Revenue Management Decision-Making Given Non-Stationary Demand and Two Alternative Decision Tasks

State-of-the-art revenue management systems combine mathematical algorithms for forecasting and optimization with human decision making. As analysts can overrule the automated demand forecast and inventory optimization, their decisions have a crucial impact on revenue management success. We present a new experimental design to examine the impact of non-stationary demand and alternative decision tasks on revenue management decision making. Furthermore, we present experimental results from a corresponding study, quantifying the gap between human decision making and mathematical optimization and evaluating decision errors. Our results highlight that human decision makers struggle to accommodate non-stationary demand. This triggers a loss-aversion bias, where decision makers accept too many low-value requests early in the booking horizon and rejecting too many requests later. When asked to set prices rather than accepting or rejecting individual requests, they overly rely on the predicted willingness to pay as an anchor for decision making. From these results, we draw implications for future research and for the design of symbiotic analytics systems.