In administrative sciences, one of the main challenge is to choose the right items for a measurement scale. The purpose of this article is to provide marketing researchers a detailed description of item response theory (IRT) for rating scales. It details the different stages of IRT using the graded response model (GRM) on two rating scales (need for cognition and satisfaction). IRT approach offers a notable advantage due to its ability to precisely assess the quality and the contribution of each of the items to the latent trait. GRM could be used either as a complement or a substitute to the confirmatory factor analysis (CFA), especially for non-normal distributed scales.