In this article we propose a case selection strategy for comparative studies of school improvement. The strategy is based on two assumptions. First, we emphasize the importance of a strategic selection of cases (schools), which allows systematic comparisons between successful and failing schools. We claim that cases should be selected on the dependent variable, and, more specifically, on the variation in the dependent variable (i.e. school success or no school success). Second, we recognize the importance of a longitudinal perspective in the case selection process. We argue that quantitative longitudinal data should be used to identify consistently succeeding and consistently failing schools. Furthermore, we demonstrate the fruitfulness of the strategy by using it in a case selection process of successful and failing schools in Sweden. In sum, the strategy addresses the demand for research designs, which allow causal inference, in school improvement and school effectiveness studies. Systematic comparisons between successful and failing schools are essential to elaborate our insights of the mechanisms behind school success.