We analyze the impact of incentivizing students' effort during the school year on performance on high-stakes assessments in a field experiment with 63 low-income high schools and over 10,000 students. We contribute to the literature on education incentives by incentivising inputs rather than output, by focusing on high stakes outcomes, and by comparing financial and non-financial rewards. We take advantage of our large sample and rich data to explore heterogeneity in the effects of incentives, and identify a "right tail" of underperforming students who experience a significant impact on high stakes assessments. Among students in the upper half of the distribution of incentive effectiveness, exam scores improve by 10% to 20% of a standard deviation, equal to about half the attainment gap between poor and non-poor students.