This study attempts to identify the key factors that will make a tsunami warning system most effective, to develop a framework in which results of natural science and engineering research can be effectively integrated into coastal natural hazard planning, and to develop a numerical example that illustrates how benefit-cost analysis may be used to assess early warning systems. Results of the study suggest that while the science of tsunami wave propagation and inundation is relatively advanced, our knowledge on the relationships between tsunami generation and undersea earthquakes, volcanic eruptions, and landslides remains poor, resulting in significant uncertainties in tsunami forecasting. Probabilities of damaging tsunamis for many coastal regions are still unknown, making tsunami risk assessment and management difficult. Thus it is essential to develop new techniques to identify paleo-tsunami events and to compile and develop size and frequency information on historical tsunamis for different locations. An effective tsunami early warning system must include not only the ocean technologies for accurately detecting an emerging tsunami, but also a civil communication system through which the population can be timely warned by the local government and other sources. Since minimizing the evacuation time is a key factor to make a warning system effective, adequate pre-event education and preparation of the population must be a critical component of the system. Results of a numerical example of the South Pacific region suggest that investments in a tsunami warning system in the region may lead to significant economic benefits.