Which customers have the highest probability of churn?
Churn is predicted for a fixed decision date and horizon. Temporal separation, leakage prevention and calibration matter as much as discrimination.
Direct answer
Prioritize human review using calibrated probabilistic risk.
Churn is predicted for a fixed decision date and horizon. Temporal separation, leakage prevention and calibration matter as much as discrimination.
logit(P(churn)) = α + XₜβESTIMAND
Estimation profile
- Unit of analysis
- Customer × decision date × horizonControlled scientific terminology
customer at decision date - Exact estimand
P(churn within H | information available at t)- Model or deliverable
logit(P(churn)) = α + Xₜβ
01—08
Verifiable analysis framework
- 01Question
Which customers have the highest probability of churn?
- 02Estimand
P(churn within H | information available at t) · customer at decision date
- 03Data
customer_id · decision_date · features_available_at_t · churn_by_horizon
- 04Model or deliverable
logit(P(churn)) = α + Xₜβ
- 05Declared calculation
logit(P(churn)) = α + Xₜβ → P(churn within H | information available at t)
- 06Uncertainty checks
Temporal holdout · Brier and log-loss · PR-AUC and calibration
- 07Method-specific validation
Check data, estimate stability and interpretation limits. Temporal holdout · Brier and log-loss · PR-AUC and calibration
- 08Limitations
Claim which action will reduce churn from the predictive model alone.
Required variables
customer_iddecision_datefeatures_available_at_tchurn_by_horizon
Checks · Controlled scientific terminology
- Temporal holdout
- Brier and log-loss
- PR-AUC and calibration
Prioritize human review using calibrated probabilistic risk.
Claim which action will reduce churn from the predictive model alone.
Scientific sources
1 source
Method connections