You don't know which users are about to churn until it's too late
Score users based on usage patterns. When their score drops, reach out before they churn.
When to use
You have enough users and data to identify churn patterns. Works best with 100+ active users.
Hypothesis template
If we score users on [usage signals] and intervene when scores drop below [threshold], we'll reduce churn by [target]% because we catch at-risk users before they leave.
Method
The problem: Users churn silently. By the time they cancel, it's too late to save them.
What works: A composite health score based on usage signals: login frequency, feature adoption, support tickets, billing issues. When the score drops, trigger an automated or manual outreach.
The result: Companies that implement health scoring reduce churn by 15-30% because they intervene while the user still cares enough to respond.
How to do it:
- Identify 4-6 signals that predict churn: login frequency drops, key features unused, support tickets increase, billing failures
- Weight each signal and create a composite score (0-100)
- Set thresholds: Green (80+), Yellow (50-79), Red (below 50)
- For Yellow: send automated check-in email with helpful tips
- For Red: personal outreach from founder or support lead
- Track: what percentage of Red users can you save?
Key insight: The best time to save a churning user is 2 weeks before they notice they've stopped using you. By cancellation day, it's over.
Success metrics
- •Churn rate reduction
- •Save rate (Red users retained)
- •Time from score drop to intervention
- •Revenue saved
Prerequisites
- Usage analytics per user
- Email or outreach system
- Enough historical data to identify churn signals
- Someone to respond to Red alerts
Common pitfalls
- •Using too many signals (4-6 is enough)
- •Setting thresholds too aggressively (too many alerts)
- •Only using automated responses for Red users
- •Not tracking whether interventions actually work
Source: Common SaaS pattern. Health scoring reduces churn by 15-30% through early intervention.
Suggested ICE scores
Ready to run this experiment?
Sign up free to use this playbook with step-by-step guidance and track your results.