study might let themes emerge from survey comments while staying close to the literal wording, while a deductive-latent study might apply a theory about power or identity to read beneath what interviewees said.
Most projects make one choice from each pair: inductive or deductive, semantic or latent. Many also mix within a pair, starting with a few expected themes, then staying open to new ones.
III: How structured is your process?
The next three approaches describe the process itself. They differ on whether you use a codebook, whether you measure agreement between coders, and whether the researcher’s perspective is treated as a problem to control or a resource to use. These are the approaches Braun and Clarke (whose 2006 paper helped make thematic analysis so widely used) now describe as the main families within the method.
#5: Reflexive thematic analysis
Reflexive thematic analysis is organic and flexible, and it treats the researcher’s perspective as an active part of the analysis, not a bias. Codes can change as your understanding deepens. Themes aren’t “found” lying around in the data, you build them through sustained engagement with it, and they’re understood as patterns of shared meaning organized around a central idea. It’s the name Braun and Clarke now use for their own approach, and it’s the most widely taught form.
Because meaning is built through interpretation, reflexive analysis doesn’t use a fixed codebook and doesn’t calculate agreement between coders. Instead, you’re expected to be transparent about how your own position and choices shaped the analysis. It suits interpretive work, especially research on emotionally charged topics or experiences tied to identity.
#6: Codebook thematic analysis
Codebook thematic analysis balances structure and flexibility. The team works from a structured codebook and applies it to the data. The codebook can still be refined as you go, but the structure keeps everyone consistent.
It’s a common choice for team-based and applied research, where several people are coding and need to stay on the same page.
#7: Coding reliability thematic analysis
Coding reliability thematic analysis uses a fixed codebook and checks that multiple coders agree on how data segments fit the predefined labels. Agreement is measured with a statistic like Cohen’s kappa. The point is to show that the coding is consistent and doesn’t hinge on one person’s judgment, which makes this the approach closest to quantitative standards of rigor.
It’s a good fit when you need to demonstrate consistency, like on a large project with many coders. It’s a poor fit for interpretive work, where the goal isn’t for everyone to code identically but to build a considered reading of the data.
Three approaches at a glance
Thematic analysis: choosing the right approach
Ask yourself these three questions:
What are your research goals?
If you’re exploring something new and want participants’ voices to lead, choose inductive. Testing a hypothesis or checking data against an existing theory? Go deductive. And, if your goal is deep interpretation, lean towards reflexive. If it’s being consistent across coders, go for coding reliability.
How deep does the analysis need to go?
Semantic analysis fits simple feedback (reviews or survey responses) where the meaning sits on the surface. Latent analysis fits work that aims to uncover the assumptions beneath what people say, which is more common with rich interview transcripts.
What are your practical constraints?
Team size, time, and data type matter. Coding reliability needs multiple coders and a significant chunk of time. A small team working through customer feedback will often be well served by a semantic, codebook-style approach, while a solo researcher with a handful of in-depth interviews may find a reflexive, latent approach more rewarding.
Thematic analysis vs. other methods
Thematic analysis isn’t the only way to analyze qualitative data, and sometimes it may not be the right tool:
A worked example: from start to finish
Alright, now that we’ve gotten the theory out of the way, seeing one quote travel from raw data to finished theme is better. Note: this is an illustrative example.
Imagine you’ve interviewed people about their first few weeks with a project management tool.
Step 1: The raw quote
One participant says: “I kept opening the dashboard hoping it would tell me what I’d missed, but it never did. I didn’t want to ask my manager because I figured I was the only one who hadn’t worked it out.”
Step 2: Initial codes
You read through it and tag the meaningful parts:
- “Dashboard doesn’t show what’s missing” (semantic, close to what was said)
- “Reluctant to ask for help” (semantic)
- “Fear of looking behind” (latent, your interpretation of what’s underneath)
Step 3: Group codes across participants
Other interviews turn up similar codes: people who skipped asking teammates, guessed at features instead of checking the docs, and described feeling embarrassed. You group them together.
Step 4: Develop the theme
You name the pattern as a claim about what’s going on, not just a topic: “New users hide their confusion instead of asking for help, so onboarding problems stay invisible to the team.”
This also illustrates the difference between what the data says vs. a label like “onboarding experiences”, which tells you what the data is about. A good theme makes an argument you could agree or disagree with, and the codes underneath it are your evidence.
When to use thematic analysis
Thematic analysis is especially useful when:
- You have lots of qualitative data: It gives you a systematic way to work through piles of transcripts, responses, or notes.
- You’re new to qualitative methods: It’s accessible, and you don’t need advanced training in a specialized method to get started.
- Your data comes in different shapes: It works on interview transcripts, surveys, and focus groups alike.
- You want to check your findings with participants: Its flexibility makes it possible to validate themes with the people you studied.
Read: The 10 Best AI Tools for Thematic Analysis in 2026 (and how to actually pick one)
Common mistakes in thematic analysis
A few errors show up again and again, even among experienced researchers:
- Confusing topic summaries with themes: “Onboarding” is a topic. A theme says something about it.
- Writing about themes in the passive voice: “Themes emerged from the data” hides the fact that you built them. Say what you did.
- Applying inter-rater reliability to reflexive work: Agreement statistics fit coding reliability analysis. In reflexive analysis, they miss the point, since the goal is a considered interpretation, not identical coding.
- Coding too shallowly or too early: Locking in your codes after one quick pass usually means missing the patterns that only show up on a closer read.
Conclusion
Thematic analysis turns raw qualitative data into something you can act on, but only if the approach matches the question. Start with the choices that shape your study: where your codes come from, how deep you read, and how structured your process is. Then pick the approach that fits your goals, your data, and your team.
At Looppanel, our AI-assisted analysis speeds up the first pass of coding, and all tags are traceable to the original source, so you can always check the reasoning behind a theme. The interpretation stays in your hands, where it belongs.
Book a demo to see how it works.
P.S. Whichever approach you pick, write down why. Your future self (and anyone reviewing your work) will want to know.
Your peers also ask
What are the types of thematic analysis?
There are seven commonly described approaches: inductive, deductive, semantic, latent, reflexive, codebook, and coding reliability. The first four describe where your codes come from and how deeply you read. The last three describe how structured your process is and how much your own interpretation shapes the themes.
What are the three types of thematic analysis?
It depends who you ask. In Braun and Clarke’s framing, the three main approaches are coding reliability, codebook, and reflexive thematic analysis. Some guides use “three types” to mean something else entirely, like inductive, deductive, and semantic. Check which framing a source is using before comparing counts.
What are inductive and deductive analysis?
Inductive analysis lets themes emerge naturally from the data. Deductive analysis is guided by existing theories or frameworks that you apply to the data.
What’s the difference between reflexive and coding reliability thematic analysis?
Reflexive analysis treats your perspective as a resource and builds themes through interpretation, with no fixed codebook and no measure of agreement between coders. Coding reliability analysis uses a fixed codebook and measures how well multiple coders agree, aiming for consistency over interpretive depth.
Is thematic analysis qualitative or quantitative?
It’s a qualitative method. It analyzes non-numerical data like interviews and open-ended responses. Coding reliability approaches borrow a quantitative-style check (agreement statistics), but the data and the goal of finding patterns of meaning stay qualitative.
.webp)
.svg)
