Improving energy efficiency in urban building stocks often relies on detailed building-code or survey inputs to define typologies and propose energy conservation measures (ECMs). In cities where codes are absent or surveys are costly, these data requirements hinder scalable retrofitting prioritization. We present a data-lean, replicable workflow combining inverse modelling and unsupervised clustering to segment buildings using only monthly electricity and outdoor temperature data. To overcome the dimensional instability of clustering raw parameters with mixed units, the method mathematically translates them into three dimensionally homogeneous performance metrics: annual baseline, cooling, and heating electricity consumption. K-means clustering is then applied to these unified metrics to identify thermally coherent building groups and screening-level ECM priorities. Demonstrated on 183 public office buildings in Santiago de Chile, baseline loads dominate total electricity use, accounting for approximately 78.6%, while cooling and heating contribute approximately 8.5% and 12.9%, respectively. Four robust clusters are obtained, revealing a small high-consumption group (11.5% of buildings) accounting for 38.8% of total electricity use, which should be prioritized for audits and targeted retrofits. Overall, this workflow enables scalable segmentation and retrofit prioritization in data-scarce urban contexts with minimal inputs.
