It has been previously demonstrated in Li et al (2016 Nucl. Fusion56 126007) that the optimum upper/lower coil phase shift ΔΦopt for alignment of RMP coils for ELM mitigation depends sensitively on q95, and other equilibrium plasma parameters. Therefore, ΔΦopt is expected to vary widely during the current ramp of ITER plasmas, with negative implications for ELM mitigation during this period. A previously derived and numerically benchmarked parametrisation of the coil phase for optimal ELM mitigation on ASDEX Upgrade (Ryan et al 2017 Plasma Phys. Control. Fusion59 024005) is validated against experimental measurements of ΔΦopt, made by observing the changes to the ELM frequency as the coil phase is scanned. It is shown that the parametrisation may predict the optimal coil phase to within 32° of the experimental measurement for n = 2 applied perturbations. It is explained that this agreement is sufficient to ensure that the ELM mitigation is not compromised by poor coil alignment. It is also found that the phase which maximises ELM mitigation is shifted from the phase which maximizes density pump-out, in contrast to theoretical expectations that ELM mitigation and density pump out have the same ΔΦul dependence. A time lag between the ELM frequency response and density response to the RMP is suggested as the cause. The method for numerically deriving the parametrisation is repeated for the ITER coil set, using the baseline scenario as a reference equilibrium, and the parametrisation coefficients given for future use in a feedback coil alignment system. The relative merits of square or sinusoidal toroidal current waveforms for ELM mitigation are briefly discussed.