Structural health monitoring primarily requires the identification of the modal parameters of structures (frequency, damping ratio and modal shape). Structural identification methods based on Ambiental vibrations are becoming more and more popular. In fact, the techniques of Operational Modal Analysis [1] allows to obtain all the modal information of structures performing easy and unexpensive tests. The challenge in these methods is to distinguish between real and spurious or mathematical modes. For the purposes of making this distinction, the techniques of Stochastic Subspace Identification [2] have been used. The issues related to this method, however, are multiple; the success of the method depends on choices and criteria decided by the user [3], often on the basis of experience, or on trial and error procedures that are very costly computationally. The goal of this work is to investigate, through experimental data, how the choice of different parameters and criteria affect the results of the tests with the objective of developing an automated methodology able to avoid the user’s intervention. Computationally efficient methods for data analysis and comparison need to be used to this aim.