This paper focuses on how to fulfill the objectivity and reliability goals, as well as the efficiency of the e-learning evaluation tools, and their integration in a blended evaluation system. In order to contribute to these goals, a new branch of statistics, i.e. "Statistical Learning", has been chosen to support this study. The proposed techniques can be very simply implemented with little knowledge of arithmetic and with the help of a standard spreadsheet. These techniques can allow us to get the whole picture of the evaluation procedure output, in order to systematically sort the main categories of the different students, and to easily identify the outliers for further assessment.