Objectives: The purpose of this study was to develop a realist evaluation paradigm in social work evidence-based practice. Method: Wraparound (at Gateway-Longview Inc., New York) used a reliable outcome measure and an electronic database to systematically collect and analyze data on the interventions, the client demographics and circumstances, and the outcomes. Results: Research designs (e.g., single-system and group) can fall into place naturally and patterns in the data can be investigated using data analysis approaches such as binary logistic regression. Conclusions: Realist evaluation can be performed at regular intervals when an outcome measure is repeated, with the purpose of investigating what interventions work and in what circumstances. (Contains 5 tables and 3 figures.)