What is thematic analysis? A developer's guide

Thematic analysis, defined

Thematic analysis takes unstructured text — interview transcripts, survey open-ends, customer reviews, support tickets, exit interviews, public-consultation responses — and produces a structured account of what people are actually saying. Rather than counting keywords or scoring sentiment in isolation, it identifies codes (short labels attached to specific excerpts) and clusters those codes into themes that capture a recurring idea, supported by evidence you can point to.

The most widely cited framework is Braun and Clarke's reflexive thematic analysis, a six-phase process that has become the reference point for rigorous qualitative coding. Its value is defensibility: every theme traces back to the text that produced it.

The six phases of reflexive thematic analysis

1. 1. Familiarisation. Read the data closely and note first impressions.
2. 2. Generating codes. Attach concise labels to meaningful segments of text.
3. 3. Constructing themes. Group related codes into candidate themes.
4. 4. Reviewing themes. Check each theme against the coded data and the full dataset.
5. 5. Defining and naming. Write a crisp description and name for each theme.
6. 6. Reporting. Present themes with representative quotes and prevalence.

Manual coding vs. a thematic analysis API

Done by hand, thematic analysis is slow but trustworthy: a researcher can code a few hundred responses well. The problem appears at scale. Coding ten thousand survey open-ends manually is impractical, and a quickly-written prompt to a language model drifts — theme names change between runs, counts don't reconcile, and you can't show the quote behind a finding.

A dedicated theme extraction API solves the engineering around the model: a stable response schema, codebooks that persist so this quarter is comparable to last quarter, confidence scores, supporting text spans, and async batch processing for large datasets. You send text and get back codes, themes, quotes, sentiment, and confidence as JSON — without owning the evaluation and consistency tuning yourself.

When to reach for an API

Use an API when you are shipping a product feature rather than running a one-off study: a review-analysis dashboard, an exit-interview tool, a citizen-feedback platform, a course-feedback feature, or any product that needs to turn large volumes of open-ended text into reliable themes. See use cases and how the API works for concrete request and response examples.