The difficulty of accurately assessing the ex ante impact of planned energy efficiency measures is a major barrier to the large-scale deployment of demand-side management (DSM) programs. The process of energy consumption in the urban built environment is dynamic in nature and comprised of the coupled interaction of multiple sub-systems. Furthermore, it usually displays significant correlation with weather and other perturbations. We propose an approach based on a novel linear/nonlinear black-box regression based model of the load driven by exogenous variables. Model estimation is performed using actual hourly load and weather data. The data is used to generate a "baseline" or "business-as-usual" energy consumption model. Modifying some physically significant model parameters, and comparing the model prediction to the baseline reveals the energy savings that can result from planned future citywide energy efficiency interventions. The baseline model can also be used to test, a priori, candidate intervention scenarios by varying some of the physically significant parameters. We test this approach in the city of Abu Dhabi, UAE. For this location, we have reliable recording of both load and weather variables at hourly resolution. The proposed procedure, often referred to as inverse load modeling, presents several novelties that result in exceptional accuracy. The final model's RMSE, does not exceed 1.6% of peak load while the MAPE is less than 2%. Finally, the descriptive nature of the model enables us to quantify the citywide impact of a DSM program comprised of several basic energy efficiency interventions. (C) 2017 Elsevier B.V. All rights reserved.