This study reports a randomized controlled trial to improve teacher-student-parent feedback, conducted in a rural county in China with many left-behind children. Data are collected from over 4,000 primary schoolchildren (8 to 10 years old) over two school terms. We find that bi-weekly student feedbacks using our special scorecard of schoolwork and behavior improve mathematics results by 0.16 to 0.20 standard deviations, with 0.09 for language. Communicating these assessments also to parents produces further large mathematics benefits for young left-behind children, about 0.30 standard deviations. A low-cost investment in better feedback thus brings significant achievement gains especially for disadvantaged children.