Topics, themes and prompts
Free-text answers (verbatims) are automatically classified so you can see what customers are talking about at scale. This page covers the taxonomy you configure — topics and themes — and the prompts that drive the classification.
Topics and themes
Open Topics & themes to manage your taxonomy.
- A topic is a specific subject a verbatim can be about — for example "wait time" or "staff friendliness".
- A theme is a broader grouping that several topics roll up into — for example "service speed" or "people".
When a verbatim is coded, it's tagged with the topics it matches, and those topics carry their parent themes. This two-level structure lets you analyse feedback both at a fine-grained level (topics) and at a summary level (themes).
Use this page to add, rename, and organise topics, and to set which theme each topic belongs to. A well-maintained taxonomy is what makes verbatim analytics meaningful.
Prompts
Open Prompts to manage the LLM prompts that power automatic coding and other text analysis.
- Versioned — prompts are kept as versions so you can change wording and track what was in effect over time.
- Template tags — prompts use placeholders that are filled in at run time:
| Tag | Replaced with |
|---|---|
%topics% | Your current list of topics, so the model classifies against your taxonomy. |
%verbatims% | The verbatim text being analysed. |
Editing a prompt creates a new version. Keep prompts clear and aligned with your topic list so classification stays accurate.
How they work together
- You define topics and group them under themes.
- A prompt (with
%topics%and%verbatims%) tells the model how to classify each verbatim against that taxonomy. - Coded verbatims feed the verbatim analytics — topic and theme breakdowns, trends, and word clouds.
Next steps
- The Verbatim analytics section — see the coded results in action.
- Features & audit — track changes to your configuration.