Urban Heat Island (UHI) effect is a well-documented challenge in urban areas. Yet, conventional analyses at large, urban scales often obscure microclimatic and physical diversity at the building level, thereby limiting actionable insights. This study introduces a building-centric UHI assessment framework using individual buildings as fundamental spatial units. We call it Urban Building Heat Island, or UBHI. Its proposal is based on a study utilising high-resolution thermal and physical data of building features for 9,000 and 8,000 buildings in Darwin and Sydney respectively.The foremost challenge in performing a building-centric UHI calculation is to accurately determine the representation of a building on a grid. We compare grid sizes of 500m, 100m, 50m with building footprints and analyse their impact on building representation and the overall accuracy of the UHI prediction. The results show that while grid-based models achieve higher accuracy as grid size decreases, the building representation gradually declines. On the other hand, building-level models combine high accuracy (R² = 0.919–0.924) with full (1.0) building representation, supporting both model fidelity and actionable intervention possibilities. We further investigate the importance of influencing variables and show that vegetation coverage (Normalised Difference Vegetation Index, NDVI, 22%) and proximity to highways (22%) as the primary regulators. Other variables include nearest-building distance, building area, and density, together accounting for half the model variance at building scale. A simple scenario-based analysis indicates that targeted enhancements in NDVI can reduce building-level UHI intensity by up to 31.8%. We also discuss the transferable aspects of this model to other cities while setting the stage for UBHI assessment and mitigation at reduced granularity.