This paper addresses the very important concern that there are many social problems for which there are few or no evidence-based interventions.  Recognizing that resources for new programs are scarce, this paper emphasizes how previous research can be used to guide innovation and help to expand the evidence base.  The paper provides a number of suggestions for developing new intervention approaches, including examination of results from meta-analyses.  In addition, approaches can be informed by an assessment of basic and longitudinal research and reliance on “kernels” – indivisible units in a behavioral process found to have a causal effect on a particular outcome.  A “kernel” might provide reinforcement, change an antecedent factor, or directly affect an outcome.  The paper emphasizes the importance of learning from the past, avoiding harm, working over time to incorporate feedback on how program elements are working, and assessing ongoing evidence on whether intended goals are being achieved.