In this study, a metamodel for cooling energy consumption is presented. In constructing the metamodel, the assumption was made that cooling energy consumption is a polynomial function of the energy consumption of each individual building component. We studied the cooling energy consumption of an office under various climate conditions and with different thermal masses and uses. The metamodel was fitted from dynamic simulations by multiple regression analysis and the accuracy obtained was very close to that of dynamic simulation. Our study highlights the value in using the Design of Experiments method to fit the metamodel since the number of simulations could be reduced without a loss of accuracy. The metamodel coefficients were analysed to gain insight into the thermal behaviour of a building. We found that the energy consumption for cooling was a convex function of the individual energy consumptions of the building components. The results suggest that it would be beneficial to study third order and exponential terms. This work opens a path to a new way of analyzing building energy performance.