Both customized and model-based benchmarks can be powerful tools, but considering the degree of detail that is required for data collection, they are not very feasible for this study. Therefore, it was aimed at establishing a benchmark based on normalized ranking. The first step was to collect a list of variables as potential energy use drivers. A national survey was conducted targeting 103 gazetted hotels, which include all the quality hotels in Singapore. Complete datasets were received from 29 of them (5 3-star, 13 4-star and 11 5-star). Data collected includes energy consumption of different fuels, building physical and operational characteristics and other related information. Regression techniques are used to identify the determinants of energy use intensities in hotels. It enables the resulting benchmark to account for the inflexible determinants and hence make comparisons fairer. The statistical model can be established directly if the independent variables to be included in it are already known. But this is usually not the case. Hence, the list of potential energy use drivers should be examined for their significance and correlation with the dependent variable as well as with the other independent variables. Stepwise linear regression is a technique that can be used to fulfill this task. As a result, only the statistically significant variables (X1, y, Xn) are left in the regression model, while the others are discarded.