How to Build a Customer Health Score from Stripe Data
A customer health score predicts which customers are likely to churn before they cancel. Here is how to build one from billing data, which signals actually predict churn, and how to use the score.
Contents
A customer health score is a single metric that tells you how likely each customer is to renew. Done well, it surfaces the customers who are quietly drifting toward cancellation weeks or months before they click the cancel button. Done poorly, it produces false positives that erode trust in the system and wastes customer success effort on accounts that were never at risk.
This guide covers how to build a health score from Stripe billing data, which signals actually predict churn versus which ones just look like they should, and how to use the score to take action.
The problem with most health scores
The most common health score mistake is using the wrong signals. Daily active users feel rigorous but do not predict churn in most SaaS products. What predicts churn is the trend of engagement relative to a customer's own baseline. A customer who went from daily to weekly logins in month three is a different signal from a customer who has always been weekly. A flat DAU metric cannot tell them apart.
The second most common mistake is flagging every account that dips below a threshold without accounting for what normal looks like for that customer. A health score that flags 30 percent of your accounts as at-risk every month produces so many false positives that no one acts on it.
A useful health score is calibrated, specific, and actionable. It flags the accounts that actually need attention, not every account that deviated from an average.
Billing signals that predict churn
Stripe billing data contains some of the strongest churn predictors available, often more reliable than product usage data because they reflect financial commitment rather than engagement behavior.
Payment failure acceleration. A customer whose payments fail and recover quickly is different from one whose payments fail more frequently over time. Increasing payment failure frequency over rolling 3-month windows predicts churn at meaningfully higher rates than isolated failures. The mechanism is behavioral: payment failure patterns predict voluntary churn before it shows up anywhere else in the data.
Annual to monthly conversion. Customers who switch from annual to monthly billing churn at 2 to 3 times the rate of customers who stay on annual. This is one of the highest-signal events in billing data and is almost never used for proactive outreach. The intervention window is 30 days before the customer fully disengages.
Post-recovery silence. A customer whose payment fails, gets recovered, and then shows no login activity after recovery is at higher churn risk than a customer who recovered and resumed normal behavior. Recovery in billing does not mean recovery in engagement.
Plan downgrade. A customer who downgrades has revealed a willingness to reduce spend. Without intervention, a percentage churns in the next 1 to 3 billing cycles.
Refund request in the past 90 days. Customers who requested a refund and stayed are ambiguous. Monitor this cohort. Some are fully resolved. Others churn 2 to 3 cycles later.
Building the score
A three-tier score (green, amber, red) built from billing signals is more actionable than a 100-point numerical score that no one can act on.
Green (healthy): Zero payment failures in the past 90 days, annual billing, no plan changes, no refund requests in 90 days, MRR stable or growing.
Amber (watch): One payment failure in 90 days (recovered), recent annual-to-monthly conversion, recent plan downgrade, or refund request in past 90 days.
Red (at risk): Multiple payment failures in 90 days, accelerating failure frequency, post-recovery silence, or recent downgrade plus low engagement.
Customer success effort should prioritize red accounts for immediate outreach, amber accounts for check-in within 30 days, and green accounts for expansion conversations.
What Recova Intelligence does
Recova's Intelligence dashboard scores every customer account automatically based on payment health, billing trajectory, and MRR movement pulled from your Stripe data. Red accounts surface in a needs-attention feed with the specific signal that triggered the flag and the recommended action.
- What is a customer health score?
- A single metric that summarizes how likely a customer is to renew, based on aggregated signals from payment behavior, billing trajectory, and engagement data.
- What Stripe billing signals predict churn?
- Payment failure acceleration over rolling 3-month windows, annual to monthly billing conversion, post-recovery silence after a payment failure, plan downgrades, and refund requests in the past 90 days.
- Why do most customer health scores produce false positives?
- They flag every account below a threshold without accounting for what normal looks like for that customer segment, lifecycle stage, or billing cycle.
- What should I do when a customer scores red?
- Immediate personal outreach for high-value accounts. For lower-value accounts, an automated check-in email acknowledging the issue and offering help.
- Can I build a health score from Stripe data alone?
- Yes. Billing signals like payment failure frequency, billing cycle changes, and plan downgrades are strong predictors. Adding product login data improves the score but is not required to build a useful starting point.
