Customer Health Scores That Actually Predict Renewal
Most health scores are a weighted average of whatever data was easy to collect. Here is an outcome-first way to design a score your CSMs trust and your renewals actually follow.
Ask a customer success team how their health score was built and you will often hear some version of: “We took logins, support tickets, NPS, and a CSM gut-feel field, and weighted them.” The result is a traffic light that looks scientific but rarely predicts what leadership cares about – who will renew, who will expand, and who is quietly leaving.
A health score is only useful if it changes what your team does on Monday morning. That starts with designing it backwards from outcomes rather than forwards from available data.
Start with the outcome, not the inputs
Before choosing a single signal, agree on what the score is supposed to predict. Renewal in the next two quarters? Expansion? Logo churn versus downgrade? Each outcome has different leading indicators. A score that tries to predict everything predicts nothing well.
Then pull historical data for customers who renewed, expanded, downgraded, and churned, and look for the signals that actually separated those groups – with enough lead time to act.
Signals that usually matter more than logins
- Breadth of adoption: how many teams, seats, or use cases are active, not just how often one champion logs in.
- Depth on value-driving features: usage of the capabilities tied to the outcome the customer bought, not generic activity.
- Relationship coverage: whether you have an engaged executive sponsor and more than one active contact.
- Trajectory, not snapshots: a 30% drop in active users over 60 days says more than any single month’s total.
- Commercial signals: late payments, procurement changes, budget freezes, or a new decision-maker.
- Support sentiment: the tone and severity of tickets, not just their count.
A health score your CSMs do not trust is just a dashboard. A score that explains itself becomes a workflow.
Make the score explain itself
Trust comes from transparency. Every score should show its top drivers in plain language – “executive sponsor inactive for 45 days,” “adoption down 28% in the finance team” – so a CSM can see why an account is red and what to do about it. A black-box number invites people to override it with intuition, and then the score stops improving.
Turn scores into plays
Each meaningful score change should trigger a defined response:
- Risk plays for declining accounts: executive outreach, a value review, or a targeted enablement session.
- Expansion plays for accounts with broad, deep adoption: introduce adjacent use cases or seat expansion while momentum is high.
- Automation for the long tail: digital nudges and in-product guidance for accounts too small for high-touch coverage.
Close the loop every quarter
Compare predicted risk against what actually happened at renewal. Which red accounts renewed? Which green accounts churned? Use those misses to reweight signals and retire the ones that add noise. A health score is a model, and like any model it drifts if no one maintains it.
Where AI helps – and where it does not
Machine learning can find signal combinations humans miss and score thousands of accounts continuously. AI copilots can summarize an account’s risk drivers before a call. But AI cannot fix an unclear outcome definition or missing data. Get the foundation right first; then let AI scale it.
In one StratApps engagement with a vertical SaaS platform, AI-powered health scoring and an expansion play system embedded in Salesforce helped lift net revenue retention from 104% to 121% over three quarters. If you want a health score that predicts renewal, see how we approach health scoring.
