This paper presents the advantages of using Weibull distribution to describe reliability figures of an electronic railway signaling system, respect to the commonly used exponential distribution. After presenting the context of reliability estimations within the railway domain, this work introduces the approach to use field-return data to build up reliability parameters instead of predictive methodologies, and it is applied to an existing electronic signaling system. A comparison between the two methodologies is also provided, as well as the introduction of further reliability indicators. A model to improve preventive maintenance tasks defining the optimum time interval is then presented and an example is provided. Some suggestions to improve the process of collecting field-return data are presented impacting both the after-sales logbook and the design of the circuit boards.