When energy and carbon improve but NOx rises: A principal-component analysis of an anonymised retrofit sequence

Energy renovation assessments often report several environmental indicators whose apparent agreement can conceal an important exception. This study examines six cumulative model states from an anonymised nonresidential energy consultancy report using correlation-based principal-component analysis (PCA). Final energy, carbon dioxide and nitrogen-oxide inventories are normalised by a common reference area, standardised and decomposed into orthogonal descriptive axes. The full sequence reduces final energy by 42.470% and CO2 by 38.494%, while NOx increases by 33.813%. The first component explains 91.058% of standardised variation but combines positive energy and CO2 coefficients with a negative NOx coefficient. It therefore expresses an environmental trade-off rather than a uniformly desirable performance axis. A second component accounts for 8.942% and preserves the distinction between the envelope-only trajectory and the final system package. Excluding that endpoint changes the energy–NOx correlation from −0.8084 to approximately +0.9999. This is a scenario-set dependence result, not a statistical robustness claim. The analysis demonstrates how an apparently strong one-component representation can be unsuitable as a sustainability score. Source-model boundaries, shared accounting inputs and the distinction between inventory emissions and local exposure are retained throughout. The contribution is a reproducible diagnostic workflow for small consultancy scenario sets, not an estimate of population effects or proof of a specific equipment mechanism.