Guide
Time series forecasting
Chopp’s technology uses advanced time series analysis to predict grocery sales. Here is what that means.
Sales as a time series
A time series is a sequence of observations ordered in time, such as daily sales of one product in one store. Time series analysis looks for structure in that sequence:
- Trend: the long-run direction of demand.
- Seasonality: patterns that repeat by day of week, month or year.
- External effects: changes explained by outside drivers such as weather, events or holidays.
- Noise: random variation that cannot be predicted.
Common method families
Exponential smoothing (including Holt–Winters) weights recent observations more heavily and handles trend and seasonality. ARIMA models describe a series in terms of its own past values and errors. Regression with external variables adds drivers such as temperature or holidays. Machine learning methods, including gradient-boosted trees and neural networks, can learn from many series and many drivers at once.
Validation
Forecasts are tested by back-testing: training on past data and checking predictions against what actually happened later. See forecast accuracy metrics.
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.