Customer Health Score: Building One That Actually Predicts

A health score based on the wrong signals produces false confidence about churn risk.

Customer Health Scores: The Predictor Most CS Teams Get Backwards

Customer health scoring is where CS teams try to predict churn 90-180 days before it happens — giving them time to intervene. Most health scores are elaborate weighted averages of the wrong signals (login count, ticket volume, NPS) that produce false confidence: green accounts that churn, red accounts that renew easily. The scores that work select signals with proven predictive power and update the weights based on outcomes.

The dimensions of health

1. Product health: usage frequency, breadth (feature adoption), depth (advanced feature use), trend (rising/flat/falling). 2. Engagement health: exec sponsor engagement, QBR attendance, response rate to CSM outreach. 3. Business health: outcomes achieved vs promised, ROI conversation status, executive relationship. 4. Support health: ticket volume, severity, resolution satisfaction, escalation count. Score each 0-3, weight by predictive power, aggregate to a 0-100 score with tier thresholds.

The signal that predicts most

Across dozens of SaaS companies, the single strongest predictor of renewal is: 'has the executive sponsor engaged in a real business conversation in the last 90 days?' Product usage matters, but exec sponsorship matters more — power users can churn if leadership disengages, and low-usage accounts can renew if leadership sees strategic value. Health scores that weight product usage above executive engagement systematically miss the biggest risk factor.

Calibrating weights against outcomes

Every quarter: pull the last 4 quarters of churned and renewed accounts, look at their health scores 90 days before the renewal event. Which dimensions predicted the outcome, which didn't? Adjust weights accordingly. Health scores calibrated against real outcomes become predictive; health scores frozen at launch become organizational theater. Expect meaningful weight shifts every 2-4 quarters as your product and customer base evolve.

The three tiers

Green (80-100): renewal probability >90%, focus on expansion. Yellow (50-79): renewal at risk, structured intervention required (exec sponsor re-engagement, value review, roadmap alignment). Red (0-49): active churn risk, escalate to CS leadership + AE + executive sponsor from your side. Each tier has a defined playbook — health scores without playbooks are dashboards without decisions.

Making it actionable

The score exists to trigger action, not to describe status. Every yellow account triggers a workflow (value review scheduled within 14 days, exec sponsor outreach within 7). Every red account triggers escalation (weekly CS leadership review, exec-to-exec re-engagement). CSMs whose accounts move green→yellow without action taken are coached; the score is only useful if the workflow is enforced.

Common mistakes

Over-weighting product usage: usage predicts activity, not commitment. Ignoring executive engagement: the single strongest signal for enterprise. Complex scores no one trusts: 27-factor weighted averages CSMs override with gut instinct. Static weights: what predicted churn 2 years ago doesn't now. No workflow triggers: dashboards without playbooks. Uncalibrated against outcomes: guesswork dressed as data.

Frequently asked questions

Should health scores be visible to customers?
Never. Health scores are internal signals; sharing them creates gaming and defensive customer behavior. Share outcomes and ROI with customers; keep the score internal.
What tools produce good health scores?
Gainsight, ChurnZero, Vitally, Catalyst all support health scoring. The tool matters far less than the framework — a spreadsheet with good weights beats a platform with bad ones.
How often should CSMs review health scores?
Weekly. Monthly is too slow to intervene in yellow accounts before they turn red.

Related fundraising guides (40)

Investor directory · Fundraising library · Articles A–Z · Company funding database