An improved forecasting approach to reduce inventory levels in decentralized supply chains

Abstract

This paper covers forecast management in decentralized supply chains. For various reasons, companies do not always agree to disclose their information. To deal with this issue, we consider a downstream demand inference (DDI) strategy in a two-level supply chain. DDI was assessed using different forecasting methods and was successfully tested using only a simple moving average. In an investigatory context using other forecasting methods, we propose the introduction of the weighted moving average method, which affects nonequal weights to past observations. First, we verify the unique propagation of demand processes. Second, we consider the forecast mean squared errors, the average inventory levels and the bullwhip effect as the supply performance metrics. Third, we formalize the manufacturer's forecast optimization problem and apply Newton's method to solve it. The optimization results, based on the simulated demands, confirm the effectiveness of our approach to produce further enhanced solutions and to improve the results of DDI. We have shown that a little change in the weights of the forecast method improves the competitiveness in the market. Conversely, the bullwhip effect is affected due to the nonequal weighting in the forecast method.