Qualitative data analysis methods - thematic, content, narrative, grounded theory, framework - with the step-by-step process and how to run each one.
Qualitative data analysis is the work of turning open-ended answers, transcripts, and documents into themes and findings. The methods — thematic, content, narrative, grounded theory — are sound; the catch was always labor, so they were applied to a sample and the rest skipped. This guide covers the methods, the step-by-step process, and how AI lets every method run on the full dataset.
By Unmesh Sheth · Founder & CEO, Sopact · Updated May 2026
Qualitative data analysis (QDA) is the process of examining non-numerical data — open-ended answers, interview transcripts, documents, observations — to identify themes, patterns, and meaning. It works by coding the data against a defined scheme, grouping codes into themes, and interpreting what they show against the research question. The main methods are thematic, content, narrative, and grounded-theory analysis.
The process of reading non-numerical data — text, transcripts, documents — for themes, patterns, and meaning, and reporting what they show.
The named approaches — thematic, content, narrative, grounded theory, framework, discourse — each suited to a different kind of question.
The six steps every method runs through: familiarize, define the codebook, code, group into themes, interpret, and report.
Tagging segments of text with codes, then grouping codes into themes. The step where the analysis is won or lost.
Thematic analysis, content analysis, grounded theory — the qualitative data analysis methods are decades old and well proven. None of them is the problem. The problem was always labor: a person had to read and code everything by hand, so in practice the methods were applied to a sample and the rest of the data went unread. AI removes the labor, not the method.
The method was never the bottleneck. The labor of reading everything was.
What changes is not the method. It is that the method finally runs on all of the data.
Thematic analysis is still thematic analysis. What the AI era changes is reach and reproducibility: every response coded, against one codebook, with a result that holds up on a re-run and a reviewer who can trace it.
Qualitative data analysis is not one technique. Each method below answers a different kind of question. They share one engine — coding text against a defined scheme — but differ in what they are looking for.
Codes the data, groups codes into themes, and maps the patterns of meaning across the dataset. Flexible across data types — the default for open-ended survey and interview analysis.
Categorizes content against a fixed scheme and counts how often each category appears. Used when the frequency of a theme matters, not only its presence.
Interprets the stories people tell — how an account is structured and what it reveals — keeping whole accounts intact rather than breaking them into codes.
Builds an explanation up from the data: coding drives categories, categories drive an emerging theory, and collection continues until the theory is saturated.
Charts cases against themes in a matrix — common in applied and policy research, where many cases must be compared against a set structure.
Examines how language itself does work — how wording, framing, and context shape meaning. Used where the how of what was said matters as much as the what.
Whichever method you choose, the analysis runs through the same six steps. The third step — the codebook — is the one that decides whether the result is consistent and can be re-run.
Read through the data — transcripts, responses, documents — to get a feel for what is there before coding anything.
Set the codes: the concepts, categories, or themes to look for, each with a clear definition. Inductive, deductive, or a blend of both.
Apply the codebook to every segment of every item. The step that, done by hand, used to force a sample — and now does not.
Cluster related codes into themes, and review each theme against the full dataset to check that it holds.
Work out what the themes mean against the research question — the reason behind the pattern, not just the pattern.
Present the themes with the evidence: frequency, example quotes, and a path from each theme back to its source.
Step two — defining the codebook — can run two ways. The choice shapes the whole analysis, and most real projects use a blend of the two.
Codes emerge from the data. You read first and name what you find, letting the themes come from the responses themselves. The approach behind grounded theory and exploratory work — strong when you do not yet know what you are looking for.
Codes are defined in advance, from a theory or a framework, then applied to the data. The approach behind framework analysis — strong when the questions are set and cases must be compared the same way.
Few projects are purely one or the other. Most applied analysis starts with a deductive codebook — the themes you already need to track — and extends it inductively as new themes surface in the data. A good tool lets the codebook grow without losing consistency on what was already coded.
The worry about AI in qualitative data analysis is that automation replaces the method with a guess. It does not have to. The method stays exactly as it is — what AI changes is who applies the codebook, and how much of the data gets read. The safeguard is three things.
The team defines the codebook — the method's coding scheme. The AI applies it; it does not invent it. The analysis stays yours.
Every transcript and response is coded as it lands, against that codebook — the whole dataset, not the sample hand-coding forced.
The same input gives the same codes on a re-run, and every theme is cited to its source line — so AI-generated themes can be validated, not just trusted.
For the tools that do this — legacy CAQDAS and the AI-native option — see qualitative data analysis software.
Qualitative data analysis methods matter most to the teams whose findings get scrutinized. The audience is broad, and for each, coding the whole dataset instead of a sample changes a different number.
