In cosmography, cosmokinetics and cosmology, it is quite common to encounter physical quantities expanded as a Taylor series in the cosmological redshift z. Perhaps the most well-known exemplar of this phenomenon is the Hubble relation between distance and redshift. However, we now have considerable high-z data available; for instance, we have supernova data at least back to redshift z ≈ 1.75. This opens up the theoretical question as to whether or not the Hubble series (or more generally any series expansion based on the z-redshift) actually converges for large redshift. Based on a combination of mathematical and physical reasonings, we argue that the radius of convergence of any series expansion in z is less than or equal to 1, and that z-based expansions must break down for z > 1, corresponding to a universe less than half of its current size. Furthermore, we shall argue on theoretical grounds for the utility of an improved parametrization y = z/(1 + z). In terms of the y-redshift, we again argue that the radius of convergence of any series expansion in y is less than or equal to 1, so that y-based expansions are likely to be good all the way back to the big bang (y = 1), but that y-based expansions must break down for y < -1, now corresponding to a universe more than twice its current size.