The threshold at which state social work services respond to children who may have been maltreated has been the subject of debate in the UK and elsewhere over a number of years. This paper presents a conceptual model that helps to explain how threshold decisions are made, based on a review of research in the UK and elsewhere. A range of factors were found to affect thresholds, including the nature of the welfare concerns for the child, the policy and organisational circumstances, the role of collaborative practice amongst a range of professionals, and the decision making of front line social workers, teams and managers. A central argument is that the technical–rational model of thresholds, which is commonplace in the UK, is insufficient and the concept of ‘thresholds’ too limiting. In their place, we adopt a naturalistic decision-making approach, arguing that ‘threshold decisions’ are mediated through various sense-making strategies at local level. These strategies may appear as shortcuts to thinking, arising from a pressurised working environment. We argue, however, that they are consequential features of the context within which children's social work services operate and that taking them into account allows for a more nuanced understanding of how thresholds are managed.