This paper describes a rigorous data collection process to find and verify the quality of what could potentially be high-functioning or high-performing schools accessible to the poor in three of South Africa’s nine provinces. A potential sample of ‘outlier’ schools is selected using system-wide Universal Annual National Assessment data corroborated against school recommendations collected from a variety of system actors expected to be informed about school quality. Unfortunately, literacy testing in 31 purposively selected schools yields no example of high-performing, no-fee schools. However, we identify outlier or resilient students even in underperforming schools. Furthermore, within the no-fee school system there exists a continuum of functionality. Schools exist that while far from reaching good (or even adequate) median levels of English literacy, exhibit relatively higher literacy levels than other sample schools after controlling for student background differences. The presence of these relatively better performing sample schools (and performance variation more generally in the no-fee system) suggests that there is a middle-ground, a rightward movement away from dysfunction that can be reached. However, it is not clear that all system actors are able to detect variations in school quality. Our sample of respondents recommending ’good’ schools are only able to identify slightly better performing no-fee schools. For certain groups, specifically education district officials, enrolment growth appears to be a better indicator of their perceptions of ‘good’ than measures of student performance.