CSAT (Customer satisfaction score)

CSAT comes from a short survey sent right after a support interaction, usually "How would you rate the support you received?" on a five-point scale. The standard buckets are 1 very dissatisfied, 2 dissatisfied, 3 neutral, 4 satisfied, 5 very satisfied, and the top two count as satisfied. Some teams use a three-point scale or a plain thumbs up and thumbs down instead; the arithmetic does not change, only which answers count as the satisfied band.

The formula is satisfied responses divided by total responses, times 100. Neutral answers stay in the denominator. Dropping them is the most common way a team accidentally inflates its own score.

Worked example. You send a one-question survey after every resolved conversation for a month and get 40 replies back: 22 rated it 5, 9 rated it 4, 4 rated it 3, 3 rated it 2, and 2 rated it 1. Satisfied responses are the 5s plus the 4s, so 22 plus 9 is 31. CSAT is 31 divided by 40, times 100, which is 77.5 percent, normally reported as 78 percent. Note what that denominator does at your size: with 40 responses, every single answer is worth 2.5 points. One extra unhappy customer next month drops you to 75 percent without anything about your support changing.

That sensitivity is the thing a team of two to ten people has to understand before acting on the number. Survey response rates on support email typically run well under a third, so a small team measuring a busy month may be reading 30 or 40 responses, and the people who bother to answer skew to the ends of the scale. A four-point move month to month is usually sampling noise. A twenty-point move is a signal. If you cannot tell which you are looking at, you do not have enough responses to run the metric monthly, and you should read it quarterly or not at all.

The part of CSAT that actually pays for itself at small scale is not the percentage, it is the free-text box under it. Ten written comments will tell you which article is missing from your help center, which policy people find unfair, and which reply of yours read as curt. That is directly actionable in a way that "CSAT was 78 percent" never is. If you only have the budget of attention for one of the two, keep the comments and drop the score.

CSAT is an interaction metric, which is what separates it from NPS. It measures how one exchange landed, so it responds to reply quality and tone rather than to how customers feel about your company overall. It also pairs naturally with first response time: a team that gets faster while getting worse will show a falling CSAT against an improving response time, and seeing the two together is the only way to catch that trade.

AcornReply does not send CSAT surveys. There is no survey builder, no rating scale, and no CSAT figure in the product. The one customer-stated signal it does collect is narrower: when the chat widget has handled a conversation with AI, it asks the customer whether that answered their question, and the analytics overview reports the counts of yes, no, and "chat with a human" for the period. A no or a request for a human reopens the conversation and pulls a person in. If you want real CSAT, run it with a dedicated survey tool alongside AcornReply.