The purpose of the study was to find out the effect of active learning on Mathematics and Science learning outcomes, and the effective length of intervention on student learning outcomes. This was a quantitative ex-post facto research conducted in 7 provinces in Indonesia consisting of 50 districts and cities. The number of samples was 368 schools (34%) out of total 1,075 partner and dissemination schools. Data were collected through test on mathematics and natural sciences ability. Test the hypothesis used Mann Whitney average difference test and Manova was used to determine the effectiveness of interventions. The result shows that students’ score in mathematics and science was better in partner schools compared to dissemination school. The Bonferroni test of mathematics scores shows that the best intervention was three years. The Bonferroni test of the science score showed a four-year as good as three-year intervention. The implementation of active learning was proven to improve students learning outcomes.