The main aim of the present study was to carry out an in-depth examination of mathematics underperformance in Greece. By applying a binary multilevel model to the PISA 2012 data, this study investigated the factors which were linked to low achievement in mathematics. The multilevel analysis revealed that students' gender, immigration status, self-constructs about mathematics, pre-primary education attendance as well as individual and school mean socioeconomic status (SES) were statistically significant predictors of student low achievement. It was also found that school accounted for a large proportion of the differences between low achievers and non-low achievers, with the final model explaining a great part of these differences. By successfully addressing the research questions, this study has demonstrated evidence that could help educators and policy makers to tackle the massive problem of mathematics underperformance not only in Greece, but also in other countries with similar educational systems.