You must predict 90-day customer value CLV_90 (many zeros, heavy right tail) at the user level from features collected up to a cutoff date: continuous spends by marketing channel (Spend_SEM, Spend_Social, Spend_Display), recency/frequency/monetary (RFM), device, region, tenure, and dozens of sparse categorical campaign IDs (high cardinality), with known issues: strong multicollinearity among spend channels, heteroskedastic errors, and nonlinear effects. Design a regression pipeline that: (1) chooses an appropriate loss/distribution (e.g., log-link GLM, Tweedie, zero-inflated, quantile regression, or gradient-boosted regression) and justifies it; (2) performs feature processing (log1p transforms, standardization, rare-category bucketing, target encoding for high-cardinality features with K-fold out-of-fold encoding to avoid leakage); (3) handles multicollinearity and feature selection (ridge/lasso/elastic net): write the elastic net objective with α and λ, and explain when each penalty dominates; (4) sets up time-based nested cross-validation that prevents leakage from future signals and campaign overlap; (5) diagnoses model assumptions: detect and address heteroskedasticity (e.g., via White-robust SE, variance-stabilizing transforms, or modeling the mean-variance relationship in a Tweedie GLM), nonlinearity (splines/interactions), and influential outliers (Huber/Tukey loss); (6) provides calibrated 95% prediction intervals for CLV_90 (e.g., conformal prediction or bootstrap) and compares them to OLS analytic intervals; (7) interprets effects: compute and interpret standardized coefficients for a regularized linear model versus SHAP values for a tree model; (8) quantifies and mitigates data leakage risks (e.g., target leakage via post-cutoff features, look-ahead in target encoding); and (9) contrasts performance and interpretability trade-offs between a regularized linear model and a gradient-boosting model, specifying metrics that are robust to heavy tails (e.g., MAE, Quantile loss, MAPE with epsilon).