Thematic analysis in qualitative research: overview and examples

Thematic analysis is a method for identifying, analysing, and reporting patterns of meaning across a qualitative dataset. It is one of the most widely used approaches in qualitative research precisely because it is flexible: it can be applied to almost any kind of text data (interview transcripts, focus group discussions, open survey responses, documents) and can be conducted inductively (letting themes emerge from the data) or deductively (applying a pre-existing framework).

Understanding what thematic analysis is (and what it is not) helps researchers choose it for the right reasons and apply it well.

What is thematic analysis?

A theme in thematic analysis is a pattern of shared meaning across a dataset. It is not simply a topic that comes up frequently; it is a pattern that captures something significant in relation to the research question.

For example, in a study of employees' experiences of hybrid working, "home environment" might be a topic (it comes up a lot), but "the boundary between professional identity and domestic life becoming harder to maintain" might be a theme: a pattern of meaning that appears across multiple data points and says something specific about what employees experience.

This distinction matters because thematic analysis is interpretive, not merely descriptive. The researcher is not just cataloguing what participants said; they are identifying what those statements mean in the context of the research question.

How does thematic analysis differ from other qualitative methods?

Method	What it analyses	How themes are generated	Level of interpretation
Thematic analysis	Meanings across text data	Inductively or deductively	Moderate to high
Content analysis	Frequency of content elements	Deductively (usually)	Low to moderate
Grounded theory	Social processes	Inductively, from data	High (generates theory)
Discourse analysis	Language use and power	From linguistic and social context	Very high
Framework analysis	Predefined dimensions	Deductively	Moderate

Thematic analysis occupies a middle position: more interpretive than content analysis, less theoretically ambitious than grounded theory. This makes it accessible to a wide range of researchers and applicable to a wide range of questions. It is also the method most commonly adapted for AI-assisted analysis, because the coding and pattern-identification steps are well-suited to computational processing.

The 6 phases of thematic analysis

The most widely cited framework is Braun and Clarke's six-phase approach, first described in 2006 and developed significantly since. These phases are iterative, not strictly sequential.

Phase 1: Familiarising yourself with the data

Read and re-read the data. For interview transcripts, this means reading each transcript at least twice before beginning to code. The goal is immersion: developing an overall sense of what is in the data before beginning to impose structure on it.

Many researchers take initial notes during this phase, jotting down ideas, observations, and early analytical thoughts. These are not codes yet; they are the raw material of later coding.

Phase 2: Generating initial codes

Coding is the process of labelling units of data with a short descriptor that captures something significant about them in relation to the research question. A code might be applied to a single sentence, a paragraph, or a longer passage.

At this stage, the goal is to generate codes without over-selecting. Err on the side of coding too much rather than too little. One piece of data can receive multiple codes if it is relevant to more than one pattern.

In reflexive thematic analysis (Braun and Clarke's current framing), codes are understood as interpretive constructs rather than neutral descriptions: the code you apply to a piece of data reflects an analytical decision, not an objective classification.

Phase 3: Searching for themes

Themes are higher-order patterns that group related codes. At this phase, the researcher reviews all the generated codes and identifies which ones can be grouped into broader themes.

This is not a purely logical exercise. It requires interpretive judgement: which of these codes actually share a meaningful connection in relation to the research question, and which are only superficially similar? A theme that groups together everything that mentions "technology" is probably a topic, not a theme. A theme that groups together "feeling replaced by automated systems," "anxiety about AI tools making roles redundant," and "manager communicating change without explaining the rationale" is capturing a pattern of meaning about how technological change is experienced emotionally.

Phase 4: Reviewing potential themes

Go back to the data with your candidate themes and check them against the evidence. Does each theme have enough supporting data? Does the data actually support the theme as you have defined it, or are you grouping disparate things together?

This phase often results in splitting themes that are too broad, merging themes that are too similar, or dropping themes that turn out to have insufficient support in the data. It is where analytical quality is built or lost.

Phase 5: Defining and naming themes

Write a brief description of each theme that captures its essence: what pattern of meaning it captures, what evidence supports it, and how it relates to the research question. The name should be descriptive of the theme's content, not just the topic it covers.

Topic label (avoid): "Manager communication" Theme name (better): "The gap between what managers know and what they share: and the conclusions employees draw from silence"

Phase 6: Producing the report

Write up the findings as a narrative, using themes as the organising structure. Each theme should be introduced, explained with specific examples and quotes from the data, and connected to the research question. The write-up should make the analytical reasoning visible, not just present conclusions.

For a detailed guide to all six phases, see thematic analysis: a complete guide.

Inductive vs deductive thematic analysis

Inductive (bottom-up) thematic analysis starts from the data and allows themes to emerge without a pre-existing framework. It is appropriate when you are exploring a topic with limited prior theory, or when you want to capture the participant's own framing without imposing the researcher's categories.

Deductive (top-down) thematic analysis starts from a pre-existing framework (a theory, a model, a previous study's coding scheme) and applies it to new data. It is appropriate when you want to test or refine existing theory, or when comparability with previous research matters.

Most thematic analyses combine both: a deductive starting framework informed by prior research, applied inductively to new data with room for themes that fall outside the framework.

Examples of thematic analysis

Example 1: employee wellbeing research (academic)

A researcher conducts 24 semi-structured interviews with employees across three organisations about their experience of hybrid working. The research question is: how do employees experience the boundary between professional and personal life in hybrid settings?

Initial codes include: home office setup, partner/family interruptions, always-on expectations, difficulty switching off, status signalling, missing informal office conversations, schedule flexibility, childcare logistics, and productivity perceptions.

