Topics and themes
Sentiment tells you how customers feel. Topics and themes tell you what they are talking about. They give your free-text feedback a consistent structure, so you can count and compare the things customers raise instead of reading every comment by hand.
How the taxonomy is organised
The structure has two levels:
- A theme is a broad area of feedback — for example Service, Pricing, or Product.
- A topic sits underneath a theme and names something more specific — for example Wait time and Staff friendliness might both sit under the Service theme.
Every topic belongs to exactly one theme. Each topic also carries a description, and that description does real work — see below.
Setting up topics and themes
You manage the taxonomy in the admin area, not on the Verbatim page itself.
- Go to Admin → Topics & themes (there is a Manage topics & themes link on the Coding tab that takes you straight there).
- Create your themes first.
- Add topics under each theme.
- Give every topic a clear description of what kind of comment belongs to it.
Keep your taxonomy small and meaningful to begin with. A handful of themes with a few topics each is far easier to act on than dozens of overlapping codes.
How the description drives auto-coding
When a comment arrives, the AI auto-coder reads it and decides which topics apply. It does this by comparing the comment against your topic descriptions — so the description is effectively the instruction you give the AI.
A vague description such as "about service" produces vague results. A specific one such as "Comments about how long the customer waited on hold, in a queue, or for a callback" gives the AI a clear target and far more accurate tagging.
| Description quality | Example | Result |
|---|---|---|
| Vague | "service stuff" | Inconsistent, over-broad tagging |
| Specific | "Mentions of staff being rude, unhelpful, or dismissive" | Accurate, repeatable tagging |
Because the description steers the AI, refining a topic's description in Admin is the main lever you have for improving auto-coding quality. After changing a description, re-run Auto-tag with AI on the Coding tab to apply the improved logic to existing comments.
Reviewing and correcting coding (the Coding tab)
Auto-coding gets you most of the way, but you stay in control. The Coding tab on the Verbatim page is your workbench for manual coding:
- Open the Verbatim page and select your survey.
- Switch to the Coding tab.
- Each comment is listed with its text and, beneath it, the topics grouped under their themes.
- Click a topic to apply it to the comment; click it again to remove it. Applied topics are highlighted.
- To let the AI fill in gaps, click Auto-tag with AI — it tags comments that are not yet coded and reports how many tags it applied.
This is the same classifier that runs automatically on capture, so manual and automatic coding stay consistent.
Seeing the result
Once comments are coded, the Topics tab visualises them: each theme gets its own chart showing the volume of comments per topic. See Word cloud and charts for how to read it.