Binary population synthesis provides a direct way of studying the effects of different choices of binary evolution models and initial parameter distributions on present‐day compact binary merger populations, which can then be compared to empirical properties such as observed merger rates. Samples of zero‐age main sequence binaries to be evolved by such codes are typically generated from a universal initial mass function (IMF) and simple, uniform, distributions for orbital period P, mass ratio q, and eccentricity e. More recently, however, mounting observational evidence has suggested the non‐universality of the IMF and the existence of correlations between binary parameters. In this study, we implement a metallicity‐ and redshift‐dependent IMF alongside correlated distributions for P, q, and e to generate representative populations of binaries at varying redshifts, which are then evolved with the COMPAS code in order to study the variations in merger rates and overall population properties.