CSAT (Customer Satisfaction Score)
Customer Satisfaction Score (CSAT) is a survey-based measure of how satisfied customers say they are with a specific interaction, outcome, product, or service experience.
How CSAT is measured
CSAT usually comes from a short question asked after the relevant experience. A team can report the average rating or the share of responses that meet its definition of satisfied. Because survey scales and scoring rules vary, the question wording, response options, scoring method, and reporting period should accompany the result.
The survey needs to name what the customer is rating. Satisfaction with the agent's courtesy is not the same as satisfaction with the resolution, and satisfaction with one call is not the same as satisfaction with the overall relationship. A precise question produces a more actionable score.
Why CSAT matters for AI phone calls
CSAT adds the caller's perspective to operational measures such as answer speed, handle time, and containment. An AI phone agent may complete a workflow quickly while leaving the caller uncertain, or it may take longer and still provide a clear, successful outcome. The satisfaction question helps reveal that difference.
For diagnosis, segment responses by call reason, outcome, transfer status, and agent configuration. Compare the feedback with transcripts, call dispositions, and quality reviews to identify why callers responded as they did. A satisfaction drop limited to transferred calls points to a different problem than a drop across every call type.
Survey results also carry selection bias. People who respond may differ from those who do not, and repeatedly asking frequent callers can distort the picture. Keep the request brief, control survey frequency, and review response coverage before treating the score as representative.
CSAT is not a substitute for resolution or effort. A caller can report satisfaction with a polite interaction even though the issue remains open. Pair the measure with first call resolution, Customer Effort Score, repeat contacts, and human escalation outcomes. Sentiment analysis may help prioritize calls for review, but inferred emotion is not the same as a customer's direct satisfaction response.