In this contribution, we study different error sources that affect the estimation of the propagation constant in broadband methods. For this purpose, a study on the state-of-the-art has been made. The behavior of different error sources was examined separately: errors in the vector network analyzer, in the line lengths, in the inhomogeneity of the substrate, and the repeatability of the launchers. It has been shown that, depending on the method used, the behavior of the random errors can be different. In this sense, an excellent strategy to reduce the errors is to increase the number of lines, choosing their length reasonably. On the other hand, it has been seen that both the errors in the length of the lines and in the inhomogeneities of the substrates will work as a bias in the solution of the estimated propagation constant. The bias will depend on the difference between lengths or substrates. Finally, it has been studied how the repeatability of transitions and connectors affects the methods. To improve this problem, it has been demonstrated that always using the same pair of connectors is an efficient way to reduce the uncertainty of the results.