Much hope is placed on education systems to reduce socio-economic learning gaps. However, in South Africa, uneven functioning of the school system widens learning gaps. This paper analyses education performance using ANA data. Weak calibration and inter-temporal or inter-grade comparability of ANA test scores limit their usefulness for measuring learning gains. However, relative performance provides meaningful information on learning gaps and deficits. A reference group that is roughly on track to achieve the TIMSS average is used to estimate the performance required in each grade to perform at TIMSS' low international benchmark. By Grade 4, patterns across quintiles of on-track performance approximate matric exemption patterns. Viewed differently, academic and labour market prospects may be bleak for children who are no longer on track. Improvement in outcomes requires greater emphasis on the Foundation Phase or earlier, before learning deficits have grown to the extreme levels observed by the middle of primary school. This statement is true whether deficits arise from weak early instruction, or simply because a disadvantaged home environment requires early remedial action. The emphasis on the early grades that this analysis of the ANAs suggests is contrary to the conclusions drawn from the ANA results by policy makers that weak test scores in Mathematics in Grade 9 require major interventions in that grade.