Data sources, collection and labels

List where each feature comes from, how the label is created, and what you would collect next. Concepts: Internal vs external data; labelling; label quality and delay Project A: Label known only after the loan matures; 1,000 labelled rows; next to collect: income, bureau score, repayment history. Project B: Label (actual load) arrives hourly; extra sources: holiday calendar, weather forecasts (not actuals) for real forecasting.