Business Context

Ancestry Marketing sends millions of lifecycle and promotional emails each month across Ancestry subscriptions and DNA-related offers. Your task is to build a model that predicts whether a recipient will convert within 7 days of receiving an email so the team can prioritize high-value audiences and improve campaign efficiency.

Dataset

You are given one row per delivered email from the last 12 months, with historical user behavior and campaign metadata available at send time.

• Size: 420K delivered emails, 45 features
• Target: Binary — converted to paid subscription or upgrade within 7 days of send
• Class balance: 11.4% positive, 88.6% negative
• Missing data: ~18% missing in DNA-related features for non-DNA users; ~6% missing in device/email-client fields

Success Criteria

A good solution should achieve strong ranking quality for campaign targeting: ROC-AUC >= 0.80, PR-AUC >= 0.35, and precision in the top decile >= 0.30. The model should also produce interpretable feature importance for marketers.

Constraints

• Batch scoring must complete daily for upcoming Ancestry Marketing sends
• Predictions should be explainable enough for channel managers to trust segment selection
• Training should avoid leakage from post-send behavior
• Retraining should be feasible on a monthly cadence

Deliverables

1. Build a binary classification model to predict 7-day conversion from email sends.
2. Explain feature engineering, preprocessing, and leakage prevention.
3. Compare at least one baseline model with a stronger non-linear model.
4. Evaluate using business-relevant ranking and classification metrics.
5. Recommend a production decision threshold or top-N targeting strategy.