Frame the use, build the dataset, explore

Most ML projects are won or lost before a model is trained. This session takes the first three steps of our churn project: deciding what a prediction is for and who acts on it, defining exactly what we are predicting and which information we'd really have at the time, and checking whether the data deserves trust. You'll write a decision statement, define the target, mark every column as available or not, lock the test set away, run five data quality checks and test your own hypotheses. No model today, on purpose. You leave with a project file that Session 2 builds on.