You are building a predictive model for a financial data and analytics platform to identify client accounts that are at risk of attrition. The commercial team wants an account-level risk score they can use to prioritize outreach before renewal decisions are made.
How would you approach building this model end to end, from framing the prediction problem through model selection, feature engineering, evaluation, and deployment? What tradeoffs would guide your choices as you move from an initial baseline to a production-ready solution?