What CSAT measures
Customer satisfaction score asks how satisfied someone was with a particular thing that just happened — a support conversation, a delivery, an appointment, a purchase. The customer answers on a short rating scale, commonly running from one to five, and the score is reported as the share of answers that fall in the satisfied end of that scale.
Two decisions define the measurement, and both are yours. The first is the scale and which points on it count as satisfied. The second is the trigger: what event sends the survey, and how soon afterwards. Change either one and the number moves without anything about the business changing, so both belong in a written definition that outlives whoever set it up.
Why CSAT matters
It is the most specific of the common sentiment measures. Because it is tied to a single interaction, a poor result usually points at something you can name — a slow reply, a courier, a booking form, a particular member of staff — rather than at a vague feeling about the brand. That makes it the fastest of these scores to turn into a fix.
It is also the one that survives a small business. A shop or a clinic without enough customers for meaningful survey rounds can still read satisfaction one interaction at a time, spot a run of poor scores in a week, and deal with it before it reaches a public review.
Where CSAT goes wrong
Response bias is the main problem. People who are delighted and people who are furious answer surveys; the quietly disappointed middle usually does not. So the score tends to be shaped like a barbell, and averaging it hides that shape completely. Look at the distribution, not just the headline.
The second trap is mixing interaction types. Satisfaction with a refund and satisfaction with a first delivery are not the same measurement, and blending them produces a number that cannot rise or fall for any traceable reason. Report each trigger separately.
The third is asking too often. Survey fatigue lowers response rates, which shrinks your sample, which makes the score jumpier — and a jumpy score invites explanations for movement that was never real.
How to act on it
Always pair the rating with a comment box, and route low scores to a person rather than to a report. A dissatisfied customer contacted quickly is a recoverable customer; the same person left in a dashboard is a bad review waiting to be written.
Then use the pattern rather than the individual score. Group the comments behind poor ratings, find the two or three causes that repeat, and fix the process behind them — often that means changing a handover, a template or a response time inside your customer support workflow. Watch the effect over the following weeks alongside retention rate, because a satisfaction score that improves while customers keep leaving is measuring the wrong moment.
