Forecasting Drivers of Business Performance
Anticipating Business Trends Through Targeted Data Mining
Cascade successfully implemented a forecasting system for a trucking company to help set bid prices for prospective accounts. Cascade determined how price, geography, and various other customer attributes impacted the amount of future business from a prospective account. The system enabled the customer to profitably optimize its prices in a targeted fashion for thousands of customer bids annually. A challenge was posed by the customer’s very sparse historical demand data, which had zero values for many lanes, and so Cascade employed a statistical methodology specifically tailored to handle this type of data.
Cascade worked with a transportation services company implementing a forecasting system to help improve the recruitment of workers. Cascade determined what factors affected the labor market for workers, which included macroeconomic effects such as the unemployment rate. The client was able to use this information to test how many new workers would be available under a variety of optimistic economic scenarios.
Cascade also implemented a system for an air cargo company to forecast market demand as part of an optimization and sales planning system. This system featured multiple seasonal effects, as well as a “demand unconstraining” feature. The client’s historical record of flown cargo often under represents the true total market demand because aircraft have finite capacity; some portion of the total market demand can not be served by the client because of insufficient aircraft capacity. The Cascade model employed a statistical technique to “unconstrain” the historical data, i.e. estimate the total market demand – including the unobserved portion – from the client’s history of flown cargo.
The forecasting system also allowed for centralized revenue management and regional business managers to review and override the forecasts before they were used for planning purposes. The regional business managers were also able to set up short-term alterations to forecasts over the upcoming week, so that the optimization system could advise them on how best to deal with short-term opportunities.