Churn Early Warning
Spot at-risk renewals 120 days out, not 12
Customer health scores lie. This skill combines product usage, support patterns, engagement trends, and stakeholder changes into a predictive model that catches churn risk months before renewal — when you can still do something about it.
INGREDIENTS
PROMPT
Create a skill called "Churn Early Warning". Build a churn prediction model for my customer base. Input signals: product usage trends (logins, feature adoption, API calls), support ticket volume and sentiment, NPS/CSAT scores, stakeholder engagement (meeting frequency, email responsiveness), champion job changes, and contract details (renewal date, term length, last price increase). Score each account 1-10 on churn risk. Alert at: score 7 = CSM warning, score 8 = manager escalation, score 9+ = executive save play. For each at-risk account, show: risk score, contributing signals, trend direction, days until renewal, and recommended save play. Weekly report: total at-risk ARR, accounts by risk tier, save plays in progress, and save rate.
How It Works
The skill builds a multi-signal churn prediction model. Usage declining? Support
tickets spiking? Champion left? NPS dropped? Each signal adds to a risk score that
triggers alerts at configurable thresholds — early enough to run a save play.
What You Get
- Predictive churn scoring based on multiple signal types
- Early warning at 120+ days before renewal
- Signal breakdown: which factors are driving risk
- Recommended save plays based on risk type
- Trend tracking: is this account improving or declining?
- Renewal pipeline report: total at-risk ARR and save play status
Setup Steps
- Connect available data sources, or provide recurring exports from those systems: product usage, support tickets, NPS, engagement
- Define signal weights (or let the model learn from historical churn data)
- Set alert thresholds: when to warn CSM, when to escalate
- Provide historical churn data for model calibration
Tips
- Champion departure is the highest-urgency signal — react within 48 hours
- Usage decline over 3+ weeks is more predictive than a single bad week
- Don't rely on NPS alone — detractors sometimes renew, promoters sometimes churn
- Run weekly save-play standup on all accounts above the risk threshold