Business planning relies heavily on time-series forecasts. However, these forecasts often prioritize objectives that do not align with the goals of the business, resulting in forecasts that do not meet business preferences. In this study, we show that optimizing traditional forecasting metrics can lead to suboptimal business performance. Specifically, we focus on inventory management and develop a procedure to compute and optimize proxies of commonly used business metrics in a differentiable manner. We explore various cost trade-off scenarios and demonstrate through empirical evidence that end-to-end optimization often outperforms the optimization of standard forecasting metrics (with improvements of up to 45.7% for a simple scaling model and up to 54.0% for an LSTM encoder-decoder model). Finally, we discuss how our findings can be applied to other business contexts to benefit their planning and forecasting processes.
Forecasting Inventory Management with a Focus on Business Metrics
