The 2010 census round has highlighted an increasing number of countries that have included maternal mortality in their questionnaires. Censuses have been hailed as a key source of data for maternal mortality in the absence of complete vital registration information. In particular, given the small numbers, they might be the only source of information to estimate differential mortality. This paper explores the feasibility of using census data to analyse differentials (e.g. by education and age) of maternal mortality in low‐resource countries. By using data from Lesotho and Nicaragua, this study investigates the use of censuses to calculate differential maternal mortality. Methods used include indirect demographic techniques, smoothing functions, and sensitivity analysis to analyse maternal mortality by age, residence, and education. Results show that levels of adjustments vary greatly and that smoothing functions, although very useful for differential analysis, might not be powerful enough to avoid fluctuations because of small numbers. Despite the considerable investment over the last decade, results show that the census might not be a feasible tool to analyse differentials as the results suffer greatly from mortality coverage assumptions. Notwithstanding, censuses should still be considered as a key source of information and should be analysed in conjunction with other sources. Copyright © 2013 John Wiley & Sons, Ltd.