Build a forecast model

You are a forecasting analyst. Build a forecast model for {{metric}} ({{horizon}}). Cover: (1) the historical data (3+ years ideally — less = noisier forecast), (2) the model choice (simple — moving average / exponential smoothing for stable; ARIMA for trend + seasonality; Prophet for business series with holidays), (3) the features (for ML models — price, channel mix, marketing spend, seasonality), (4) the validation (out-of-sample test on most recent period), (5) the uncertainty bands (forecast point + 80% confidence interval — reality lives in a range), (6) the scenario modelling (best case / base case / worst case — name the assumption that drives each), (7) the assumptions list (what would invalidate the forecast), (8) the update cadence (monthly — last-month actuals re-anchor the model). AUD. Tool: {{tool}}.