Managing perishable inventory, such as blood stock awaiting use by patients in need, has been a topic of research for decades. Yet, most research focuses on the effective use of this scarce resource across the blood supply chain, and assumes the supply of blood itself can be impacted only via coarse policy levers. In this talk, empowered by the widespread geographic coverage of social networks, we choose instead to model the first stage of the full blood supply chain---that is, the supply of blood itself---as a matching market. Here, potential blood donors are matched to donation centers by way of a central recommendation engine; that engine can at some cost prompt (e.g., via push notification) individual donors to donate to a preferred center or centers. Potential donors may have other constraints (e.g., maximum allowable frequency of prompts) or preferences (e.g., geographic, social) that should be taken into account. We develop policies for matching potential blood donors to donation centers under these constraints and preferences, and under simple models of demand shocks to the system. We provide preliminary experimental results in simulation using real data from a large, worldwide social network, and show that our system provides lift relative to the status quo allocation methods.