The building sector has a high potential to reduce energy consumption. Achieving this depends on household’s choices, which are known to be highly heterogeneous. Agent-based models are tools used to describe energy choices, but require data demanding calibration. Here we combine a novel, cross-country European household-level survey -including socio17 demographic characteristics, energy-saving habits, energy-saving investments, and metered household electricity consumption- with a global agent based energy choice model. Cluster analysis reveals that households who demand and consume energy in very similar ways cannot easily be mapped to standardly used socio-demographic classes or attitudes. However, the data also shows interesting patterns both between and within the clusters. Most noticeably, income, consistently, has the largest effect on demand, dwelling efficiency and energy-saving investments. Dwelling improvement potential also incentivizes energy efficiency investments. We use this cluster analysis to calibrate agents of the residential sector of an agent based model, including also the within cluster variations and uncertainty. Including these various sources of heterogeneity affects the timing and speed of the transition, two key indicators in the context of climate change mitigation. The results reinforce the need for grounding agent28 based models in real data, to take real advantage of their capabilities and contribute to a better understanding of energy transition dynamics.