He argued that if we found such stars in another galaxy, we could assume that they had the same luminosity - and so calculate the distance to that galaxy.  If we could do this for a number of stars in the same galaxy, and our calculations always gave the same distance, we could be fairly confident of our estimate.