We study a mentoring program that aims to improve the labor market prospects of disadvantaged adolescents. Our randomized controlled trial investigates its effectiveness on three outcomes highly predictive of later labor market success: math grades, patience/social skills, and labor market orientation. For low-SES (socioeconomic status) adolescents, the mentoring increases a combined index of the outcomes by over half a standard deviation after 1 year, with significant increases in each outcome. Effects on grades and labor market orientation, but not on patience/social skills, persist 3 years after program start. By that time, the mentoring also improves early realizations of school-to-work transitions for low-SES adolescents. The mentoring is not effective for higher-SES adolescents.