The Geiger-Nuttall (GN) law relates the partial $α$-decay half-life with the energy of the escaping $α$ particle and contains for every isotopic chain two experimentally determined coefficients. The expression is supported by several phenomenological approaches, however its coefficients lack a fully microscopic basis. In this paper we will show that: 1) the empirical coefficients that appear in the GN law have a deep physical meaning and 2) the GN law is successful within the restricted experimental data sets available so far, but is not valid in general. We will show that, when the dependence of logarithm values of the $α$ formation probability on the neutron number is not linear or constant, the GN law is broken. For the $α$ decay of neutron-deficient nucleus $^{186}$Po, the difference between the experimental half-life and that predicted by the GN Law is as large as one order of magnitude. ; Comment: 4 pages, 5 figures, to appear in Phys. Lett. B