Risk factors that increase the likelihood of children experiencing poorer school functioning (academic and behavioural difficulties) have been widely researched. However, risk factors do not occur in isolation, but instead cluster together across different ecological domains (Oldfield, Humphrey, & Hebron, 2015). As they accumulate, the likelihood of negative outcomes is proposed to increase, often disproportionately. This is referred to as cumulative risk theory (Evans, Li, & Whipple, 2013). The Good Behaviour Game (GBG), a universal, classroom-based intervention, has the potential to successfully improve academic and behavioural outcomes for at-risk pupils (Flower et al., 2014). However, at-risk is often used as a proxy for highly aggressive in GBG research, thus failing to account for pupils exposed to multiple risks. Indeed, the impact of the GBG on pupils at varying levels of risk is poorly researched, and so any potential differential effects are unknown. Furthermore, research suggests that the way the GBG is implemented can have an impact on the achievement of its intended outcomes (Ialongo et al., 1999), although the evidence base is extremely limited. However, the research carried out to date on other school-based interventions does appear to suggest an interaction between high implementation and gains for pupils with higher levels of risk exposure (Abbott et al., 1998). The study aimed to contribute to the knowledge base on cumulative risk exposure and child development, assess the impact of the GBG on pupils at different levels of risk exposure, and determine the extent to which differential intervention gains varied among at-risk pupils as a function of implementation. A hybrid concurrent embedded mixed methods design was employed to allow for a deeper exploration of the perceived ways in which the GBG influenced at-risk pupils' outcomes. Quantitative data were collected from 3,084 pupils in 77 primary schools as part of a longitudinal randomised controlled trial of the GBG. Demographic and attainment data were collected from the National Pupil Database, while teachers completed surveys regarding pupils' disruptive behaviour. A reading test was administered to pupils to obtain attainment data. Qualitative data were collected from six self-selecting case study utilising semi-structured interviews and focus groups with teachers and pupils. Data analysis was conducted using multi-level modelling and thematic analysis. Analyses revealed six risk variables for disruptive behaviour, and seven risk variables for reading attainment, operating at pupil- and school-levels. A cumulative risk effect was identified, and a quadratic risk-outcome relationship was determined to be present, indicative of an exponential increase in negative outcomes as risk exposure increased. However, there were no statistically significant main or subgroup effects of the GBG on at-risk pupils' outcomes. Higher levels of implementation fidelity and quality were significantly associated with worsening disruptive behaviour scores for high-risk pupils. Qualitative analysis identified perceived differential gains for pupils exposed to six different risk variables; 10 proximal outcomes also emerged, along with five common GBG elements considered to be the mechanisms through which these outcomes were influenced. The implications of these findings and directions for future research are discussed.