Most AI agent dashboards show a thumbs-up count and a thumbs-down count. That's a start. But a raw count doesn't tell you whether things are getting better or worse, which answers are causing the most dissatisfaction, or whether a knowledge base update actually improved anything.

Satisfaction Score Trends turn feedback data into something you can act on.

From raw feedback to a trend line

Every message in Companin can receive a thumbs up or thumbs down from the user. The Satisfaction Score dashboard aggregates this feedback into a CSAT percentage (positive ratings / total ratings) and plots it as a weekly trend line.

What you can see:

Overall CSAT % for the current period
Week-over-week change — is satisfaction improving or declining?
CSAT by topic — which categories of questions get the best and worst ratings?
Lowest-rated responses — a ranked list of the AI answers that received the most thumbs-down, with the actual response text

Using CSAT to validate knowledge base updates

One of the hardest things to know after updating your knowledge base is whether the update actually helped.

CSAT trends give you a before/after comparison. If you updated your pricing documentation on a Tuesday and satisfaction for pricing-related questions improved the following week, that's validation the update worked. If it didn't improve, you know something caused a problem and you can investigate.

This closes a feedback loop that most teams are missing. Without it, you're making changes and hoping. With it, you're making changes and measuring.