Social assistance has attracted renewed interest in countries where economic growth is doing too little on its own to address high levels of income inequality and poverty. Research into the material effects of such programmes is important but can be misleading if it fails to capture their full meaning to intended beneficiaries and other stakeholders. This is illustrated by a case study of Peru's 'glass of milk' programme, drawing on mostly qualitative evidence of its material, social and cultural dimensions. The programme is found to be well adapted to diverse contexts, but in a way that enhances its efficacy as a gendered instrument of mass patronage rather than as a means of addressing Peru's structural inequalities. The paper also suggests that a switch to conditional cash transfers is unlikely, on its own, to change this.