Building operational data is widely used for the development and validation of models and good design practices. This paper aims to analyze the potential impact of inadequate representation of energy-vulnerable groups in residential building datasets and to demonstrate the diversity of occupancy behavior by analyzing national surveys. Emphasis is placed on the need to closely examine how these data sources can introduce bias, particularly in relation to the socio-economic realities of occupants. Biases can lead to policies that fail to address the needs of vulnerable populations, leaving gaps in our understanding of the real energy challenges these groups face. Based on an exploratory literature review, a reflection on the challenges associated with data collection, analysis, and use in this specific context is made. Furthermore, the diversity of occupant behavior is investigated using a Time Use Survey. The data were subjected to descriptive analysis, time-series K-means clustering, and decision tree classification to identify the socio-demographic patterns of occupants based on their activity and occupancy behaviors. Using the silhouette score and the elbow method, five clusters were identified. Based on the gain ratio (entropy) and maximum depth of the wanted decision tree, occupants’ characteristics were associated to different clusters. However, due to the heterogeneity of occupants and the complexity of human behavior, accurately representing activity occupancy behavior is challenging. This underscores the significance of diversity in datasets for accurately simulating building energy consumption.