Business Context

Frost Bank wants to identify retail customers who are likely to enroll in Frost's mobile banking app within the next 30 days. The analytics team will use the model to prioritize digital adoption campaigns and branch outreach across roughly 1.2M active consumer customers.

Dataset

The training data is a customer-level snapshot built from Frost Bank transaction systems, CRM, branch interactions, and digital channel events.

• Rows: 420K customers observed monthly over the last 18 months
• Target: Whether the customer enrolls in Frost mobile banking within 30 days of the feature snapshot
• Class balance: 14.6% positive, 85.4% negative
• Missing data: Moderate missingness in branch interaction fields and some digital activity fields for offline-only customers

Success Criteria

A strong solution should achieve ROC-AUC >= 0.82, PR-AUC >= 0.42, and top-decile lift >= 2.5 on a held-out time-based test set. The model should also produce explainable drivers so marketing and branch teams can understand why a customer is scored as high propensity.

Constraints

• Batch scoring only; nightly scoring for ~1.2M customers must complete within 30 minutes
• Model outputs must be explainable enough for campaign targeting review
• Avoid leakage from post-enrollment events
• Retraining should be feasible on a monthly cadence

Deliverables

1. Build a binary classification model to predict 30-day Frost mobile app enrollment.
2. Describe your end-to-end modeling approach, including preprocessing, feature engineering, validation, and thresholding.
3. Compare at least one interpretable baseline with one stronger nonlinear model.
4. Report evaluation metrics on validation and test data.
5. Recommend how the model would be deployed for monthly retraining and nightly batch inference.