It is known that standard dipole showers yield incorrect subleading-colour contributions already at leading-logarithmic level for certain observables. In this talk, I will present two new, simple prescriptions to palliate this issue. I will show how the resulting colour-corrected matrix elements are cured at fixed order in the appropriate limits. I will then demonstrate how the two algorithms allow us to achieve full-colour next-to-leading logarithmic accuracy for various observables (global observables and multiplicities). I will conclude by commenting on the comparison to full-colour NLL-resummed results for non-global observables.