Build a health score your data can back up.
Most health scores are a gut feeling wearing a number. Weight the signals that actually predict retention, then score a live account and watch the model work — and notice which signals you can truly measure versus the ones you’re guessing at.
What actually predicts a renewal?
Set how much each signal matters. Weights normalise to 100% automatically — the relative sizes are what count.
Now rate one real customer, 0–100 per signal.
Use an account you know. Where you have to guess instead of pull the number, that’s a data gap the score is quietly hiding.
—
—
Each bar is that signal’s contribution = its weight × the account’s value. The biggest bar is what’s holding the score up; the smallest is your first intervention.
The catch: —
A health score is only as honest as the data feeding it.
If “executive engagement” or “usage” isn’t a real, current field, that weight is fiction — and the score launders a guess into a number people trust. Whether your signals are actually captured and reliable is Data Foundation, the heaviest dimension in the Readiness Index. See how solid yours is.
How to build a health score that predicts churn instead of decorating dashboards
Most customer health scores are green. Most churning customers were green three months before they left. This paradox exists because most health scores weight signals by intuition rather than by their actual correlation with retention. This tool helps you build a model that weights the right signals and produces scores that CSMs can act on.
Enter the signals you want to include in your health score (product usage, support tickets, stakeholder engagement, contract history, NPS) and assign weights based on how strongly each signal predicts churn in your customer base. The tool shows you the resulting model, lets you score a live account, and highlights which signals are driving the score up or down.
Who this tool is for
CS leaders building or rebuilding their health scoring model. RevOps teams integrating health data into Salesforce or a CS platform. Anyone whose current health score says "green" for accounts that churn.
How to interpret your results
If your model produces mostly green scores, your weights are probably too generous or your signals are too activity-based (logins, ticket volume) rather than outcome-based (executive sponsor engagement, usage depth, expansion/contraction trends). Back-test against your last 20 churned accounts: if the model does not flag at least 70% of them as at-risk 90 days before churn, adjust the weights.
This tool maps to the Data Foundation dimension of the AI-Ready RevOps Framework. For a broader assessment, take the free assessment.
Related articles
How to build a customer health score model that predicts churn · How to simulate churn risk across your renewal book