Forecast Enrichment

Empirical studies show that incorporating weather and event data into forecasting models can produce dramatic accuracy gains. One study cited in the Journal of Textile Science and Technology in February 2023 found that including weather data could reduce sales forecasting errors by 8.6% to 12.2% on average and by 50.6% during summer weekends. Another study cited in the same article found that adding location, season, and product category to the weather data would allow retailers to increase revenue by 2%.

In the grocery industry, weather-adjusted demand forecasting can reduce forecast errors for weather-sensitive items during atypical conditions, such as heat waves, deep freezes and severe storms. Retailers have used this capability to reduce spoilage and improve replenishment. For example, Albertsons Companies Inc. observed sharp demand increases for soups and chilis during seasonal temperature drops in the Northeastern United States, while coffee sales fluctuated by an average of 5%—rising as much as 10% during cooler periods.

Instacart, which delivers orders to consumers from more than 100,000 stores in North America, used to spend hundreds of hours mapping out events such as parades and marathons to determine how they would impact delivery routes. By using AI-powered event-tracking technology from PredictHQ that uses APIs to automatically show maps of affected areas, Instacart has cut that research time in half, according to a PredictHQ case study.