Customer Health Score
A customer health score is a single rating, usually 0–100 or a color tier, that summarizes how likely a customer is to renew, churn, or expand, based on signals like product usage, support tickets, communication sentiment, and billing.
What feeds a health score
A useful score triangulates several signal families, because each one alone lies. Usage tells you whether people show up: logins, feature adoption, active seats. Support tells you whether the product is hurting: ticket volume, severity, open escalations. Communications tell you how the relationship feels: email sentiment, slowing reply times, a champion who stopped answering. Billing tells you what the money is doing: late invoices, downgrades, a renewal date approaching with no conversation booked.
An account can look healthy on any one axis while dying on another, steady usage with a furious inbox is still an account at risk.
Rules-based vs AI scoring
Rules-based scoring is the classic approach: an admin assigns weights, usage 30%, tickets 20%, NPS 15%, and the platform adds them up. It’s transparent and easy to explain, but rigid. The weights encode one person’s guess at what predicts churn, they age silently as the product and customer base change, and anything unstructured, like the tone of an email thread, never makes it in.
AI scoring reads the signals directly, including the unstructured ones, and can be calibrated against ground truth: which accounts actually churned and which renewed. The tradeoff is trust, a model’s score is only actionable if it shows its reasoning. A number with no evidence behind it gets ignored just as fast as a stale rules formula.
Why most health scores get ignored
The most common failure isn’t a wrong score, it’s a vanity score: a dashboard number nobody acts on. It happens when the score can’t answer “why?”, when it updates too slowly to catch a churn story in motion, when it never gets checked against real churn outcomes, or when everything scores yellow and the team stops looking. A health score earns its place only if it changes what someone does this week, which account gets the call, which renewal gets the save.
No score without evidence.
Keply scores every account from real signals, email, tickets, usage, billing, and each score names its drivers on screen: “renewal in 21 days, usage −40% MoM, 3 open tickets.” Scores calibrate on your own churn history, and when one drops, Keply drafts the save and sends it with your approval by default. Flat $299/month. See how it works.
Health scores, FAQ
- What signals go into a customer health score?
- The common inputs are product usage (logins, feature adoption, seat activity), support activity (ticket volume, severity, escalations), communication signals (email sentiment, response times, a champion going quiet), and billing (late invoices, downgrades, upcoming renewals). The best scores weigh several of these together rather than leaning on one.
- What's the difference between rules-based and AI health scoring?
- Rules-based scoring applies weights an admin configures: logins count for 30%, tickets for 20%, and so on. It's transparent but rigid, and it only sees what someone thought to encode. AI scoring reads the underlying signals, including unstructured ones like email tone, and can be calibrated against which accounts actually churned. Either way, a score you can't trace to evidence is hard to act on.
- What is a good customer health score?
- Scores are relative to your own scale, a 72 means nothing outside the system that produced it. What matters is that the bands separate real outcomes: accounts scored healthy should renew at a visibly higher rate than accounts scored at risk. If they don't, the score needs recalibrating, not the customers.
- How often should health scores update?
- As often as the signals change, daily in practice. A score computed monthly can miss a churn story that unfolds in weeks: usage drops, then tickets spike, then the champion goes quiet. The point of a health score is catching that sequence while the renewal is still saveable.
See every account scored, with the evidence on screen.
Connect your stack or drop in a spreadsheet. Live in a day.
Start freeFlat $299/mo. First month free. 30-day money-back. Cancel anytime.