The document proposes a framework to use social network data and transaction records to identify cohesive customer subgroups for targeted advertising. It defines a cohesion measure to identify subgroups where customers have strong social ties. Transaction data is used to estimate each product category's "liking" for each subgroup. A linear program then selects the optimal set of customers from high-liking subgroups to target for a given product. The approach was evaluated on a university library circulation dataset and outperformed random advertising.