Predict Campaign Response for Retail Marketing

Business Context

BrightCart, an e-commerce retailer sending 8-10 million promotional emails and push notifications per month, wants to use predictive analytics to improve targeting. The marketing team currently sends broad campaigns with low conversion and wants a model that predicts which customers are likely to purchase within 7 days of receiving an offer.

Dataset

You are given a historical campaign dataset where each row represents a customer-campaign impression. The goal is to predict whether the customer will convert after receiving the message.

• Size: 420K customer-campaign rows, 42 features
• Target: Binary — purchased within 7 days of campaign send (1) vs no purchase (0)
• Class balance: 11.6% positive, 88.4% negative
• Missing data: 18% missing in push engagement for email-only users, 6% missing in demographic fields, sparse missingness in campaign metadata

Success Criteria

A good solution should improve campaign targeting enough to support selective sends. Aim for ROC-AUC >= 0.82, PR-AUC >= 0.38, and precision in the top 20% scored customers >= 0.30. The model should also produce interpretable feature importance for marketers.

Constraints

• Batch scoring must finish in under 20 minutes for 2 million customers
• Predictions are generated daily, so training can be offline
• Marketing stakeholders need understandable drivers of response
• Budget limits rule out large deep learning systems for the first version

Deliverables

1. Build a binary classification model to predict 7-day campaign conversion.
2. Explain preprocessing choices for mixed feature types and missing values.
3. Compare at least one baseline model with a stronger ensemble model.
4. Select a decision threshold or ranking policy for campaign targeting.
5. Report evaluation metrics and provide a short interpretation of business impact.