Despite the widespread occurrence of humanitarian emergencies such as epidemics, earthquakes, droughts, floods and violent conflict and despite the significant financial resources devoted to humanitarian assistance, systematic learning from such interventions using rigorous theory-based impact evaluations is very rare. The objective of this paper is to examine the extent to which scientific impact evaluation methods can provide evidence to improve the effectiveness and efficiency in humanitarian action. This paper explores the methodological options and challenges associated with generating high quality evidence needed to answer key questions about the performance of humanitarian assistance, including whether assistance is reaching the right people, at the right time, is bringing about the desired changes in their lives (effectiveness) and is being delivered in the right doses, ways and with manageable costs (efficiency). With the help of six case studies and drawing on real-life examples from the small but growing academic literature, we demonstrate how impact evaluation methods can be used successfully and in an ethical manner to improve humanitarian assistance. A key lesson from our review is that it pays to be prepared. Much information is being collected these days about the risks of various emergencies unfolding, be they sudden onset or slow onset emergencies. Hence national actors and international donors can prepare for these events and for conducting meaningful impact evaluations. Given the overwhelming needs and the lack of funds, doing more with limited resources is a key challenge for humanitarian assistance and impact evaluation is one way of achieving this.