This guide explains how the Conversation Score is calculated to measure the overall quality of your AI Employee's interactions. It details how star ratings translate into the score and provides calculation examples to help you interpret the metric.
Overview
The Conversation Score is the average Conversation Success Score (CSS) across all assessed sessions for a given customer. It is calculated per customer and displayed on the Organizations page as "Conversation Score."
Conversation Score = SUM(CSS ratings) ÷ Number of assessed sessions
Higher scores (closer to 100% or 1.0) indicate better outcomes, while lower scores suggest that the AI Employee may need attention or logic adjustments.
Only sessions that have been given a star rating are included. Unassessed sessions and test sessions are excluded from the calculation entirely — they do not appear in the numerator or the denominator.
How star ratings map to CSS
Each session can be rated using a 1–5 star widget on the Sessions page (the "Conversation Quality" column). The star rating is converted to a CSS value using the formula:
CSS = 0.2 × Star Rating
Star rating | CSS value |
5 stars | 1.0 |
4 stars | 0.8 |
3 stars | 0.6 |
2 stars | 0.4 |
1 star | 0.2 |
Unassessed | Excluded from calculation |
Example 1: High performance
An AI Employee has completed 28 total sessions, of which 20 have been assessed with star ratings:
15 sessions rated 5 stars (CSS = 1.0 each)
3 sessions rated 4 stars (CSS = 0.8 each)
2 sessions rated 3 stars (CSS = 0.6 each)
8 sessions unassessed (excluded)
Sum CSS values: (15 × 1.0) + (3 × 0.8) + (2 × 0.6) = 15 + 2.4 + 1.2 = 18.6
Divide by assessed sessions: 18.6 ÷ 20 = 0.93
Result: The Conversation Score is 93%. This indicates the agent is performing well.
Example 2: Low performance
An AI Employee has completed 28 total sessions, all of which received a 1-star rating (CSS = 0.2 each):
Sum CSS values: 28 × 0.2 = 5.6
Divide by assessed sessions: 5.6 ÷ 28 = 0.2
Result: The Conversation Score is 20%. This indicates the agent requires immediate attention to resolve interaction friction.