Thematic analysis produces four themes:

Permeability of the home-work boundary: the experience of work spilling into personal time and vice versa, and how employees manage (or fail to manage) this
Visibility and trust: anxiety about whether managers trust remote workers to be productive, and the compensatory behaviours this produces
The social texture of work: what employees miss from the office environment and what they do not miss
Autonomy as double-edged: the ways in which flexible scheduling is experienced as both a benefit and a source of pressure

Each theme is supported by 6-12 specific quotes from the transcripts. The write-up discusses what each theme means for organisational policy and management practice.

Example 2: customer churn research (business)

An HR software company runs 18 exit interviews with customers who churned in the previous quarter. The research question is: what were the drivers of the decision to cancel, and what could have prevented it?

Codes include: onboarding friction, integration limitations, pricing model confusion, support response speed, product gaps vs competitors, champion leaving the customer organisation, procurement process, and alternative products trialled.

Thematic analysis identifies three themes:

The expectation gap: the difference between what customers expected when purchasing and what they found in practice, particularly around integration complexity
The moment of re-evaluation: the specific triggers (a new vendor in procurement, a product update that broke a workflow, a failed support ticket) that moved customers from dissatisfied to actively evaluating alternatives
The absence of a relationship: customers who churned had limited contact with anyone at the company beyond their initial onboarding, and described no relationship to pull them toward staying

These findings directly inform changes to the customer success programme: specifically, the design of an ongoing touchpoint structure for the 6-18 month period after onboarding, which was identified as the high-risk window.

Example 3: policy consultation (public sector)

A government team receives 4,000 written responses to a public consultation on proposed changes to housing policy. A thematic analysis of a stratified sample of 400 responses identifies six themes across the response corpus, with supporting quotes and frequency data. The six themes are: affordability concerns, planning system complexity, community impact and infrastructure, environmental considerations, property rights, and housing for key workers. Cross-tabulation by respondent type (individual citizen vs developer vs local authority vs housing association) shows markedly different theme profiles across groups. For this type of application, see stakeholder consultation analysis.

Reflexive thematic analysis vs codebook thematic analysis

Braun and Clarke's current framing distinguishes between two distinct variants:

Reflexive thematic analysis treats themes as the researcher's interpretive construction rather than patterns that objectively exist in the data. The researcher's positionality, assumptions, and analytical decisions are foregrounded and disclosed. This approach is common in academic qualitative research and is associated with a more epistemologically sophisticated analysis.

Codebook thematic analysis uses a pre-developed coding scheme that is applied consistently across coders and datasets. Inter-rater reliability is assessed to ensure the coding scheme is being applied consistently. This approach prioritises reproducibility and is more common in applied research, multi-coder studies, and AI-assisted analysis.

For most business research applications, codebook thematic analysis is more practical. For academic research with a strong qualitative epistemological commitment, reflexive thematic analysis is the more appropriate approach.

Thematic analysis and AI

Designing AI that augments qualitative researchers instead of replacing them

Frequently asked questions

Is thematic analysis qualitative or quantitative?

Thematic analysis is a qualitative method. It interprets meaning rather than measuring frequency. While it is possible to count how often a theme appears across a dataset (which adds quantitative information about prevalence), the core analytical work is qualitative: interpreting what themes mean in relation to the research question.

How many themes should a thematic analysis have?

There is no fixed rule. Most published thematic analyses identify 3-7 themes. Fewer than 3 may indicate under-analysis; more than 7 may indicate the researcher has not consolidated themes sufficiently. The right number depends on the scope of the research question and the complexity of the dataset. A study of a narrow phenomenon in a homogeneous sample may have 3-4 themes. A study of a complex, multi-faceted experience across a diverse sample may have 6-8.

Can thematic analysis be used for survey data?

Yes. Any text data can be analysed thematically, including open-ended survey responses. The process is identical to interview analysis, though open-text survey responses are typically shorter and less contextually rich than interview transcripts. The resulting themes should be interpreted in light of the survey context: a response to a specific survey question provides less latitude for the participant's own framing than an open-ended interview.

How does thematic analysis handle conflicting data?

Conflicting data should not be averaged away or excluded. If two participants describe the same experience in contradictory terms, both descriptions are analytically significant. Good thematic analysis accounts for variation and contradiction within themes: "most participants described X, but a minority described an experience that contradicted this, particularly in the context of Y." This variation often points to the most analytically interesting dimensions of the phenomenon being studied.

What software is used for thematic analysis?

Traditional options include NVivo, MAXQDA, ATLAS.ti, and Dedoose. AI-assisted options include Skimle, which automates initial coding while maintaining full traceability from every theme to source quotes. For a full comparison, see qualitative data analysis tools complete comparison.

Ready to run thematic analysis on your qualitative data without weeks of manual coding? Try Skimle for free and see AI-assisted thematic analysis in action: every theme traced back to specific quotes, across your full dataset.

About the authors

Henri Schildt is a Professor of Strategy at Aalto University School of Business and co-founder of Skimle. He has published over a dozen peer-reviewed articles using qualitative methods, including work in Academy of Management Journal, Organisation Science, and Strategic Management Journal. His research focuses on organisational strategy, innovation, and qualitative methodology. Google Scholar profile

Olli Salo is a former Partner at McKinsey & Company where he spent 18 years helping clients understand the markets and themselves, develop winning strategies and improve their operating models. He has done over 1000 client interviews and published over 10 articles on McKinsey.com and beyond. LinkedIn profile