From Pop Tarts to snow blowers, consumer product sales are being driven by the weather and predicted with increasing accuracy by IBM.
AI-driven analytics pair with detailed weather forecasting to help merchants know what specific items shoppers will want to purchase, when and where they will shop, and how much they will buy. IBM has fine-tuned its forecasting models so they can provide store-level forecasts for specific products (SKU’s).
Interestingly, the most important drivers in the models are not economic but weather related.
Video Spotlight: How to Improve Retail Demand Forecasting with Cognitive
This post is based on the Forbes article, The Science of IBM’s Holiday Retail Forecast—and Why Companies Count On IT, by Matt Hunter, November 27, 2019, and the YouTube video, How to Improve Retail Demand Forecasting with Cognitive, by IBM Services, January 12, 2018. Image source: Shutterstock/ra2studio
1. How do weather related forecasts predict buyer behavior?
Guidance: Current weather conditions can affect buying behavior, driving up demand for umbrellas or hot coffee, for instance. However, forecasts of longer range patterns tend to have a greater influence. For instance, this year much of the United States faced an early cold snap in November, prompting earlier than usual purchases of winter gear like coats and snow blowers. Retailers were prepared, because IBM’s forecasts included these weather patterns and alerted them ahead of time so they had plenty of inventory on hand. Similarly, some stores use IBM’s forecasts to stock up on items like prepared chicken or Pop Tarts that people look for when hurricanes are likely to strike.
2. What other factors are considered in IBM’s holiday forecasting models for retailers? What has helped improve their accuracy in recent years?
Guidance: Other variables associated with holiday forecasting models are economic inputs like the consumer price index, disposable income, inflation, unemployment, consumer debt, and regional economic conditions. IBM’s forecasting models go beyond these traditional inputs, however, by using proprietary AI algorithms and extremely detailed weather data. Advances in neural network technology in recent years have enabled Michael Haydock, Chief Scientist, Global Business Services for IBM, to really drill down into the specific buying habits of particular regional markets and provide store-specific forecasts.