Bearings are omnipresent in mechatronic systems such as motors and drive trains, as they are the interface between rolling and stationary components. Bearings in critical machinery are continuously monitored to detect faults in an early stage and avoid system damage and downtime. This condition monitoring is typically done with electrical accelerometers or microphone arrays. Optical fiber Bragg grating-based (FBG) sensors are ideal candidates to overcome the shortcomings of their electronic counterparts as they are thin and lightweight, immune to EMI and crosstalk, and have multiplexing capabilities. This implies that a multitude of sensors can be implemented directly on the bearing.In this work, we report on the quantitative investigation of the repeatability and reliability of FBG-based bearing fault detection. We document the influence of the installation of the fiber on the bearing on the acquired strain signals, and the impact thereof on the selection of an adequate damage indicator. To the best of our knowledge, this is the first time that such an extensive study has been carried out in view of quantifying the repeatability and reliability of FBG-based fault detection in bearings.