The aim of this study is to provide a better understanding of the heterogeneities in user-product relationships and their consequences regarding the household energy predictions. Several supervised and unsupervised machine learning algorithms have been applied to a comprehensive data set of residential energy consumptions collected by the US Energy Information Association. The results of the analyses reveal that, while the heterogeneities in the use-phase of consumer electronics could skew their environmental assessment results, they do not possess the same discriminant influences on the household electricity consumption compared to certain socio-demographics or usage of home appliances. Various cross-comparisons among product features and use-phase behaviors have been made and the most important predictors of the residential electricity consumption based on the data have been introduced. Product-level and user-level discussions on the findings have also been provided.