Unpacking the cognitive processes of the boundedly rational newsvendor

We report on the findings of five controlled experiments that test the effectiveness of proposed interventions aiming to improve both the process and outcomes of decision making when performing the newsvendor task. We use eye-tracking technology aided by interviews to gain new insights into people's cognitive processes. Our results confirm that human newsvendors exhibit the tendency of mean-reverting demand prediction using past demand even when the demand is independent and identically distributed (i.i.d.). This points toward the presence of the cognitive bias known as the law of small numbers in the decision-making process. The interventions involving ordering multiple products instead of the same product repeatedly, feedback frequency reduction, training and provision of optimal solution, all reduce reliance on the past demand, thereby reducing the extent of the mean-reverting tendency. Yet, none of these interventions eliminated this tendency completely; overall, judgmental forecasting persisted, albeit at a lower extent or in a different form. This tendency provides a plausible explanation for the fact that the interventions (and the resultant reduction in reliance on past demand) fail to lead to performance improvements. While the qualitative analysis of the interviews shows the subjects' intention to not rely on past demand in some instances, their gaze behavior shows that the past demand continued to impact their orders to varied degrees. Mere reduction (or even removal) of irrelevant information does not automatically make the subjects shift their focus to the relevant information and use them as intended.