Traditional approaches to analyse educational outcomes frequently assume there is a linear relationship between non-attendance (via school absence and/or exclusions) and attainment, with policy implications usually focusing on the detrimental effect of non-attendance on attainment. This paper shows that implementing a non-linear approach to analyse the relationship between non-attendance and attainment can be useful to understand heterogeneous pathways of socioeconomic and health-related disadvantages into unfavourable educational outcomes. We followed a cohort of pupils in Scotland who started primary school in 2008/09 (age 4/5) and finished compulsory education in 2018/19 (age 15/16) and implemented a multilevel mixture regression model using linked population-level administrative data from Education Analytical Services (Scottish Government), Public Health Scotland (NHS) from the period between 2007-2019, and the 2011 Census (National Records Scotland). We found five unobserved groups of pupils that are distinct from each other in their patterns of attendance, exclusions and attainment, as well as their socioeconomic and health status, and level of needs. Furthermore, we found evidence that a good level of attainment could be theoretically possible without an equally good level of attendance, since the two groups with the highest attainment have markedly different profiles of attendance and exclusions, as well as distinct patterns of disadvantage in terms of attainment. We conclude that addressing the attainment gap requires a multi-layered approach that considers heterogeneous profiles of non-attendance. Policy and school guidance must also recognise the complexity of upstream and downstream factors that interact and condition children and young people’s lives, experiences and outcomes.