Lindsay Vance, Business Intelligence Manager
Pabst Brewing Company
In addition to their flagship brand of Pabst Blue Ribbon Beer, the Pabst Brewing Company distributes around 30 additional brands of beer, using 700 distributors through the U.S. Each distributor submits weekly orders to Pabst for cases of beer in eight different packages, which represents the demand for Pabst products. Based on the history of these orders for more than 16,000 combinations of distributors, products, and packages, Pabst wanted to anticipate the orders for the next two years. Accurate beer demand forecasts would help Pabst reduce overproduction, which is wasteful due to the perishable nature of beer, and minimize underproduction, which represents opportunity costs.
With the support of Philadelphia-based CompassRed Data Labs, the Pabst Brewing Company is now using time series modeling to produce more than 16,000 customized beer demand forecasts by distributor, product, and package. The historical data has been modified as needed to reduce the influence of unusual fluctuations ‘noise,’ which is particularly important for the ‘innovation’ products that Pabst frequently adds to their portfolio.
These forecasts are distributed through a business intelligence report that allows the sales managers to select forecasts for a specific distributor, product, and package. In addition to the beer demand forecasts, this report includes the inventory of cases on hand and the ideal number of cases in inventory to enable the sales managers to anticipate the number of cases of beer required for future weeks’ shipments.
These forecasts enable Pabst to better align their production with demand, thereby reducing the waste associated with overproduction and the opportunity costs of underproduction. Forecasts are generated for a two-year horizon. The forecasts are distributed to Pabst managers throughout the country using a business intelligence product. In addition to the forecasts, additional information is provided, such as inventory total and inventory targets. Each week, more than 16,000 forecasts are generated based on its algorithms.