Major, trace, and rare earth element compositions of both tonalite–trondhjemite–granodiorite (TTG) and modern adakite-like magmas are typically used in conjunction with batch melting experiments and models to infer source rock composition, depth of melting and tectonic setting. However, batch melting does not capture the impact of melt segregation processes on magma geochemistry. We have used melting experiments in conjunction with numerical modelling to investigate the impact of melt segregation on TTG arc crust formation. Our melt segregation equilibrium (MSE) experiments are designed to reproduce the local changes in bulk composition that are predicted by the numerical model to occur as buoyant melt migrates upwards along grain boundaries and accumulates to form a magma that leaves the source region. The MSE experimental results show distinct differences in the melt and solid phase compositions and solid phase stability when compared with direct partial melting experiments. They yield a significant reduction in hornblende and plagioclase modal proportions at lower temperatures and partial melt compositions that are lower in An and have higher Mg-numbers. These results suggest that dynamic melt segregation and equilibrium processes may have a significant impact on modes, melt compositions and geochemical indicators such as Mg-numbers. Mantle wedge interaction may not be necessary to generate varying Mg-numbers in TTG and adakite magmas. Moreover, the use of batch melting models or experiments to interpret these geochemical signatures may not be appropriate.