Nutrition surveillance is a rather recent application of epidemiological surveillance to nutrition. It covers not only the morbidity and mortality of nutritional disorders, but is at least as much interested in their major determinants. After some fifteen years of trial and error, nutrition surveillance has today found its way: its concepts, methods and uses are well established by now. After retrieving experiences conducted since the seventies and the eighties, and on the basis of their own experience, the authors have come to identify a few major features of nutrition surveillance: predominance of the supply of information over the collection of data; the early identification of users; the importance of causal analysis as a departure point for conceiving a surveillance activity; the highly selective choice of data to be collected; a permanent concern for sustainability; the participation of all actors (or "stakeholders") and, from a technical viewpoint, the place of qualitative data collection techniques (borrowed from social anthropology) to complement quantitative data. The authors summarize the major steps followed in the design and implementation of a nutrition surveillance activity, emphasizing the benefits of a careful and in-depth preparation during the so-called "pre-surveillance" phase.