Thematic analysis is a qualitative research method for identifying, analyzing and reporting patterns of meaning, called themes, across data such as interview transcripts, focus group discussions and open-ended survey responses. A widely used approach is Braun and Clarke's six phases: familiarization, coding, generating themes, reviewing themes, defining and naming themes, and writing up.
Summary:
|
Thematic analysis is a qualitative research method for identifying, analyzing and reporting patterns of meaning, called themes, across data such as interview transcripts, focus group discussions and open-ended survey responses. A widely used approach is Braun and Clarke's six phases: familiarization, coding, generating themes, reviewing themes, defining and naming themes, and writing up.
Qualitative data is rich, but hard to use. A dozen interviews can produce hundreds of pages of text, and no stakeholder will act on a raw transcript. Thematic analysis closes that gap. This guide covers what it is, how to do it step by step, which type to choose, and where software and AI fit.
What Is Thematic Analysis?
If you are asking what is a thematic analysis, the thematic analysis meaning is simple: it is a way to find patterns of meaning in qualitative data. Instead of counting words, you read closely, label what matters, and group those labels into themes that answer your research question. In research terms, something is thematic when it recurs across participants and carries a shared meaning. That is the core thematic analysis definition.
The most cited framework comes from Virginia Braun and Victoria Clarke. Their 2006 paper, "Using thematic analysis in psychology," set out a clear process and argued that the method is flexible across theories and disciplines.
Teams now also analyze transcripts produced through AI qualitative research, which makes a structured approach more important, not less.
Thematic analysis works with any qualitative material that can be turned into text. Common sources include:
One-to-one interview transcripts
Focus group discussions and their transcripts
Open-ended survey responses
Diaries, field notes and social media content
The format shapes what you analyze. Interviews give depth on one person, while group discussions add interaction between people, so the choice between AI moderated interviews vs focus groups changes what your transcripts look like and what questions they can answer.
Codes vs Themes
Two terms matter most in thematic analysis coding. A code is a short label attached to a data extract. A theme is a broader pattern of shared meaning built from several codes. Codes describe pieces of data, while themes explain what the pieces mean together.
Level | What it is | Example |
Data extract | A quote from a participant | "I took the free month, forgot about it, and only noticed when the charge appeared." |
Code | A short label on the extract | Forgot about trial |
Theme | A pattern across many codes | Billing surprises erode trust |
A common mistake is to treat a topic as a theme. "Pricing" is a topic. "Customers accept the price but resent not being warned about it" is a theme, because it makes a claim about meaning.
Why Thematic Analysis Is Important
The importance of thematic analysis in qualitative research comes down to one thing: it gives structure to human language without flattening it. Researchers keep the nuance of what people said while producing findings others can read in minutes.
The main strengths of thematic analysis are:
Flexibility. It works across theories, disciplines and data types.
Accessibility. The steps are easy to learn, so new researchers can produce credible work quickly.
Decision-ready output. It converts large, unstructured datasets into a small set of clear patterns.
In business research, this surfaces customer needs, motivations and pain points that surveys miss, and helps teams turn qualitative research data into usable insights.
The method has limits too. It takes time, especially during coding, and quality depends on the researcher's skill and reflexivity. Those limits explain why so many teams now use AI qualitative data analysis to speed up the early phases.
When to Use Thematic Analysis
Thematic analysis fits four common situations.
Exploratory research. When little is known about a topic, themes show you where to look next. A team entering a new market might interview ten potential customers and use themes to shape a larger study. Our guide to exploratory research covers how this stage works.
Complex phenomena. Decisions with many moving parts, such as how a household chooses a bank or how a clinician adopts a new tool, rarely fit a survey scale. They are better explored through in-depth interviews, where participants explain context in their own words, and then analyzed thematically.
Large qualitative datasets. Large does not have to mean hundreds of interviews. A systematic review of 23 empirical studies by Hennink and Kaiser found that saturation was reached with roughly 9 to 17 interviews or 4 to 8 focus groups in relatively homogeneous samples. A well-scoped study needs a manageable number of conversations, but each one produces a lot of text to organize.
Interview and focus group data. Thematic analysis is the default method for transcripts. If you are still deciding how to collect that data, this guide on when you need AI moderated interviews explains which research questions benefit most.
Thematic analysis is common across several fields:
UX research: understanding why users abandon a flow
Market research: exploring brand perception and unmet needs
Customer experience: analyzing open-ended feedback at scale
Healthcare: studying patient experiences
Psychology: the field where the method was formalized
Approaches to Thematic Analysis
Approaches are choices the researcher makes along two dimensions. The first is whether the analysis is driven by the data or by existing theory. The second is whether you focus on surface meaning or underlying meaning. You can combine any position on one dimension with any position on the other.
