Skip to main content

Sentiment

Review generated sentiment signals for conversations in a selected period.

Accessing Sentiment

  1. Open your agent in the dashboard
  2. Go to Analytics > Sentiment

Choosing a Period

The same period controls as elsewhere in analytics: this month, last month, last 7 days, last 30 days, or a custom from/to range.

What Sentiment Measures

Sentiment is a signal generated from the conversation itself — how the exchange went, not a survey response. Treat it as a way to find conversations worth reading, not as a satisfaction score you report to the board.

How to Use It

Find the failures

Sort toward negative sentiment and read those transcripts. Negative conversations usually have one of three causes:

  1. Missing knowledge — the agent could not answer. Fix with training.
  2. Wrong knowledge — the agent answered confidently and incorrectly. Fix with a Q&A pair, which takes priority over document content.
  3. Wrong expectation — the visitor wanted a human. Consider the Technical Support agent so escalation is structured rather than a dead end.

Watch the trend

A single bad conversation is noise. A week of declining sentiment after a release is a signal.

Cross-reference topics

Combine with Topics: frequency tells you what matters, sentiment tells you what hurts. The intersection is your work queue.

Exporting

Export CSV takes the sentiment data into a spreadsheet for reporting or deeper analysis.

No Data Yet

Sentiment needs conversations in the selected window. New agents and narrow ranges show an empty state.

Next Steps