In reservoir characterization, rock-physics models provide the link between seismic observables (density, compressional and shear wave speeds) and reservoir parameters, such as porosity, lithology and fluid saturation. However, the accuracy of these predictions is rarely explored. In fact, the validation of a model representing a dataset is often limited to the analysis of a cross-plot of two arbitrary magnitudes. The objective of this paper is to improve the calibration procedure through a quantitative assessment of the reservoir property predictions using various rock-physics models. The analysis is based on an inverse rock-physics modelling that organizes the rock-physics transforms into constraint data so that the seismic variables are direct functions of the reservoir parameters. It is revealed that the predictions of reservoir quality can assist in the diagnosis of the rock microstructure itself, such as the location of clay particles in clay-rich sediments. In addition, we found that a quantitative analysis is the only way to evaluate accurately the performance of various models when studying heterogeneous datasets.