Teacher evaluation in higher education has been a controversial topic and its use is frequent in this field for taking decision. In most of cases, the evaluation is based on the assumption that students learn more from highly qualified teachers and its validity is based on the fact that students observe the teachers performances in the classroom. Therefore, the students should be the evaluators under the assumption that they will respond sincerely when asked about the teacher performance. However, many studies question about the methodologies used for getting such measurements, in general, because the averages by categorical responses have little statistical sense. In this document, the measurement of teaching performance is proposed through a multi-faceted TRI model that takes into account parameters associated with the severity of the evaluator and an additional parameter that is related to the effect of the evaluated course. For the model estimation, Bayesian inference techniques were used due to the large number of parameters to be estimated. The proposal was applied to a data set obtained from a survey of perception of teaching performance conducted by the Faculty of Science of the National University of Colombia to students of the same faculty. The proposed model was evaluated with goodness of fit statistics and compared with a model that did not include the additional parameter. The results obtained indicate that the proposed model had a better fit and the selection criteria indicate that the proposed model was more appropriate for the measurement. Additionally, using auxiliary variables, differences were observed between the qualifications given by female students and male students and the qualifications by postgraduate students and undergraduate students, in addition, there were differences between the estimated average difficulties of the courses according to the study area that offers.