It is estimated that, on a global scale, neonatal deaths now contribute to nearly 40% of all mortality in children under the age of five. However, as most neonatal deaths occur at home in countries with no vital registration, estimates of mortality are normally based on large national surveys such as the Demographic and Household Surveys (DHS). However, these have major limitations which restrict their accuracy.This study explores the potential contribution of DHS data in improving knowledge of trends in neonatal mortality in developing countries. It analyses the potential causes and extent of both sampling and non-sampling errors using review of existing literature as well as original analysis. The study suggests that one of the greatest limitations for DHS data is the wide confidence intervals. This makes it impossible to use DHS data to detect relatively small changes over time. While analysis suggests that in most countries data on neonatal mortality conform to expected patterns, there is also some evidence of age-heaping and back-dating of deaths