Teams coding open-ended survey responses, reviews, and check-in comments to explain customer behavior.
Teams applying a named method — thematic, framework — to pre- and post-training open-ended responses.
Teams applying a coding framework to essays, narratives, and reviewer notes a decision will rest on.
Works the same way for fellowship reviews, accelerator selection, and program intake — the same methods, different data.
Bring a set of transcripts or open-ended responses. We help you pick the method, define the codebook, and run it live — the whole dataset coded, every theme cited.
Qualitative data analysis (QDA) is the process of examining non-numerical data — open-ended answers, interview transcripts, documents, and observations — to identify themes, patterns, and meaning. It works by coding the data against a defined scheme, grouping codes into themes, and interpreting what they show against the research question. The main methods are thematic analysis, content analysis, narrative analysis, and grounded theory.
The main qualitative data analysis methods are thematic analysis (recurring themes across the data), content analysis (systematic categorizing and counting), narrative analysis (interpreting the stories people tell), grounded theory (building a theory up from the data), framework analysis (a matrix of cases against themes, common in applied and policy research), and discourse analysis (how language is used). Thematic and content analysis are the most widely used.
Qualitative data analysis runs in six steps: familiarize yourself with the data; define a codebook of the concepts to look for; code the data against that codebook; group the codes into themes; interpret what the themes mean against the research question; and report the themes with the evidence behind them. The codebook step decides quality — defined once, applied the same way to every item.
Choose the method by the research question. Use thematic analysis to map patterns of meaning, content analysis when you need the frequency of categories, narrative analysis to keep whole accounts intact, grounded theory to build an explanation where none exists yet, and framework analysis to compare many cases against a set structure. Most applied work uses thematic or framework analysis; the methods can also be combined.
Thematic analysis is the most widely used qualitative data analysis method. It codes the data, groups the codes into themes, and reviews those themes against the full dataset to identify the patterns of meaning that run through the responses. It is flexible across data types and research questions, which is why it is the default for open-ended survey and interview analysis.
Content analysis is a qualitative data analysis method that systematically categorizes content against a fixed coding scheme and counts how often each category appears. It is used when the frequency of a theme matters, not only its presence. Content analysis can be qualitative, quantitative, or both, and it is valued for producing results that are countable and comparable across a dataset.
Grounded theory is a qualitative data analysis method that builds an explanation up from the data itself, rather than testing a theory set in advance. Coding drives the categories, the categories drive the emerging theory, and data collection continues until the theory is saturated. It fits research questions where you do not yet have a framework and need one to come from the evidence.
Coding is the core step of qualitative data analysis: tagging segments of text with codes — short labels for a concept, category, or theme. Codes are then grouped into themes. Coding can be inductive, where codes emerge from the data, or deductive, where codes are defined in advance. A consistent, defined codebook is what keeps coding from drifting between coders or over time.
Inductive coding lets codes emerge from the data — you read first and name what you find, the bottom-up approach used in grounded theory and exploratory work. Deductive coding starts from a codebook defined in advance, from theory or a framework, and applies it to the data — the top-down approach used in framework analysis. Most applied analysis is a blend: a starting codebook, extended as new themes appear.
Analyze open-ended survey responses with thematic or content analysis: define a codebook of the themes you expect and want to detect, code every response against it rather than a sample, group codes into themes, and report each theme with its frequency and example quotes. Reading every response — instead of skimming a sample — is what turns open-ended answers from a word cloud into a finding.
Yes. AI does not replace the method — thematic analysis is still thematic analysis — it removes the labor. A model applies the codebook to every transcript and response as it arrives, far faster than hand-coding. Automation is reliable when the codebook is fixed, the same input gives the same codes on a re-run, and every theme is cited to its source, so a reviewer can validate the AI-generated themes.
Quantitative data analysis works on numbers, using statistics to find magnitude and pattern — it answers how much and how many. Qualitative data analysis works on words, using coding and thematic methods to find meaning — it answers why and how. They answer different questions and are strongest read together. For the combined practice, see qualitative and quantitative analysis.
This page is the methods and the process. The guides below cover the broader practice, the data, the tools, and the pillar that joins the qualitative half to the quantitative one.
A working session, not a demo. Bring a set of interview transcripts or open-ended responses. We help you choose the method, define the codebook, and run the analysis live — the full dataset coded on arrival, every theme cited to its source. You leave with a method chosen on purpose, a working codebook, and a coded sample of your own data.
Live walkthrough · 30 min · with Unmesh Sheth, Founder & CEO · bring transcripts or open-ended data you want coded
