The education of the poor and disadvantaged population has been a long-standing challenge to the education system in both developed and developing countries. Although computer assisted learning (CAL) has been considered one alternative to improve learning outcomes in a cost effective way, the empirical evidence of its impacts on improving learning outcomes is mixed. This paper intends to explore the nature of the effects of CAL on student academic and non‐academic outcomes for underserved populations in a developing country. To meet this goal, a randomised field experiment of a CAL programme involving over 4000 third grade students, mostly from poor migrant families, in 43 migrant schools in Beijing was conducted. The main intervention is a math CAL programme that is held out of regular school hours. The programme is tailored to the regular school math curriculum and is remedial in nature. The results show that, the CAL programme improved the student standardised math scores by 0.14 standard deviations and most of the programme effect took place within two months after the start of the programme. Low‐performing students and those with less educated parents benefited more from the programme. Moreover, CAL also significantly increased the levels of self‐efficacy of the students and their interest in learning. At most a moderate programme spillover in Chinese test scores was observed. Our findings are robust to the Hawthorne effect and CAL programme spillovers that might potentially bias the estimates of the programme effects.