The Cox-Strack method is commonly applied to assess the contact resistivity between a metal and a semiconductor since the 1960s, while the underlying assumptions have not yet been rigorously assessed. In this article, a combination of finite-element modeling and mathematical analysis is used to investigate the accuracy of the conventional Cox-Strack equation for generic metal-semiconductor junctions. A systematic error in the spreading resistance equation is quantified, and alternative, more accurate equations are presented. Furthermore, it is shown that commonly used experimental configurations can lead to highly overestimated contact resistivities. Guidelines are formulated for accurate extraction of the contact resistivity from the Cox-Strack measurements.