This paper presents and analyzes the results of a decision-making experiment in inventory management under uncertainty. The experiment included 81 participants who played the role of a small car importer facing random demand as in the (Q,R) model. The results show strong evidence of learning and convergence, and the average reorder point (R) closely approaches the optimal level for maximizing profits. However, the participants' decisions are still biased by realizations of extreme values of demand and loss of potential sales. We argue that participants are affected by recency, loss aversion, and, possibly, their own risk aversion.