Adjusting supply chain forecasts for short‐term temperature estimates: a case study in a Brewing company

One of the key drivers of seasonal behaviour in time series is weather, most notably temperature. Very often weather effects are inter‐correlated with the impact of other one‐off events like public holidays or promotional activity. Due to the uncertainty that comes with short‐term weather forecasts, the exact effect of weather on weekly or daily sales has been given very little attention in the forecasting literature. The present study evaluates the impact of weather‐driven adjustments to forecasts in a Brewing company. The forecasting team applies a decomposition approach, where: (a) an exponentially smoothing model is used in order to produce weekly sales forecasts; (b) an econometric model is built once a year in order to estimate the impact of 10‐day ahead temperature changes in sales (as an input to this model, weekly weather forecasts from the Met office are used); (c) the sales forecasts from the former are adjusted based on the impacts from the latter, as well as for promotions, one‐off events and regular seasonal behaviour. Empirical findings suggest that the weather adjustment mechanism improves the forecasting function in the company.