Financial education is a high priority for policymakers around the world. Many developing economies, including China and India, and the vast majority of OECD member countries, have implemented national financial education strategies. As enormous resources are being invested into financial education, the effectiveness of financial education programs is an exigent issue. Accordingly, the issue of effectiveness is intensively debated (see Hastings et al., 2013; Lusardi and Mitchell, 2014), albeit with a change in focus. Some years ago, studies raised concerns about the reliable effectiveness of financial education (e.g., Cole et al., 2011; Fernandes et al., 2014), but with a rapidly increasing number of eld experiments, new meta-analyses show that effects are positive, on average (Miller et al., 2015; Kaiser and Menkhoff, 2017; Kaiser et al., 2020). Thus, more recent debate focuses on improving effectiveness of interventions. Therefore, researchers are evaluating alternatives to typical lecture-based classroom-programs. Proposals include tailoring interventions to narrow target groups (Doi et al., 2014; Drexler et al., 2014), introducing personalized elements like counseling (Carpena et al., 2017), and using mass media (Berg and Zia, 2017), among others. In contrast, we study an approach that aims for improving effectiveness within the still pervasive classroom-based programs, i.e., by applying the concept of active learning (Freeman et al. 2014) to the field of financial education: We conduct a cluster randomized eld experiment in 83 villages to study the differential effects of two financial education interventions offered to 1,291 small-scale business owners in rural western Uganda. The treatments contrast "active learning" with "traditional lecturing" within standardized lesson-plans while keeping content, teachers, and intensity of training constant. Three results emerge: First, and our main finding, is that the group randomly allocated to the "active learning" program experiences a significant improvement in financial behavior whereas there are no effects on the group allocated to "traditional lecturing." Specifically, the active learning program causes an increase in two out of five outcome-domains, each captured by an index. The strongest impacts occur in the savings and investment domains: in the short-term, i.e., until midline at 6 months post treatment, total savings increase by 21 percent relative to the control group, and total investments into the own business by 33 percent. In the long-term (3.5 years after treatment) these increases relative to the control group do not flatten out but become stronger, suggesting that at least part of the "active learning" group has changed their financial behavior persistently. Second, we show in heterogeneity analyses that the benefits of "active learning" are especially pronounced above the median of the respective outcome distributions, indicating that average treatment effects of the financial education program may be driven by responses from individuals who face relatively fewer constraints prohibiting behavior change. Third, regarding potential mechanisms of "active learning" treatments, we provide tentative evidence that the strong impact of this approach may be supported by stimulating increases in financial literacy, self-control, and financial confidence. An additional social mechanism may explain the observed treatment effects to a greater extent, i.e., active learning leads individuals to join social groups in order to discuss their personal financial matters. Our research adds to the growing literature that advances the understanding of the differing impact of delivery channels of financial education and seeks to contribute to the debate on how to design effective and scalable training programs in developing countries.