I have a raw customer dataset that mixes demographics, usage logs and transaction history, and I need it turned into an end-to-end churn-prediction solution within four days. Day one I expect rigorous data cleaning in Python (pandas, NumPy preferred) and an initial exploratory notebook that highlights outliers, missing values and early churn patterns. Day two moves to feature engineering—tenure metrics, usage statistics, complaint counts and any other sensible transformations—followed by first-pass models in scikit-learn (logistic regression and a benchmark random forest). On day three you will refine those models, compare accuracy and ROC-AUC, pick the champion model, and build an interactive churn-risk dashboard in either Tableau or Power BI that surfaces overall churn probability, customer-level scores and the top drivers. Finally, the fourth day is reserved for polishing a concise report and leaving a small buffer for tweaks. Deliverables I need to see at hand-off: • A cleaned, well-documented dataset plus the complete Python code/notebooks • Saved, reproducible model artefacts with comments on hyper-parameters and performance metrics • A live Tableau or Power BI dashboard (cloud-shared link or packaged workbook) visualising churn risk and key drivers • A short PDF or slide deck summarising insights, model evaluation, and clear retention recommendations linked to the data I’ll review on the spot using my own test data, so please include instructions for re-running the pipeline and refreshing the dashboard.