Whenever you're ready, walk me through how you'd approach this.
We want to build a feature that predicts which customers are most likely to churn this quarter. Walk me through how you would design this system from data collection to model deployment.
Address how you would define churn, identify useful data sources, select and evaluate a modeling approach, determine how predictions would drive customer-success actions, and deploy and monitor the system. Discuss trade-offs involving model accuracy, interpretability, data leakage, fairness, operational cost, and prediction timing.