Impact of Supply Learning When Suppliers Are Unreliable

Dual sourcing and inventory are two prevalent and widely studied strategies firms use to manage yield risk. A pervasive but implicit assumption in the literature is that a firm knows its suppliers' yield distributions with certainty. This is a strong assumption in many circumstances. A firm is more likely to have a forecast of a supplier's yield distribution and to update that forecast based on its experiences with the supplier. We introduce and analyze a Bayesian model of "supply learning" (i.e., distribution updating) and investigate how supply learning influences both sourcing and inventory strategies in dual-sourcing and single-sourcing models, respectively. In the case of Bernoulli all-or-nothing yield distributions, we completely characterize the firm's optimal sourcing and inventory decisions for the supply-learning model. Among other results, we prove that for a given expected supplier reliability (i.e., the mean of the firm's forecast for the probability of successful delivery) an increase in the reliability forecast uncertainty increases the attractiveness of a supplier, but it reduces the firm's desire to invest in inventory to protect against future supply failures. We extend our analysis to allow for general yield distributions, multiple sourcing (i.e., more than two suppliers), and inventory carryover in the dual-sourcing model.