Modal parameters are estimated from vibration data, thus they are inherently afflicted with statistical uncertainties due to the unknown ambient excitation and measurement noise. While the point estimates of the modal parameters can be obtained with several system identification methods, only few of them also provide the associated uncertainties. The quantification of these uncertainties is important for many applications, since they are a means to assess the precision of the estimates, and to evaluate if changes between different datasets are statistically significant or not. As such, they are an added value in modern modal analysis practice and used in applications to e.g., damage detection and localization, reliability analysis, modal tracking and model calibration. For subspace-based system identification, efficient methods for uncertainty quantification have been developed for the last 15 years, yielding reliable estimates of the uncertainties at reasonable computational cost. They cover a wide range of subspace methods and their application areas. In this paper, an overview of the developments is given and the importance of the knowledge of the uncertainties is illustrated.