How we think about forecasting
Methodology
Chopp describes its technology as AI forecasting that uses advanced time series analysis. This page sets out the principles that approach rests on, in general terms.
Principles
1. Start from the time series
Grocery sales are recorded over time, per product and per store. Trend, weekly cycles and annual seasonality are modelled explicitly rather than averaged away.
2. Explain variation with real-world drivers
Chopp names six families of drivers: weather patterns, local events, traffic flows, seasonality trends, holidays and schedules. Each is a measurable reason why demand departs from the usual pattern.
3. Judge forecasts by what they do in store
A forecast is only useful if it reduces waste and prevents stockouts. Accuracy metrics such as WAPE and bias are a means to that end. See forecast accuracy metrics.
4. Perishables first
The cost of a bad forecast is highest where shelf life is shortest: fresh produce, bakery, dairy, meat and prepared food.
Client to provide · Model and validation details
Add specifics if the client wants to publish them: model families used, forecast horizon, retraining cadence, validation results from pilots. No accuracy figures are stated on this site because none were published.
Ready to cut away the waste?
Tell us about your store or chain, and we will show you how AI sales forecasting can help you order the right products in the right quantities.