Each industrial field seems affected by revolutionary innovations that will push forward the quality and variety of disparate services: driverless trains running 24 hours/day, self-driving cars and interconnected objects are only some of the latest trends. The intrinsic common aspect among all these promising concepts is the increase in reliability needs: human intervention for restoring and rescuing is not tolerated anymore. Despite this consideration, electronic equipment datasheets of many vendors do measure reliability through fuzzy metrics, the MTBF (Mean Time Between Failures) being the first among those. Firstly, this paper aims to provide information on the origin of the MTBF and on the reason why it is nowadays the most common mean for referring to reliability. Furthermore, the reasons behind the fallacy of the MTBF are explained. Using MTBF as synthetic indicator of reliability does often represents the wrong choice. Indeed, high MTBF figures could be measured even if the actual reliability is low. However, MTBF use is often imposed by regulations or contracts, and some examples are reported in this paper related to the railway signaling domain. Finally, alternative structured ways of communicating about reliability, based on further concepts than the bare MTBF, are provided as conclusion of the article.