Our research was built upon two studies by Noyoo (2016) and Barnett et al. (2020), even though the former was in the South African context while the latter was in the American context. There was a possibility that the companies’ social impact of their corporate social investment (CSI) initiatives was overstated. Consequently, it contributes more to the status quo of inequality, poverty and unemployment (unintended social impact) than alleviating it (intended social impact). To solve this problem, this paper answered the question, “What is the relationship between social impact (dependent variable) and CSI (independent variable), if any?” A self-administered survey questionnaire was used to collect the quantitative data. Using panel data, we did quota sampling. The final dataset was 427 completed responses. The least squares regression analysis was guided by critical systems heuristics to analyse the data, focusing on the legitimacy source of influence. There were violations of the assumptions required for discriminant analysis. Looking at the unique individual contributions of the predictors, relevance (β = 0.032, t = 0.642, p = 0.521), coherence (β = 0.007, t = 0.127, p = 0.899), effectiveness (β = 0.203, t = 3.689, p = 0.000), efficiency (β = 0.343, t = 6.912, p = 0.000) and sustainability (β = 0.358, t = 7.779, p = 0.000) positively predicted social impact of CSI in South Africa. We concluded that our proposed model justified the source of legitimacy and positively influenced social impact. Therefore, these are suitable constructs to consider when developing CSI initiatives.