Purpose: The goal of this article is to find out if an old Life Cycle Assessment (LCA) study remains valid or not after a period of time. To answer this we re-perform an LCA of a bus stop in the city of Barcelona that was performed about 20 years ago. Methods: The LCA of a bus stop, performed and published in 1998, is re-performed and its results compared with those of the original study, keeping the same scope of the original system. The software used by the original study was SimaPro; the data came from IVAM LCA data, BUWAL 250, IDEMAT 96 and PRe4 databases; and CML 1992 impact assessment methodology was used. The new study used thinkstep GaBi6 software; GaBi6 Professional + extension databases; and CML 2001 methodology. The assessment includes an analysis of the key influencing factors that cause the discrepancies, such as models and databases. Moreover, a specific focus on evolution of the methodologies and its influence on the results is described. A 30% of difference between results is accepted as the threshold value to be able to state that the results differ. Results and discussion: The overall results obtained in the two studies are quite similar or, at least, comparable. However, when analysing and comparing the systems disaggregated stage by stage, higher differences in each impact category are found. Therefore, the lower discrepancy at system level may be due to coincidence or compensation. The main causes of discrepancy have been found to be: (i) the update of the assessment methodologies and characterization factors, (ii) the improvement of the databases, and (iii) the change in the techno-sphere and the improvement of the environmental policies. Conclusions and recommendations: After 15–20 years, LCA results cannot be considered reliable. Results can, however, be used as an indication for the expected order of magnitude of the impacts and for the relative importance of the processes in the different life cycle stages. The comparison is made through one case study only; therefore, it can barely be used as a generalization, neither for the difference in results nor for the sources of discrepancy. Nevertheless, this type of analysis can be considered a first step in quantifying the longevity of LCA results.