This article investigates the performance of Maximum Likelihood (ML)method to estimate the usability of e-commerce sites and compares theperformance between the software Bilog-MG ® and Excel ® in theestimation of these usability. For this, we used real data from a study onthe degree of usability of 361 e-commerce sites, which were applied 32items calibrated by the unidimensional logistic model of two parameters(MLU2) of Item Response Theory (IRT). The estimation process by MLwas developed in BILOG® and Excel® softwares. The results showedthat the ML method is flawed when there is a constant pattern of response,which may occur during application of the first items on the questionnaire.However, the method performs well when the pattern of responses is notconstant. Moreover, process performance prepared in Excel® was betterthan in conventional software BILOG-MG®. The parameters of the itemsalso influence the estimation of ML
