MVP lab · temporal validation · decision support
ForecastOps
Operational forecasting with uncertainty: transparent baselines, temporal backtesting, interval fan charts and inventory what-ifs on synthetic demand.
Caveat: synthetic operational lab — not production financial forecasting, and not unattended purchase automation.
Demo dataset
365 synthetic daily observations for one SKU.
Forecast cockpit
Choose a baseline, inspect the fan chart, then read backtest metrics before trusting the curve.
Model comparison
Rolling temporal folds — never shuffled. Lower MAE wins for this demo SKU, not for every business.
Scenario simulator
Stress demand, then read order quantity, stockout risk and days of cover as decision cards — still human-reviewed.
Action plan
Methodology (interview-ready)
- Regularize daily sales and keep time order intact.
- Fit transparent baselines before any complex model.
- Score with rolling temporal folds (MAE, MAPE, RMSE, coverage).
- Communicate uncertainty with a horizon-widening interval.
- Translate forecast into inventory what-if cards with explicit limits.
Deep dive: docs/methodology.md · architecture: docs/ARCHITECTURE.md