Should firms invest in demand forecasting? Benefits of improving forecasting accuracy on order smoothing, dual-sourcing and multi-stage supply chain problems

We investigate how improvements in demand forecasting–defined as transitioning from constant demand forecasting to minimal-mean-squared-error demand forecasting–affect managing inventory using the proportional order-up-to replenishment policy. While the existing literature extensively explores the impacts of demand forecasting under the normal order-up-to policy, its impacts under the proportional order-up-to policy remain largely unexplored. Our analysis covers order-smoothing, dual-sourcing, and multi-stage models. We emphasise the impact of auto-correlated demand processes, which are common in practice, on the desirability of minimal-mean-squared-error forecasting. Our results reveal that minimal-mean-squared-error demand forecasting only yields significant performance improvements for a small subset of AR(1) demand processes and can even lead to deteriorated performance in certain cases. Additionally, strategies such as dual-sourcing and collaboration through order smoothing reduce the need for high forecasting accuracy.