BRAC has over 40,000 schools worldwide. It is widely praised for serving disadvantaged students and for matching or outperforming government schools. Using data that we collected from Dhaka’s slums, we test these claims. We find that BRAC serves the most disadvantaged students in our survey, but contrary to popular belief, BRAC students perform significantly worse than comparable students at other school types when we control for family demographics in a matching procedure. Anticipating our need to control for selection, we collected data on family demographics and the child’s fluid intelligence; since the latter affects both types of school and student performance, it unambiguously should be included in the propensity score. Once we control for fluid intelligence, the performance difference with other NGO schools disappears. The gaps between government and JAAGO schools have narrowed, but they still remain large and statistically significant.