A customer success team that tracks everything tracks nothing. The metric map below sorts the CS measurement space into four jobs — retain revenue, predict risk, prove the product gets used, and capture sentiment — and tells you which two or three numbers belong on the executive dashboard versus which stay inside the CS tool for the CSM to work. The mistake most teams make is putting a 12-metric scorecard in front of a board; the board has one question (is the revenue base safe?) and you answer it with two numbers.
The four jobs and the metrics that do them
Retain revenue — NRR and GRR. Net Revenue Retention and Gross Revenue Retention are the two outcome metrics. GRR measures the leak; NRR measures the leak net of expansion.
GRR = (Starting ARR − Churn − Downsell) / Starting ARR
NRR = (Starting ARR − Churn − Downsell + Expansion) / Starting ARRGRR is capped at 100% and tells you how leaky the bucket is. NRR is uncapped and tells you whether existing customers grow the account faster than they shrink it. Top-quartile B2B SaaS runs GRR 95%+ and NRR 120%+; below 100% NRR means expansion isn’t covering the leak. These are the two numbers the board sees.
Predict risk — churn rate and health score. Outcomes are lagging; churn rate and health score are how you see the leak before it hits NRR.
Logo churn = Customers lost in period / Customers at start of period
Gross $ churn = Churned ARR / Starting ARRLogo churn and dollar churn diverge when you lose small accounts (logo churn high, dollar churn low) or one whale (the reverse). Report both; the gap is itself a signal about where churn concentrates.
The health score is the leading indicator — a composite that predicts which accounts will churn or expand. Build it from weighted inputs, not vibes:
- Product adoption (40%) — license utilization, depth of feature use, frequency. The single most predictive input.
- Engagement (25%) — exec sponsor active, QBR attendance, support ticket sentiment, email responsiveness.
- Commercial (20%) — on-time renewal history, payment health, contract term remaining.
- Relationship (15%) — multi-threading depth, champion still employed, NPS/CSAT trend.
Calibrate the weights against your own churned-account history, then recompute weekly. A score that doesn’t predict churn in a backtest is decoration.
Prove the product gets used — adoption and TTV. A renewal is the bill arriving for value already delivered. Adoption metrics measure that delivery: license utilization (active seats / purchased seats), feature adoption (% of accounts using the features tied to their use case), and DAU/MAU stickiness for usage-based products. Time to Value (TTV) is the onboarding outcome metric — days from contract signature to the customer hitting their first defined success milestone. Shorter TTV correlates with higher retention; track median and the long tail separately, because the accounts stuck at 90+ days are your earliest churn cohort.
Capture sentiment — NPS, CSAT, CES. Three different questions, often confused:
- NPS (“how likely to recommend?”, 0-10) — relationship-level loyalty, surveyed quarterly or after a QBR. Score = % promoters − % detractors, range −100 to +100.
- CSAT (“how satisfied?”, 1-5) — transactional, fired after a specific interaction (support ticket close, onboarding milestone).
- CES (“how easy was it?”, 1-7) — effort on a specific task; the best predictor of repeat behavior and the most actionable of the three.
Sentiment is an input to the health score, not a dashboard headline on its own. NPS in isolation moves on survey-population noise.
Which metrics go where
| Audience | Metrics | Cadence |
|---|---|---|
| Board / exec | NRR, GRR | Quarterly |
| CS leadership | NRR, GRR, gross $ churn, health-score distribution, TTV median | Monthly |
| CSM working view | Per-account health score, adoption %, renewal date, open risks | Weekly / real-time |
| Onboarding team | TTV median + long tail, milestone completion rate | Weekly |
The executive dashboard is two numbers, not twelve. Everything else is a working metric that lives in the CS platform — Gainsight, ChurnZero, Vitally, or Planhat — where the CSM acts on it, not where the board reads it.
Common pitfalls
- NRR without GRR. 130% NRR with 80% GRR is a churn problem masked by expansion. Guard: always report the pair; the gap is the story.
- Health scores nobody backtests. A score that didn’t predict your last 20 churns predicts nothing. Guard: validate against churned-account history before trusting it, and re-validate quarterly as the product changes.
- Vanity adoption. “Logins” measures nothing; a user can log in daily and get zero value. Guard: define adoption as use of the features tied to the account’s stated use case, not raw activity.
- Survey-driven NPS theater. A 4-point NPS swing on a 30-response quarter is noise. Guard: report confidence intervals or response counts alongside the score, and treat sentiment as a health-score input rather than a standalone KPI.
- Cohort blindness. Blended NRR/GRR hides which segments leak. Guard: cut retention by acquisition quarter, segment, and ICP fit.
Related
- NRR vs GRR — the two retention metrics in depth
- Gainsight — health scoring and the CS command center
- ChurnZero — adoption and churn-risk automation