Role play: you have 15 minutes to review a slide deck on predicting airline departure delays and then 20 minutes to present to the Director of Operations. The current analysis uses linear regression with date encoded as an integer and includes plots that mix weekdays and months without seasonality controls. Tasks: (1) Critique the feature engineering (e.g., date as categorical or cyclical features, weather joins, route-level effects, leakage risks such as using wheels-off time). (2) Propose a modeling approach (regression vs. classification on delay > 15 minutes), metrics (RMSE/MAE or AUC-PR/cost), and cross-validation that respects time order. (3) Explain two charts you would replace, how, and why; produce a clear narrative for a non-technical audience linking insights to actions (e.g., crew buffers by route and time-of-day). (4) Deliver a concrete recommendation that would reduce compensation payouts by 5% without increasing cancellations, and define the experiment or KPI plan to verify impact within four weeks.