Natural and man-made emergencies are regular occurrences in the Greater Horn of Africa region. The underlying impoverishment of whole populations is increasing, making it more difficult to distinguish between humanitarian crises triggered by shocks and those resulting from chronic poverty. Shocks and hazards can no longer be seen as one-off events that trigger a one-time response. In countries that are both poor and exposed to frequent episodes of debilitating drought or chronic conflict, information needs tend to be different from the straightforward early warning/commodity accounting models of information systems that have proven reliable in past emergencies. This paper describes the interdependent components of a humanitarian information system appropriate for this kind of complex environment, noting the analytical links between the components and operational links to programme and policy. By examining a series of case studies from the Greater Horn region, the paper demonstrates that systems lacking one or more of these components will fail to provide adequate information--and thus incur humanitarian costs. While information always comes with a cost, the price of poor information--or none--is higher. And in situations of chronic vulnerability, in which development interventions are likely to be interspersed with both safety nets and emergency interventions on a recurrent basis, investment in improved information is a good investment from both a humanitarian and a financial viewpoint.