Teaching students with externalising problem behaviours is difficult for teachers, as it challenges the relationship that teachers engage in with their students. In this study, effects of Key2Teach on externalising and social-emotional problem behaviours and the mediating role of conflict in the teacher-student relationship were studied using a randomised controlled trial (RCT)-design. In two cohorts, 103 teacher-student dyads and peer-students (n = 1643) were assessed two times during a school year. Fifty-three dyads received the intervention (experimental group), whereas 50 dyads received no intervention (control group). Data were collected on teacher-reported externalising and social-emotional problem behaviours in students, and on teacher-student conflict. For dyad-students, results showed a direct effect of Key2Teach on conduct problems and an indirect effect, via teacher-student conflict, for hyperactivity problems. No effects on social-emotional problems were established. For peer-students, results showed indirect effects on externalising problems and direct effects on social-emotional problems. Implications for practice and future research are discussed.