A SIMULATION-BASED ROLLING FORECAST APPROACH FOR THE SELECTION OF ORDERING STRATEGY IN THE SUPPLY CHAIN

The accuracy of demand forecast is critical to inventory management. Rolling forecast is a useful tool for lowering risk of inventory management in practical. Customers' estimation of rolling forecast is still deviated from realized orders because of turbulent environment. The issue of rolling forecast has bothered firms until now, yet little research addresses it. We adopted a simulation approach, in which a model is built up and examined in several experiment settings. How to decrease inventory costs by introducing ordering strategies and considering lead-time and stock-out penalties is explored Ordering strategies include net requirements and adjustments determined by average, SD and RMSEA of differences between historical forecast and realized data. Levels of estimation error and bias in rolling forecast are comprised with experimental settings. The results reveal that the ordering strategy considering RMSEA is the best in the situations of normal and under estimation bias, whereas the strategy considering average is the best in the situation of over estimation bias. The level of estimation error is irrelevant to selection of ordering strategies, but is positive association with inventory costs. Stock-out penalties and leading time are positively associated with inventory costs. So, accuracy of rolling forecast is emergent in the situation of long leading time and high stock-out penalties.