A growing body of research has demonstrated an important relationship between work experience and teacher productivity. This implies that educational quality can be improved through reduction in turnover or acceleration of the return to experience. Mentoring has become an extremely popular policy to achieve these goals, but little is known about its general impact on teachers. I study the impact of mentoring on new teachers in New York City, which adopted a nationally recognized mentoring program in 2004. I first measure the average impact of the program using a difference-in-differences framework and find positive but statistically insignificant estimates of the effects of mentoring on teacher outcomes. I then use detailed program data to examine the relationship between outcomes and measures of mentoring quality. Although assignment of teachers to mentors was non-random, I use within-school comparisons and instrumental variables to address potential sources of bias in my estimates. I find strong relationships between measures of mentoring quality and teachers' claims regarding the impact of mentors on their success in the classroom. I also find some, albeit weaker, evidence that these measures of mentoring quality had significant impacts on teacher absences, retention, and student achievement. The most consistent finding is that retention within a particular school is higher when a mentor has previous experience working in that school. Importantly, I find evidence that student achievement in both reading and math are higher when teachers receive additional hours of mentoring. I also find that retention is higher among teachers who received other types of support (e.g., common planning time), consistent with work by Smith and Ingersoll (2004).