Website XYZ, a music-listening social networking platform, operates on a freemium model — offering basic services for free, with additional premium features via a subscription. This project aims to predict which users are most likely to convert from free to premium subscribers within 6 months if targeted by a marketing campaign.
To identify high-potential customers likely to adopt premium features, enabling targeted promotional campaigns that increase ROI, reduce wasted outreach, and maximize customer value.
Given user behavioral and social features, build a classifier to predict the probability of adoption if the user is included in the next marketing campaign.
- Addressed class imbalance (few adopters)
- Nested Cross-Validation for robust model selection
- Hyperparameter tuning for XGBoost
- Threshold tuning for recall optimization
- SHAP explainability to uncover key drivers of adoption
- ROI simulation and LLM campaign suggestions
- Model: XGBoost Classifier
- Target Metric: Maximized Recall (important for reaching all potential adopters)
- Threshold: Tuned to optimize lift and reduce false negatives
- Top Model Features:
- Increase in songs listened over time
- Count of loved tracks
- Social network activity
- Engagement spikes
Use the deployed dashboard to:
- Upload customer data
- Get live adoption predictions
- Simulate ROI with custom cost/revenue assumptions
- Visualize SHAP explanations and lift curves
- Receive LLM-based campaign suggestions