Demand Intelligence
What will customers buy, where and when?
Forecasts demand across 32 store and SKU combinations with 18 months of history, learning calendar and promotion effects and excluding censored days (799 in the sample). Holdout WAPE is 22.1% against a 27.9% weekday baseline.
- Demand by store, SKU and week
- Promotion and calendar effects
- Scenario testing before publishing
- Forecast published to inventory
- Duration: 4–6 weeks for an outcome-based POC
How the AI works. Forecasting models predict demand for every item and location over twelve weeks and are tested against weeks that were held back before anyone relies on them, with the error (WAPE), bias and forecast range shown openly. Planners see where the AI forecast is strong and where it is not, and the what-if carries the forecast straight into a stocking decision.
Planners can try scenarios (stronger promotion, demand up or down) and publish a reviewed forecast that Inventory then consumes.