Digital Thematic Analysis — Identifying Patterns in Digital and Online Data

They overlap but differ in emphasis. Qualitative content analysis typically involves a structured coding scheme applied to classify and sometimes count content categories. Digital thematic analysis, following Braun and Clarke, focuses on identifying and interpreting patterned meanings across the corpus rather than categorising or tallying content. Thematic analysis is more interpretive and inductive; qualitative content analysis is often more systematic and closer to a classification task.

Qualitative data analysis software (NVivo, ATLAS.ti, MAXQDA) can organise, search, and retrieve coded segments efficiently and is appropriate for large digital corpora. Automated text mining, topic modelling, or machine-learning classifiers are different tools that produce different outputs — they can be used as a preliminary corpus exploration step but should not replace the interpretive coding and theme-development phases that constitute thematic analysis.

Institutional requirements vary, but most qualitative methodologists and research ethics frameworks recommend seeking ethics review whenever human-generated data are analysed, even if publicly posted. The key considerations are: whether participants could be identified through their posts, whether the community norm treats the space as public or private, and whether the platform's terms of service permit research use. Anonymisation of usernames and paraphrasing of distinctive quotes is standard practice.

There is no universal rule, but the operative criterion is the same as for interview-based thematic analysis: sufficient richness and diversity to reach thematic saturation — the point where additional data yield no substantially new themes. For focused phenomena in niche communities, saturation may be reached with a relatively small corpus; for broad social phenomena on major platforms, a larger corpus is needed. Always prefer a well-delimited, thoroughly read corpus over an enormous but superficially skimmed one.

Extracts should be anonymised (usernames replaced with pseudonyms or generic labels such as 'User 1'). Highly distinctive phrasing that could be traced through a search engine should be paraphrased. The platform, thread, and date of the post should be noted in your methods section to enable readers to assess the context, even if the specific post cannot be attributed publicly. Follow your institution's ethics approval conditions and relevant disciplinary reporting conventions.