Demand Forecasting & Targets
Three models forecast every SKU and territory, get scored on their real error, and turn into targets automatically.
- 3
- models, scored on MAPE
- Every
- override is recorded
- 12
- capabilities on this page
01Inside the product
02What ships
Built from Projection Engine · Manager Dashboard · Management Dashboard · Distributor Portal
- Before
Last year plus ten percentNowPer-SKU, per-territory forecasts - Before
One analyst's spreadsheetNowThree models, scored on MAPE - Before
Overrides lost in emailNowEvery override on the record
Three models, compared
Projection Engine- Weighted-average forecast
- Prophet
- XGBoost
- MAPE tracking
Judgement stays in the loop
Projection Engine · Platform- Offer-uplift factor
- Manual override
- Forecast accuracy
- Audit logging
The forecast becomes the plan
Manager · Distributor · Manufacturing- Auto-target generation
- Target gauges
- Stock forecast
- Demand signal
03How it runs
- 01
Learn
Secondary sales train all three models.
- 02
Adjust
Planned schemes lift the forecast.
- 03
Override
Planners can overrule — on the record.
- 04
Cascade
Targets, stock and the plant plan follow.
Who works in this
- GMManagementWeb
- AMArea & regional managerWeb
- PHPlant headWeb
Fed by, and feeding
- Secondary sales
- Validated uploads as history.
- Prophet & XGBoost
- Beside a weighted baseline.
- Targets
- Published after manager review.
- Plant & distributor
- Demand signal and stock forecast.
04Questions
Because no single model wins everywhere. A steady staple is forecast well by a weighted average; a festival-driven SKU needs Prophet's seasonality; a line with many interacting drivers suits XGBoost. All three run, all three are scored on MAPE, and you can see which is winning for each SKU and territory.
Next in the chain
Every platform page- Production PlanningPlant schedule, materials and capacity against real demand7 capabilities
- Sales AnalyticsTerritory P&L, sell-through gap and distributor performance12 capabilities
See demand forecasting on your own chain.
Thirty minutes, on demo data shaped like your business — your roles, your beats, your SKUs. Pilot one area before you commit to anything.