Inductive vs Deductive Thematic Analysis
In inductive thematic analysis, themes emerge directly from the data. You do not start with a fixed framework. You read, code what you see, and build themes from the bottom up. This approach suits exploratory work and unexpected findings.
In deductive thematic analysis, you start with predefined codes or themes drawn from theory, prior research or your research questions. You apply them to the data and then refine them. This is a top-down approach and gives a structured frame, which helps when you must test a specific idea or compare results across studies.
Many studies combine both. A team studying subscription cancellation might start with deductive codes from a retention framework, such as price, usability and content. While coding, they notice guilt-inducing cancellation screens that no framework code covers, so they add an inductive code. The result keeps structure while staying open to surprises.
Semantic vs Latent Themes
Semantic themes capture the explicit, surface-level meaning of what participants said, and suit a clear account of what people reported. Latent themes look beneath the surface at the assumptions and ideas that shape what people say, and require more interpretation.
Take one extract: "I just didn't have time for it anymore."
Semantic coding: Lack of time.
Latent coding: Product never became part of the daily routine, and "time" is a socially easy way to explain low value.
Semantic themes answer "what did people say?" Latent themes answer "what is going on underneath?"
Types of Thematic Analysis
Braun and Clarke describe three main types of thematic analysis. They differ on the use of a codebook, the number of coders and the role of researcher subjectivity.
Type | Codebook | Multiple coders | Researcher subjectivity |
Reflexive | No fixed codebook; codes evolve | Optional, used for reflection, not agreement | Treated as a resource |
Codebook (including framework and template analysis) | Developed early and applied | Often used | Managed within a structure |
Coding reliability | Fixed codebook | Required, with agreement measured | Treated as bias to reduce |
Codebook thematic analysis includes framework analysis and template analysis. A team builds a coding framework early and applies it across the dataset, which suits applied research where several people code. Coding reliability approaches go further: two or more coders apply the same codebook independently, and inter-rater reliability shows how closely they agree.
Reflexive Thematic Analysis
Braun and Clarke later renamed their own approach reflexive thematic analysis to separate it from codebook and reliability styles. The name signals the main idea: the researcher's perspective is part of the analysis.
In reflexive thematic analysis, researcher subjectivity is treated as a resource, not a bias to remove. Themes are not "found" in the data like objects. They are produced through the researcher's engagement with it. Reflexive work therefore values careful reflection on how your assumptions shape your choices over agreement statistics.
How to Do Thematic Analysis: Braun and Clarke's 6 Steps
The Braun and Clarke thematic analysis process has six phases: familiarization, coding, generating themes, reviewing themes, defining and naming themes, and writing up. These thematic analysis steps are iterative, not linear, so you will often move back to earlier ones as your understanding grows.
If you collect data through AI moderated interviews, transcripts are produced automatically, so you can start Step 1 sooner.
Steps 1 and 2: Familiarize Yourself With the Data and Generate Codes
Step 1: Familiarization. Read every transcript at least twice: first for the overall story, then to note what stands out and why.
Step 2: Generating initial codes. Go through the full dataset systematically and attach short labels to meaningful extracts. Code every relevant segment, not only those that fit your expectations. One extract can carry several codes.
Here is a short coded excerpt from the subscription example:
"I only watched one show on it. Once the season ended, I had no reason to open the app." Codes: Single-use motivation; Nothing new to return to
"Cancelling took five screens. I nearly gave up and just kept paying." Codes: Cancellation friction; Reluctant retention
Keep codes specific and close to the data. Many teams now use tools to speed up this phase, and automating thematic analysis for the first pass is one option when the dataset is large.
Steps 3 and 4: Generate and Review Themes
Step 3: Generating themes. Cluster related codes into candidate themes. "Single-use motivation," "Nothing new to return to" and "Stopped opening the app" might form a candidate theme about fading value. A thematic map, which is a simple diagram showing how codes group into themes and subthemes, helps you see relationships and gaps.
Step 4: Reviewing themes. Check each candidate theme against two things: the coded extracts that support it, and the full dataset. Ask whether enough data supports the theme, whether it is distinct from others, and whether it tells a coherent story. Then merge, split or discard themes as needed. Some candidates turn out to be a single code in disguise, and others are too broad and need subthemes.
Steps 5 and 6: Define, Name and Write Up Themes
Step 5: Defining and naming themes. Write a clear definition and scope for each theme in a sentence or two. A definition states what the theme covers and what it does not. Then give each theme a name that makes a point. "Billing surprises erode trust" is stronger than "Billing."
