We analyse the use of the concept of household in sample surveys, with evidence drawn from a review of survey definitions, a series of in-depth interviews with data producers and users, and a systematic study of recent literature. We consider the place of the concept within the discipline of demography, and demonstrate how its definition and use interact with cultural values and core concepts integral to the discipline. Focusing on Tanzania as a case study, we examine the diversity of factors that influence the construction of household-level data from cross-sectional household surveys. Throughout the survey process, contrasting interpretations of the meaning of household and different motivations for using specific definitions of the term interact. This generates data and outputs with potential for undercounting, bias, and misrepresentations, with adverse effects on the quality of data used for monitoring development indicators. Some ways of improving data collection on households are proposed.