Health information systems (HIS) in emergencies face a double dilemma: the information necessary to understand and respond to humanitarian crises must be timely and detailed, whereas the circumstances of these crises makes it challenging to collect it. Building on the technical work of the Health Metrics Network on HIS and starting with a systemic definition of HIS in emergencies, this paper reviews the various data-collection platforms in these contexts, looking at their respective contributions to providing what humanitarian actors need to know to target their intervention to where the needs really are. Although reporting or sampling errors are unavoidable, it is important to identify them and acknowledge the limitations inherent in generalizing data that were collected in highly heterogeneous environments. To perform well in emergencies, HIS require integration and participation. In spite of notable efforts to coordinate data collection and dissemination practices among humanitarian agencies, it is noted that coordination on the ground depends on the strengths and presence of a lead agency, often WHO, and on the commitment of humanitarian agencies to investing resources in data production. Poorly integrated HIS generate fragmented, incomplete and often contradictory statistics, a situation that leads to a misuse of numbers with negative consequences on humanitarian interventions. As a means to avoid confusion regarding humanitarian health statistics, this paper stresses the importance of submitting statistics to a rigorous and coordinated auditing process prior to their publication. The audit trail should describe the various steps of the data production chains both technically and operationally, and indicate the limits and assumptions under which each number can be used. Finally emphasis is placed on the ethical obligation for humanitarian agencies to ensure that the necessary safeguards on data are in place to protect the confidentiality of victims and minority groups in politically sensitive contexts.