Non-cognitive skills programs may be an important policy option to improve the academic outcomes of adolescents. In this paper, we evaluate experimentally the EPIS program, which is based on bi-weekly individual or small-group non-cognitive mediation short meetings with low-performing students. Our RCT estimates, covering nearly 3,000 7th and 8th-grade students across over 50 schools and a period of two years, indicate that the program increases the probability of progression by 11% to 22%. The effects are stronger amongst older students, girls, and in language subjects (compared to maths).