The influence of match length and re-matching on the effectiveness of school-based mentoring was studied in the context of a national, randomized study of 1,139 youth in Big Brothers Big Sisters programs. The sample included youth in grades four through nine from diverse racial and ethnic backgrounds. At the end of the year, youth in intact relationships showed significant academic improvement, while youth in matches that terminated prematurely showed no impact. Those who were re-matched after terminations showed negative impacts. Youth, mentor, and program characteristics associated with having an intact match were examined. Youth with high levels of baseline stress and those matched with college student mentors were likely to be in matches that terminated prematurely, while rejection-sensitive youth and mentors who had previous mentoring experience were more likely to be in intact relationships. Implications for research and practice are discussed.