Customer satisfaction score (CSAT)
CSAT measures how satisfied customers were with a specific interaction, from a short rating survey sent after the conversation or ticket ends.
Customer satisfaction score (CSAT) is a measure of how satisfied customers were with a particular interaction, usually collected with a one-question rating survey right after a conversation or ticket ends.
How it is collected. The survey asks a single question about the interaction just finished, sometimes with a follow-up. Zendesk, for example, lets administrators ask the CSAT question on a 1-2, 1-3 or 1-5 scale, shown as numbers, emoji or custom labels, and add a follow-up question when the rating is negative. For email tickets it sends the survey one day after the ticket is solved. For web, mobile and messaging conversations it presents the survey immediately after the ticket is solved.
How it is calculated. There is no single formula, so a CSAT figure only means something alongside its method. Zendesk maps each response to good or bad (on a 1-5 scale, 4 and 5 count as good) and reports the score as a percentage. Salesforce’s CX Insights dashboard computes CSAT as the average of the CSAT scores it receives. The same customers could produce two different numbers under these two methods, so compare CSAT only within one method.
What it does not show. CSAT reflects the customers who answer, about the interaction they just had. It says nothing about people who gave up before reaching anyone, and little about whether the underlying problem stayed fixed. Teams usually read it next to operational measures such as first contact resolution and average handle time.
When agents handle the request. If a customer’s AI agent settles a request with a business’s agent, no person may be present to rate the exchange as it happens. A survey can still go to the customer about the outcome, and outcome-based measures, such as whether the request was resolved without a repeat contact, carry more of the weight.
Neighbouring terms. First contact resolution counts issues solved in one contact. Containment rate counts conversations an automated agent finished without a person.