Background: Consistent identification and prevention of failure for at-risk medical students is challenging, failing courses is costly to all stakeholders, and there is need for further research into duration, timing and structure of interventions to help students in difficulty. Aims: To verify the value of a new exam two weeks into medical school as a predictor of failure, and explore the requirements for a preventative intervention. Methods: Students who failed the two-week exam were invited to a series of large-group workshops and small-group follow-up meetings. Participants' subsequent exam performance was compared with non-participants. Results: About 71% of students who performed poorly in the new exam subsequently failed a course. Attendance at the workshops made no difference to short- or long-term pass rates. Attendance at more than three follow-up small group sessions significantly improved pass rates two semesters later, and was influenced by teacher experience. Conclusions: Close similarity between predictor task and target task is important for accurate prediction of failure. Consideration should be given to dose effect and class size in the prevention of failure of at-risk students, and we recommend a systemic approach to intervention/remediation programmes, involving a whole semester of mandatory, weekly small group meetings with experienced teachers.