Downstream Demand Inference in decentralized supply chains

For many years, the main objective of studying decentralized supply chains was to demonstrate that a better inter-firm collaboration could lead to a better overall performance of the system. The literature has demonstrated that collaborating by sharing information, even in a partial way, can lead to near-optimal solutions. In this context, many researchers have studied a phenomenon called Downstream Demand Inference (DDI), which presents an effective demand management strategy to deal with forecast problems. DDI allows the upstream actor to infer the demand received by the downstream actor without the need of information sharing. Recent research showed that DDI is possible with Simple Moving Average (SMA) forecast method, and was verified for an autoregressive [AR(1)] demand process, a moving average [MA(1)] demand process, and an autoregressive moving average [ARMA(1, 1)] demand process. In this paper, we extend the strategy's results by considering causal invertible [ARMA(p, q)] demand processes. We develop Mean Squared Error and Average Inventory level expressions for [ARMA(p, q)] demand under DDI strategy, No Information Sharing (NIS) and Forecast Information Sharing (FIS) strategies. We compute the Bullwhip effect generated by employing SMA method and we simulate the resulted improvement compared to employing MMSE method. We analyze the sensibility of the three performance metrics in respect with lead-time value, SMA and ARMA(p, q) parameters. We compare DDI results with NIS and FIS strategies' results and we show experimentally that DDI generally outperforms NIS. Finally, we provide a revenue sharing contract as a practical recommendation to incite supply chain managers to adopt DDI strategy.