Customer Satisfaction Score (CSAT)

Customer Satisfaction Score measures how satisfied a customer is with a specific interaction, product, or service, usually by asking them to rate their satisfaction on a short scale immediately afterwards. The score is typically reported as the percentage of respondents selecting the top ratings, such as four and five on a five-point scale, though variations in scale and calculation are common enough that comparisons between organizations are rarely meaningful.

Its defining characteristic is specificity of timing. Where a loyalty metric asks about an overall relationship, CSAT asks about a particular event while it is still fresh: a support conversation, a delivery, an onboarding session, a checkout. This makes it diagnostic in a way that relationship-level metrics are not, because a low score can be attached to an identifiable interaction, a specific agent, a particular product, or a defined process step, and therefore acted upon.

The metric's principal strengths are immediacy and response rate. Asked at the moment of completion, in the channel where the interaction occurred, CSAT surveys achieve much higher response rates than periodic relationship surveys, and the answers reflect actual experience rather than reconstructed memory. The single question is also cheap enough to deploy across every touchpoint, which produces the volume needed to segment meaningfully by channel, team, product, and time period.

Its limits follow from the same specificity. Satisfaction with an interaction does not predict loyalty or future behavior reliably: customers frequently report satisfaction with a support conversation that resolved a problem the product should never have caused, and then leave anyway. High scores on individual interactions can coexist with deteriorating retention, particularly where the underlying product experience is poor but the service recovering from it is good. CSAT should therefore never be read as a proxy for the health of the customer relationship.

Measurement artifacts are also common. Scores skew positive because dissatisfied customers frequently disengage rather than respond, and because rating a person, such as a support agent, produces higher scores than rating a process. Asking too often causes fatigue and declining response quality. Attaching scores to individual staff performance leads directly to score solicitation, where agents ask customers for top ratings, which inflates results without improving anything. And because the wording, scale, and calculation vary widely, industry benchmarks are much less comparable than they are usually presented as being.

Timing decisions carry more weight than they are usually given. Surveying immediately after a support conversation captures the interaction while it is fresh but before the customer knows whether the promised resolution actually occurred, which measures the conversation rather than the outcome. Surveying after the resolution is confirmed measures something more meaningful but suffers lower response rates and weaker recall of the interaction itself. Businesses that care about the distinction typically ask twice with different questions, or choose the timing according to which they intend to act on, and the choice should be documented, since a change in survey timing produces a shift in scores that will otherwise be interpreted as a change in performance.

Used properly, CSAT is an operational metric rather than a strategic one, and its value is concentrated in the free-text follow-up and in the segmentation. Tracking the score by touchpoint identifies which steps in a journey are failing, and pairing it with behavioral outcomes shows whether those failures matter commercially. In practice this means routing the verbatim comments into the qualitative pipeline of a user research programme and connecting the scores to retention and repeat purchase data through data analytics, so that satisfaction is evaluated against behavior rather than reported as an end in itself.