We analyse deep g′ and r′ band data of 97 galaxy clusters imaged with MegaCam on the Canada–France–Hawaii telescope. We compute the number of luminous (giant) and faint (dwarf) galaxies using criteria based on the definitions of de Lucia et al. Due to excellent image quality and uniformity of the data and analysis, we probe the giant-to-dwarf ratio (GDR) out to z ∼ 0.55. With X-ray temperature (Tx) information for the majority of our clusters, we constrain, for the first time, the Tx-corrected giant and dwarf evolution separately. Our measurements support an evolving GDR over the redshift range 0.05 ≤ z ≤ 0.55. We show that modifying the (g′ − r′), and K-correction used to define dwarf and giant selection does not alter the conclusion regarding the presence of evolution. We parametrize the GDR evolution using a linear function of redshift () with a best-fitting slope of α = 0.88 ± 0.15 and normalization β = 0.44 ± 0.03. Contrary to claims of a large intrinsic scatter, we find that the GDR data can be fully accounted for using observational errors alone. Consistently, we find no evidence for a correlation between GDR and cluster mass (via Tx or weak lensing). Finally, the data suggest that the evolution of the GDR at z < 0.2 is driven primarily by dry merging of the massive giant galaxies, which when considered with previous results at higher redshift, suggests a change in the dominant mechanism that mediates the GDR.