Analysis of Various Forecasting Approaches for Linear Supply Chains based on Different Demand Data Transformations

Supply Chain Management (SCM) has become a popular topic in recent years. One of the main discussion points is the effective management of inventories along the supply chain. Several inventory policies have been analyzed and compared to overcome the traditional tradeoff between high service levels and high inventory costs and vice versa. This paper describes how different demand data aggregation affects reorder point calculations in continuous inventory review systems and analyzes the impact on inventories, safety stocks, and service levels of the supply chain members. Based on a linear three-stage supply chain, four different forecasting scenarios based on simple means and variance calculations as well as moving average and moving variance estimations have been tested. To analyze potential effects for different supply chain settings, four demand patterns were implemented (stationary, season, trend, trend and season). The simulations reveal different effects depending on demand data aggregation and customer demand structure. Since the assumption of normally distributed demand data is violated for upstream suppliers in linear supply chains, difficulties arise particularly in calculating safety stocks. Aggregating order data can mitigate some of the biases in several cases. It is shown that forecasting monthly aggregated orders outperforms the other strategies in terms of lower mean inventories and safety stocks, but may lead to slightly lower service levels.