NPS vs CSAT vs rating scales: pick the right one

They are not interchangeable

NPS, CSAT, and a generic rating scale get used as if they were three flavors of the same thing. They are not. Each answers a different question, on a different time horizon, about a different unit of experience, and picking the wrong one is the most common reason a feedback program collects data that never moves.

Before choosing a scale, decide what decision the answer will inform. 'Should we keep investing in this relationship' is a different question from 'did this specific interaction go well', and no single metric answers both.

NPS measures the relationship

Net Promoter Score asks how likely someone is to recommend you, 0 to 10. Nine and ten are promoters, seven and eight passives, zero to six detractors; the score is the percentage of promoters minus the percentage of detractors.

Its strength is that it is one comparable number about the overall relationship, which makes it good for tracking direction over quarters and across segments. Its weakness is the flip side: it is a lagging, low-resolution signal. NPS tells you the relationship is drifting; it almost never tells you which interaction caused the drift.

Use NPS as a trend line, not a diagnostic. And place it early — opinions drift as a survey goes on, and you want this one before the rest of the form colors it.

CSAT measures the interaction

CSAT asks how satisfied someone was with a specific thing that just happened — this support ticket, this onboarding call, this checkout. It is usually a short scale reported as the percentage who answered in the top one or two boxes.

Its strength is precision in time and scope. Fired right after an interaction, it points at a fixable moment while the experience is fresh. Its weakness is that it says little about the long-term relationship; someone can rate every interaction fine and still churn.

The clean split: CSAT for 'did this go well', NPS for 'is this relationship healthy'. Many programs need both, asked at different moments, not one stretched to cover both jobs.

Plain rating scales measure attitude strength

A symmetric agree-to-disagree scale — the Likert scale — measures how strongly someone holds a specific attitude. It is the right tool when you want nuance on a particular statement rather than a single headline number.

Two rules carry most of the value. Keep the direction and endpoints consistent across the whole survey, because a respondent who learns 'right is good' and then hits a flipped item produces noise, not data. And resist long matrices: stacking many rows is efficient to build and exhausting to answer, and fatigue shows up as straight-lining down a column.

Five points is plenty for most attitudes. Reserve the 0 to 10 range for NPS, where the cutoffs are defined, so your scales stay legible across the form.

Choose by the decision, then stay consistent

The selection rule is short. Tracking the relationship over time: NPS. Evaluating a specific interaction: CSAT. Understanding the strength of a particular attitude: a rating scale. If a question does not map to a decision, it should probably not be on the form.

Whatever you choose, do not change it casually. Because answers reference stable question IDs rather than labels, you can re-word a prompt for clarity without breaking the trend — but switching from a 5-point scale to NPS mid-program resets the comparison. Consistency is most of what makes a metric worth tracking at all.