An Evaluation of Statistical Methods For Workload Forecasting at the DDNV Material Processing Center

Workload surges can occur intermittently throughout the year at the Defense Logistics Agency Distribution Norfolk, Virginia (DDNV) Material Processing Center (MPC). These unexpected backlogs cause receipt delays, incur extra costs for overtime and requirements for additional personnel, and negatively impact fleet readiness. The DDNV MPC is one of the largest materiel processing centers in the world,handling more than 40,000 transactions per month. Much of the workload demand is from the local fleet forces, to include more than 70 Norfolk-based Navy surface ships, submarines, numerous shore commands,and other Air Force and Army installations throughout the Hampton Roads area. Currently, there is no codified way forecasting is done aside from examining historical data and adjusting based on MPC input,and time-series analysis techniques do not work well for this type of intermittent demand. Currently,workload is forecasted using ad-hoc techniques in Excel. In this thesis, various forecasting techniques with a primary emphasis on variations of Croston's intermittent demand forecasting were evaluated. Pertinent data were collected primarily from the Defense Logistics Agency, with additional data collected from its Distribution Standard System (DSS) and Fleet/Type Commanders. The forecasting methods were evaluated using historical data.