Evaluating a Bayesian approach to demand forecasting with simulation

At The Boeing Company, stock levels for maintenance spares with substantial lead times must be established before fielding new aircraft designs. Initial calculations use mean time between demand estimates developed by the engineering department. After sufficient operating hours, stock levels are recalculated using statistical forecasts of maintenance history. A Bayesian forecasting method was developed to revise engineering estimates in light of actual demand on new aircraft programs. Three forecasting methods were evaluated: Engineering Estimates, traditional Statistical Forecasting, and Bayes´ Rule. Stock levels were established using inventory optimization, and fill rate performance was evaluated using warehouse simulation. The proposed Bayesian approach outperforms the other methods, enabling the inventory optimization model to establish stock levels that achieve higher fill rate, resulting in better initial inventory investment decisions. This paper´s contribution is comparing spares forecasting approaches for a well-defined set of airplane parts using a carefully constructed inventory optimization and simulation test environment.