This technological paper identifies the educational indicators that best differentiate school performance obtained by students from the 5 th grade of primary education in state schools of Espírito Santo in the Math test and Portuguese Language test in the Saeb 2017 through logistic regression. The sample consisted of 364 schools of the public education network of Espírito Santo (Brazil), classified as the best performance and worst performance. The results showed that the Socioeconomic level of students, the Index of teacher regularity and the Indicator of school management complexity formed an optimal set of variables to discriminate the performance of schools in the Portuguese Language test. This same set of variables, and the variable Students per class, proved to be more relevant to evaluate the performance of schools in the Math test. These results can contribute to educational public policy decisions, reorganizing their investments based on priority educational indicators. The findings reinforce the importance of the school environment in school performance.