The manufacturing of ceramic tiles is a very complex process, where a wide range of variables has an important influence in the final product. With regard to external appearance, the most of the production defects take place in the decoration station. Nevertheless, these defects are usually detected before baking, when the product is already finished, causing an important loss of effectives. Under this perspective, a mechanism able to detect the printing defects in the green parts would achieve 2 goals: on one hand, the reduction of the nonquality costs since green parts can be more easily recycled; and on the other hand, it would point out the real root cause of the failure by indicating, for instance, which ink is causing the problem. Color Prediction Models (CPM) are mathematical approaches which relate the microscopic distribution of the printed dots of a halftone image with the resulting macroscopic color. Its usage is extended in the field of the Graphic Arts, especially for calibration and fine image reproduction. However, they are barely known in the ceramic tile industry, a sector that keeps many similarities with the Graphic Arts one in terms of decorating. In this paper, we analyzed the prediction quality of 4 successful CPM (Murray-Davies, Yule-Nielsen, Neugebauer and Neugebauer Modified Yule-Nielsen) on 1 and 2 inks halftones printed on ceramic substrates, setting a comparison between them by means of linear and non-linear optimization techniques. Moreover, we proposed a value for the enigmatic “n” parameter on ceramic surfaces, which is said to model the optical dot gain phenomenon.