Where Did the Demand Go? Teaching Demand Censoring with the Newsvendor Challenge

Abstract

The newsvendor model is widely used to teach decision making under uncertainty. Traditionally, analytical methods have been taught to determine the optimal order quantity that balances missed profit from ordering too few units against the cost of excess inventory from ordering too many. In practical settings, however, organizations must estimate these costs using censored data as only sales are observable in the data, not the true demand. Neglecting demand censoring can, therefore, lead to underestimating lost sales. These concepts can be difficult to grasp in a classroom setting. Hence, we developed a fun classroom challenge to simulate the newsvendor problem with data and demand censoring. In the challenge, students are provided with hypothetical, long-term sales data and are tasked with determining the optimal number of units to order per period. The challenge extends the traditional newsvendor model by including censored data such that traditional approaches result in suboptimal decisions. The challenge also emphasizes the importance between time spent on predictions (forecasting demand) and prescriptions (order decisions). The challenge can be adapted to diverse student cohorts, and experience reveals productive class discussions as students compete to determine the best order policies.