CSAT vs NPS: which one to use for measuring customer experience
CSAT and NPS come up together in almost any conversation about customer experience, but they answer different questions. Using the wrong one gives you a number that looks like a serious metric but doesn't help you decide anything specific.
What each one measures
CSAT measures satisfaction with something specific: a purchase, a delivery, a support interaction. The question is "how satisfied are you with [this thing that just happened]?", on a short scale (typically 5 points). NPS measures loyalty toward the whole brand, built up across the entire relationship: "how likely are you to recommend us?", on the standard 0-to-10 scale.
When to use CSAT
Right after a specific interaction, while it's still fresh: closing a support ticket, after a purchase, at the end of an onboarding flow. Sent as a transactional survey (triggered by the event, not on a fixed calendar), CSAT gives you a precise read on how that particular interaction went.
When to use NPS
As a periodic tracking metric (quarterly or semi-annual), sent to the whole customer base without depending on something specific having just happened. NPS is useful for seeing loyalty trends over time, not for evaluating a specific interaction.
Why a high CSAT doesn't guarantee a high NPS
Someone can be very satisfied with a specific support interaction (high CSAT) and still not recommend the brand, because their loyalty decision depends on the whole relationship, not a single moment. The reverse also happens: one bad interaction doesn't necessarily destroy the loyalty of someone who generally has a good relationship with the brand.
Using both without them stepping on each other
CSAT after every support interaction, NPS once a quarter to the whole base — the two numbers coexist because they measure different questions at different moments. In fact, a sustained pattern of high CSAT over time often ends up showing up later as a higher NPS.
The question that settles it
"Do I need to know how this thing that just happened went, or how the whole relationship is doing?" — the answer tells you which of the two metrics applies. Neither one is "better" in the abstract.