Step 6: Writing up. Present each theme with an analytic narrative and supporting quotes. The narrative explains what the theme means and why it matters, and the quotes give evidence. A quote on its own is not analysis, so always say what the reader should take from it. Tie the themes back to the research question.
Thematic Analysis Example
Here is a worked example. The team interviewed 12 customers who cancelled a subscription app and wanted to know why. The sample size is not arbitrary. In a study of 60 in-depth interviews, Guest, Bunce and Johnson found that saturation occurred within the first twelve interviews, and that the basic elements of the main themes were present as early as six. For a focused question with a similar kind of participant, 12 interviews is a reasonable starting point.
The path from data to theme looked like this:
Raw extract: "I took the free month, forgot about it, and only noticed when the charge appeared."
Code: Forgot about trial.
Candidate theme: Billing problems.
Final theme: Billing surprises erode trust.
Note how the candidate theme was a topic, and the final theme became a claim about meaning. That shift is the real work of Steps 3 to 5.
If you are unsure when to stop recruiting, read more on saturation in qualitative research.
The analysis produced three final themes. The quotes below are illustrative.
Theme | Definition | Supporting quote | Business decision it informs |
Value fades after the first use case | Customers join for one reason and leave when it is satisfied | "Once the season ended, I had no reason to open the app." | Build a 30-day onboarding journey that introduces a second reason to return |
Billing surprises erode trust | Customers feel misled when a trial converts without warning | "I only noticed when the charge appeared." | Send renewal reminders before every charge |
Leaving feels harder than joining | Complex cancellation flows turn mild disappointment into resentment | "Cancelling took five screens." | Simplify the cancellation path and offer a pause option |
Each theme leads to a specific action, and that link is what makes the analysis valuable to a stakeholder.
Thematic Analysis vs Other Qualitative Methods
Thematic analysis is one of several ways to analyze qualitative data. The right choice depends on your goal and the kind of output you need.
Method | Goal | Output | When to use |
Thematic analysis | Identify patterns of meaning | A set of themes with narrative | Most interview and focus group projects |
Content analysis | Categorize and count content | Frequencies of categories | When you need to quantify how often something appears |
Grounded theory | Build a new theory from data | A theory grounded in the data | When no adequate theory exists |
Interpretative phenomenological analysis (IPA) | Understand lived experience | Detailed accounts of individual experience | Small samples exploring how people make sense of a major experience |
Thematic Analysis vs Content Analysis
The core of content analysis vs thematic analysis is meaning versus frequency. Content analysis counts how often categories appear, which helps when you need numbers. Thematic analysis interprets what patterns mean, which helps when you need explanation.
The line is not absolute. Qualitative content analysis also exists and overlaps with thematic analysis when researchers interpret categories rather than only count them. If your question is "how often?", lean toward content analysis. If it is "why?", lean toward thematic analysis.
Thematic Analysis vs Grounded Theory
In grounded theory vs thematic analysis, both involve coding, but the aims differ. Thematic analysis identifies patterns across a dataset. Grounded theory builds a new theory from the data, using constant comparison between cases and codes.
Sampling also differs. Grounded theory uses theoretical sampling, choosing later participants based on what the emerging theory still needs. Thematic analysis typically samples to cover the question and population. Choose grounded theory for a new explanation of a process, and thematic analysis to describe and interpret patterns.
Thematic Analysis Software and AI Tools
You can run thematic analysis with a word processor and a spreadsheet, but qualitative data analysis software (CAQDAS) such as NVivo, ATLAS.ti and MAXQDA handles large datasets more easily.
Inside these tools, the steps look familiar. You import transcripts, highlight extracts and assign codes, organize codes into a code tree or codebook, write memos, retrieve all extracts for a code or theme, and export results for your write-up. This is the workflow people usually mean by NVivo thematic analysis.
When you choose a thematic analysis tool, use these criteria:
Coding workflow: how easy is it to code, recode and merge?
Collaboration: can several researchers work on one project and compare their coding?
Multilingual support: can it handle transcripts in more than one language?
AI assistance: does it suggest codes or summarize extracts, and can you review them?
Export: can you move codes, themes and quotes into reports?
If you are collecting the interviews as well as analyzing them, it also helps to compare AI moderation platforms, because the quality of the transcript affects every later step.
For a closer look at the analysis side, this roundup of AI tools for thematic analysis compares the main options.
Can AI Do Thematic Analysis?
AI thematic analysis can do part of the work. The practical question is which part. Teams exploring AI for qualitative data coding will find the same split.
