Seasonal false alarms in customer churn and decline early-warning systems: adjacent-window labels confound seasonality with decline, and a year-over-year correction

How many days of early behavior suffice for subscription churn prediction? In the public KKBox dataset, the early indicator of churn is typically an indicator of someone's contract status; however, when looking in the heavily churned manual-renewal segment, having access to early behavior creates a substantial increase in prediction for that specific segment (PR +0.10 at 120 days). A nine-window sufficiency curve shows diminishing returns; descriptive inflection detectors place a knee between 30 and 120 days depending on the metric and model class. Stress-testing over three cohort/task designs shows that this curve is singular to the design being tested; for example, in our test with a moving target, the curve inverts and can shift depending on the feature set used. Therefore, any window-sufficiency claim should state its cohort construction, target definition, and feature families. A survival-filtered replication on educational-attrition data (OULAD) confirms the rising-curve shape in one behavior-dominated setting (floor ROC 0.56 vs. 0.88), though signal composition inverts.

Customer churn prediction is a central task in customer analytics, particularly in non-contractual, pay-per-use service environments where disengagement is not explicitly observed and must be inferred from behavioral inactivity. Existing churn prediction approaches often rely on simplified temporal assumptions or single-point representations of customer behavior, which limit their ability to support continuous risk assessment, interpretability, and realistic deployment over time. This study proposes a temporally explicit churn prediction framework that models customer behavior using rolling behavioral windows, enabling repeated and instance-level churn risk estimation as customer activity evolves. Customer behavior is summarized within a fixed 30-day observation window, followed by a 30-day future churn evaluation window, ensuring a clear temporal separation between behavioral evidence and churn outcomes. The framework integrates feature-based and sequence-based learning approaches within a unified temporal design. The proposed approach is evaluated on a large-scale, real-world dataset from a non-contractual service platform. Empirical results demonstrate strong and stable predictive performance, with accuracy reaching 87.6% and ROC-AUC of 0.94 for the feature-based model, while the sequence-based model achieves recall as high as 96.1% by capturing temporal disengagement patterns. Evaluation on future unseen data confirms meaningful robustness under temporal shift, with accuracy remaining above 83% and ROC-AUC exceeding 0.91 without model retraining. Overall, the findings highlight that carefully designed temporal framing, rather than model complexity alone, is critical for achieving robust, interpretable, and deployment-ready churn prediction. The study provides a practical foundation for churn-oriented decision support in dynamic service environments.

Customer churn prediction on the IBM Telco Customer Churn benchmark (n = 7,043) routinely reports test accuracies above 95%, with the most cited published study reporting 99.01%. We audit this benchmark for four trustworthiness failures invisible to the accuracy- and F1-centred reporting that dominates the literature. First, pre-split SMOTE inflates churn-class F1 by 13.1 percentage points across ten classifiers and fifteen seeds (Wilcoxon p < 10^-4 per classifier); the same leaky pipeline ordering paired with class weighting yields no inflation, isolating the effect to SMOTE's geometric construction. We measure the mechanism directly: approximately 36% of synthetic training points are nearest-neighbour interpolations of test-set instances. Second, the TotalCharges field is approximately determined by tenure multiplied by MonthlyCharges (R2 = 0.999); removing it changes accuracy by less than 0.2 percentage points, yet TreeSHAP ranks it ninth in mean absolute attribution - a pattern that materially corrupts SHAP-based interpretation. We propose an R2 > 0.95 pre-modelling diagnostic. Third, in a 15-seed calibration audit, isotonic regression is the strongest default; temperature scaling fails on class-weighted tree ensembles whose predicted-probability distribution is bimodal. Fourth, the cost-optimal decision threshold (under a 50 USD retention offer and 24-month CLV proxy) is approximately 5-10 times lower than the F1-optimal threshold, saving approximately 77,000 USD per 1,000 customers. We replicate F1 and F2 on Iranian Telecom Churn (within domain) and Bank Customer Churn (across domain): F1 generalises; F2 generalises only within telecom. We synthesise these findings into a four-component reporting checklist - pipeline disclosure, redundancy diagnostic, calibration audit, and cost-sensitive thresholds - and release a reproducible implementation.