BackgroundChild mortality remains one of the major public-health problems in Tanzania. Delays in receiving and accessing adequate care contribute to these high rates. The literature on public health often focuses on the role of mothers in delaying treatment, suggesting that they contact the health system too late and that they prefer to treat their children at home, a perspective often echoed by health workers. Using the three-delay methodology, this study focus on the third phase of the model, exploring the delays experienced in receiving adequate care when mothers with a sick child contact a health-care facility. The overall objective is to analyse specific structural factors embedded in everyday practices at health facilities in a district in Tanzania which cause delays in the treatment of poor children and to discuss possible changes to institutions and social technologies.MethodsThe study is based on qualitative fieldwork, including in-depth interviews with sixteen mothers who have lost a child, case studies in which patients were followed through the health system, and observations of more than a hundred consultations at all three levels of the health-care system. Data analysis took the form of thematic analysis.ResultsFocusing on the third phase of the three-delay model, four main obstacles have been identified: confusions over payment, inadequate referral systems, the inefficient organization of health services and the culture of communication. These impediments strike the poorest segment of the mothers particularly hard. It is argued that these delaying factors function as ‘technologies of social exclusion’, as they are embedded in the everyday practices of the health facilities in systematic ways.ConclusionThe interviews, case studies and observations show that it is especially families with low social and cultural capital that experience delays after having contacted the health-care system. Reductions of the various types of uncertainty concerning payment, improved referral practices and improved communication between health staff and patients would reduce some of the delays within health facilities, which might feedback positively into the other two phases of delay.