What is customer satisfaction score (CSAT)?

Customer satisfaction score (CSAT) is the share of customers who rate an interaction positively in a one-question survey asked right after it: "How was this conversation?" with stars, a 1 to 5 scale, or thumbs up and down. The usual formula is positive ratings divided by all ratings, with positive meaning 4 or 5 on a five-point scale or a thumbs-up. Some tools report the average score instead, which is a different number from the same votes.

It matters because CSAT is the only standard support metric that asks the customer, and because it is cheap: one tap, no form. For an AI agent it is the counterweight to containment and resolution, which can be flattered by never handing off; a visitor who was contained and unhappy will say so here. Read it by topic, and read the transcripts behind the negative votes, since the number alone does not say why.

Where it runs out: most customers never rate. Response rates of a few percent are common, and the people who do vote skew toward the delighted and the furious, so a CSAT of 85 percent from 40 votes describes 40 people. Ratings also measure the whole interaction, so a slow human reply after a good AI answer lowers the AI's score. Compare CSAT across products only when the question, the scale and the moment it is asked are the same, and never compare a share of positive ratings with an average score.