This paper investigates children’s well‐being in Burundi during and after participation in a graduation model programme using a longitudinal mixed‐methods approach, which combined newly collected qualitative data with existing quantitative and qualitative data. The programme is found to improve children's well‐being, particularly in relation to housing, food security and education. Covariate shocks (such as drought and political uncertainty) undermine the sustainability of effects but greater knowledge, experience and prioritisation of children’s needs contribute to harnessing improved outcomes.