How household-level data from censuses and surveys are analyzed to study household structure is an issue that has received little attention. The present study proposes a new methodological approach to address this gap. Specifically, we introduce the idea of the household configuration as a mathematical representation of observations from the household roster that uses the tools of sequence analysis to study relationships between household members. This "household configuration approach" is statistically efficient, captures the heterogeneity of family forms in a population, and is computationally simple. An application to Canadian census data for Indigenous and non-Indigenous peoples shows that our approach can yield interesting insights into household structure, otherwise not readily obtained.