The Bayesian newsvendors in supply chains with unobserved lost sales

We consider two-echelon supply chains with one supplier and two retailers. Retailers are censored newsvendors facing general parametric demand distributions involving unknown parameters. Using a Bayesian MDP formulation, we investigate how the supplier can make use of the combined information gathered from the retailers' sales data to increase channel pro ts. We compare among the following three scenarios: (i) supplier shares her (pooled) demand distribution updates with the retailers, (ii) supplier does not use pooled updates, and (iii) integrated firm. We show that buy-back contracts achieve channel coordination in the above system.