In this paper we explain some of the difficulties of providing forecasts of the financial benefits of early intervention programmes, focussing on those delivered during the early childhood period. We highlight the diversity of early intervention, and the complexity and multiplicity of outcomes. We summarise recent work at the Early Intervention Foundation to assess the evidence on the impacts of early intervention, recognising the diversity of approaches to delivery and the importance of innovation and local practice as well as of rigorous approaches to evaluating causal effects. We also describe new ways of assessing accurately the local fiscal costs of late intervention and consider the implications of this for addressing the well-established barriers to investment in prevention. Our analysis brings to the fore gaps in the evidence from which even the most rigorous ‘gold-standard’ research is not immune. These limitations prevent the production of an accurate and realistic cost-benefit ratio or net present value for the majority of programmes as delivered in practice. We suggest some paths towards a firmer foundation of evidence and a better alignment of evidence and policy.