Use cases at a glance
Churn doesn't arrive without warning signs. Most at-risk accounts show multiple signals weeks before cancellation: failed logins, support tickets, usage drops, or payment failures. The at-risk agents in this guide catch those signals, score them, and surface flagged accounts before your support team notices empty inboxes.
⚠️ At-Risk Account Scoring
Real-time scoring system that weights multiple signals (login, usage, support, payment) and flags accounts above the risk threshold daily.
⏰ Trial Expiry Alerts
Monitor trials expiring in 3, 7, and 14 days without converting. Send timely reminders before they auto-expire.
📉 Downgrade Velocity Tracking
Detect when customers downgrade plans, especially in the same month as a prior upgrade (likely dissatisfaction).
📊 Cancellation Spike Detection
Alert when cancellations in a single day exceed your 7-day average. High-velocity churn deserves immediate attention.
🔄 Reactivation Tracking
Monitor cancelled subscriptions reactivated within 30 days, and follow up with high-value reactivations.
How churn signal scoring works
The scoring system assigns a weight to each churn signal based on its predictive power. Each account's total score is the sum of its signals' weights. Accounts exceeding a threshold (typically 18-22, configurable) are flagged as at-risk.
| Signal | Weight | Definition |
|---|---|---|
| No login in 14+ days | 8 | Account not accessed (may indicate switching or low priority) |
| Support tickets this week | 6 | Spike in help requests often precedes churn |
| Usage dropped >50% vs last month | 9 | Significant drop signals disengagement |
| Failed payment in last 7 days | 7 | Payment issues are strong churn predictors |
| Downgraded plan in last 30 days | 5 | Plan reduction shows reduced commitment |
| Trial expiring in 3 days without upgrade | 10 | Highest weight; trial expiry is near-certain churn if not converted |
Setting up the at-risk agent
The churn-signal-scorer agent runs daily and computes each account's risk score. Configure it with your weighted signals and the threshold above which accounts are flagged.
Trial expiry monitoring
Trials are your highest-risk segment. Send reminders at 14 days, 7 days, and 3 days before expiry. Most teams see 20-40% conversion from a 3-day reminder email.
Downgrade velocity tracking
A single downgrade is not alarming. But a downgrade within 30 days of an upgrade signals buyer's remorse — and probable churn within 60 days. Track and alert on these patterns.
Cancellation spike detection
Normally your churn is steady: maybe 1-3 accounts per day. If you see 5+ cancellations in a single day, something's wrong (product incident, competitor launch, billing issue). Alert immediately.
HEARTBEAT templates
Sample at-risk report
FAQ
How does the churn signal scoring system work?
The system assigns a weight (0-10) to each churn signal based on its predictive power. Signals include no login in 14+ days (weight 8), support tickets opened this week (weight 6), usage dropped >50% vs last month (weight 9), failed payment in last 7 days (weight 7), downgraded plan in last 30 days (weight 5), and trial expiring in 3 days without upgrade (weight 10). Each account's total score is the sum of weighted signals. Accounts exceeding the threshold (typically 18-22) are flagged as at-risk and included in daily reports.
Will I get false positives from healthy accounts?
Yes, especially for the 'no login' signal which can indicate integration completion for low-touch products. Adjust weights based on your product: lower the login weight for embedded tools, raise it for usage-based products. Also use the 'ignore_if' exceptions in the agent config to exclude accounts by plan tier (e.g., ignore trials under 7 days old, ignore free plans with short tenure). Track outcomes over 4 weeks and refine weights based on which flagged accounts actually churned.
What should I do when an account is flagged as at-risk?
Recommended workflow: (1) Review the signal breakdown to understand which factors triggered the flag. (2) If it's a high-MRR account or strategic customer, send a proactive check-in via email or CS outreach that same day. (3) For trial expiry alerts, a simple reminder email with one-click upgrade has 20-40% conversion rates. (4) Track outcomes (customer replied, upgraded, churned) to refine your weights over time. In 4-6 weeks, you'll have enough data to optimize which signals predict real churn in your specific market.