Assessing the sensitivity of material-intensity-based building stock estimates to design parameters

Accurate estimates of building material stocks (BMS) are critical for advancing circular economy strategies. The widely used material intensity (MI) method applies fixed values per square meter and therefore overlooks variations introduced by architectural design. To assess the sensitivity of MI-based estimates to design parameters—a question difficult to address with real-world data—we developed an automated workflow that combines parametric modeling and deep learning to generate synthetic building datasets and calculate their component-based material stocks. We generated 48,600 residential buildings, all with a constant gross floor area (GFA) of 2,400 m² m 2 , while systematically varying footprint shape, corridor type, number of units, room layout, number of floors, and structural system configuration. Statistical analysis reveals that within this controlled design space, even under identical GFA, total material stock ranges from 758 to 2,888 m³ m 3 . Multiple regression analysis shows that number of floors and structural system configuration exert the strongest influence, followed by footprint shape and number of units, while corridor type and room layout contribute only marginally. These results demonstrate how component-based approaches can be used to assess design-induced variability in MI-based estimates. The findings also provide guidance on when MI values may require design-specific refinement by considering dominant design parameters.