Business analytics has become the latest fad in management practice. Carried out by largely positivist statistical analysis, anecdotal evidence claims great success for these practices. This paper takes a critical realist approach to a philosophical analysis of these practices. Reviewing CR ontology and epistemology, it applies those to the practice of business analytics, showing that where BA success occurs it is because the analysis has encountered relatively unimpeded actions by relatively enduring structures of causal mechanisms. It may fail in those areas undergoing rapid structural change, or where there exists a confusing welter of mechanisms. The danger in this approach from a CR perspective is that other causal mechanisms may intrude, or the structure of the mechanisms may change causing the models built to fail. This paper argues that the practice of BA may be improved by performing retroductive analysis to identify these structures so that anticipatory action can be taken to avoid failure of the models.