Forecast StrategyFOOD & US / DEMAND PLANNING / GOVERNANCE
Forecast Strategy Workbench
Seasonality & Events Calendar
October 2026
1 · Scope
2 · Methods & inputs
3 · Backtest
4 · Compare
5 · Approve
1 · Set the forecasting scope
Select the history and the time bucket used to compare past forecasts against actual demand. Each model gets the same training and holdout cutoffs.
2 · Candidate methods
Select models to test. Each method lists its input contract.
Input data for selected methods
Required fields are fixed by each method. Choose optional drivers only when they are available at forecast time.
Sales historyRequired
Required · Client VAT sales register; transaction/returns rule and units to confirm
Required for active method
Pipeline & one-off upliftRequired for client formula
Required for client formula · Sales_Prediction; mapped daily uplift
Required for active method
CalendarOptional
Optional · Holidays / calendar dates; availability to validate
Use this input where supported
PromotionsOptional
Optional · Promotion dates and intensity; not in verified snapshot
Use this input where supported
PricesOptional
Optional · Historical and known future prices; not in verified snapshot
Use this input where supported
3 · Backtest and compare
Inspect actual-versus-forecast time buckets. Live backtesting evaluates candidate methods against canonical sales invoices from the database.
Choose methods and inputs, then run the backtest. Ready to compare selected candidates.
4 · Select and approve
Select a tested method, inspect its input contract and approval impact, then create an approved strategy version.
Proposed strategy v2: Seasonal / moving baseline for the selected scope. Run a comparison to enable selection. Approval requires validated backtest evidence in the live product.
SCM Platform Demand Governance · Live transactions evaluated from canonical sales ledger. Comparative metrics reflect empirical backtest holdouts.