AI does well at:
Transcribing audio and video
Translating transcripts
First-pass coding
Clustering large datasets into candidate groups
The wider AI picture explains why the gains are uneven. McKinsey's 2026 State of AI survey of 1,719 participants found that 80% of respondents say AI has improved their individual productivity, yet only 37% attribute any EBIT impact to AI at the enterprise level. For researchers, faster coding does not automatically mean better findings. The value shows up when teams rebuild the workflow around the tool.
Human researchers remain essential for:
Interpretation, especially latent meaning
Reflexivity about how assumptions shape the analysis
Checking for themes that are not grounded in the data
Deciding which themes matter to the business
Evidence on accuracy supports this caution. In a case study using Discord community data, GPT-4 reached a Cohen's kappa of 0.57 in inductive coding, which the authors describe as moderate agreement with human coders. That helps a first draft, but a researcher still has to review the codes.
Governance is the other half. Deloitte's 2026 State of AI in the Enterprise report, based on 3,235 leaders, found that only one in five companies has a mature governance model for autonomous AI agents. Research teams should not wait for that gap to close, so decide who reviews AI-generated codes and how disagreements get resolved.
Common Mistakes and Best Practices in Thematic Analysis
Most weak thematic analysis fails in a few predictable ways:
Themes that are only topic summaries. "Pricing" describes a subject, not a finding.
Too many themes. A long list usually means the analysis stopped at coding.
Quotes without interpretation. Quotes are evidence, not analysis.
Mixing reflexive and reliability logic. Do not claim reflexive depth while using inter-rater statistics as proof of quality.
Strong work follows a few rigor practices:
Keep an audit trail. Record how codes changed and why themes were merged or split.
Use a reflexive journal. Note how your assumptions shape your reading.
Write clear theme definitions. Each theme needs a scope someone else could apply.
Report transparently. State your approach, coding process and limits.
Keep human oversight over AI-generated codes. Review them against the original extracts before they become themes.
Running Thematic Analysis at Scale With AI-Moderated Interviews
The slowest parts of thematic analysis are the early ones: transcribing, reading and first-pass coding. Teams running studies across markets feel this most, because every extra interview multiplies the work before interpretation begins.
AI-moderated interviews shorten those phases. With AI moderated qualitative interviews, participants are interviewed by an AI moderator, and the platform produces automatic transcription, translation and theming. Researchers can then spend their time on the part that needs them: checking themes, interpreting meaning and deciding what to do next.
Language is a second constraint. With 70+ languages supported, Decode helps teams run multilingual research and compare themes across markets without a separate translation step.
Emotion adds another layer. Facial coding at 90%+ accuracy, built on 62 facial expressions, lets researchers see the emotional response around a theme, such as the frustration that shows up when a participant describes a cancellation flow.
Decode also includes a generative AI research assistant that summarizes findings and surfaces patterns, so teams can move from raw conversation to decision-ready themes faster. The platform is backed by 17 patents and trusted by 150+ global brands.
The goal is to reduce the time between data collection and decision-ready themes without losing human interpretation. Decode by Entropik is built around that principle.
Frequently Asked Questions
1. What is thematic analysis?
Thematic analysis is a qualitative research method for identifying, analyzing and reporting patterns of meaning, called themes, across data such as interviews, focus groups and open-ended survey responses.
2. When should you use thematic analysis?
Use it to explore a new topic, understand a complex phenomenon, organize a large qualitative dataset, or analyze interview and focus group data, especially when your question asks why.
3. What is the difference between inductive and deductive thematic analysis?
Inductive thematic analysis builds themes from the data without a predefined framework. Deductive thematic analysis starts with codes or themes drawn from theory or research questions. Many studies combine both.
4. How does thematic analysis differ from content analysis?
Content analysis counts how often categories appear, so it focuses on frequency. Thematic analysis interprets the meaning behind patterns, so it focuses on explanation. Qualitative content analysis also exists and overlaps with thematic analysis.
5. What are the 6 steps of thematic analysis according to Braun and Clarke?
The six steps are familiarization, generating initial codes, generating themes, reviewing themes, defining and naming themes, and writing up. The process is iterative, so you often return to earlier steps.
6. What is reflexive thematic analysis?
Reflexive thematic analysis is Braun and Clarke's approach in which researcher subjectivity is treated as a resource, not a bias to remove. Reflection on assumptions is central to quality.
7. Can AI chatbots do a thematic analysis?
A general-purpose AI model can help with first-pass coding and summarizing, but it should not replace the researcher. Its codes need checking against the original extracts, and interpretation still needs a human.
8. Can I use qualitative analysis software for thematic analysis?
Yes. CAQDAS tools support the thematic analysis workflow: importing transcripts, coding extracts, organizing a code tree, writing memos and retrieving coded data. The software supports the process, but the analytic decisions remain yours.
Related Topic:
