Natural disasters have many consequences in terms of human lives, material, economic, and/or environmental damages. Among preventive and mitigation measures, it is recognized that early warning systems (EWS) are an effective and essential tool to minimize damages caused by natural disasters. Using the experience of CEMADEN (National Early Warning and Monitoring Centre of Natural Disasters) in Brazil, this paper aims to investigate, through a systems approach, what factors may interfere with the effectiveness of EWSs. A case study was developed based on interviews with experts from CEMADEN. Those interviews generated cognitive maps that translated the perceptions of the experts and were used to structure the problem and to support the construction of a systemic model. The model allowed the analysis of the EWS, identifying behaviors, as reinforcement and balancing loops, not always intuitive, to support better management and planning decisions to improve the system effectiveness.