NPS, CSAT, CES — pick the metric for the decision
Three CX metrics, three different decisions. The disciplined operator picks the metric for the decision — not the metric for the headline.
What each measures
NPS measures relationship — likelihood to recommend on 0–10, calculated as %promoters minus %detractors. Composite, lossy, useful for relational tracking. CSAT measures transaction — satisfaction with a specific interaction on a 1–5 or 1–10 scale, typically %top-box. Contact-level signal; recency-biased. CES measures effort — how easy the resolution was. Strong inverse signal of repeat-contact and churn.
Each measures something specific; each misses something specific.
The metric-as-decision discipline
Pick the metric for the decision: trending the relationship → relational NPS; sustaining contact quality → transactional CSAT segmented by contact type; operational journey ease → CES segmented by journey; problems resolved → FCR + repeat-contact + complaint rate; what’s emerging → theme volume + verbatim review.
The undisciplined operator picks one metric and treats it as the answer to everything. A single NPS does not tell you about contact effort, repeat-contact, complaint emergence, or vulnerability.
Failure modes worth naming
Gamed by routing (only happy customers surveyed). Used at touchpoint level where it doesn’t fit. Composite indices treated as decision-grade without decomposition. Industry benchmarks treated as target. Verbatims unread, so the score is discussed but the why isn’t.
A metric is a prompt; the verbatim and the complaint are signal.
Composite, with caution
A composite VoC index can be useful as an executive dashboard; it is dangerous as the only metric. Composites should be decomposable — the leader who can’t drill into them to see what moved doesn’t actually know what’s happening.
One number on the wall is comforting and uninformative.
The closing principle
Pick the metric for the decision, segment where it matters, read the verbatims, and don’t let one number become the answer to everything.
See also
- Voice of Customer is not the survey
- Reading VoC data honestly behind the headline