Many traditional forecasting platforms try to make predictions about the future by analyzing patterns that occurred in the past. More sophisticated methods exist, such as exponential smoothing or the Holt-Winters method. Nonetheless, as sophisticated as these methods are, they all suffer from the same deficiency: they only consider the past. While the past is indeed a statistically significant indicator of what will happen in the future, it isn't the only indicator. In multi-echelon supply chain systems, there exists a plethora of highly relevant data that can be tapped into. We have built an Arena model to simulate our manufacturing plant's supply chain. As we go forward, we will explain how this model was used to explore the significance of various external data sources such as DC and store inventory levels, open orders, and lead times. Using these findings, we derive a regression-based forecasting algorithm that improves accuracy by 35 percent. More accurate forecasts allow our company to maintain a higher service level, lower safety stock, and a more predictable transportation schedule, all of which lead to significant cost savings.
展开▼
