Driven by gravity, galaxies continuously grow through accretion of smaller systems. Stellar streams are nice illustrations of this hierarchical build-up, but the accreted stars quickly disperse. I will present advanced dynamical models that can convert the observed positions and velocities of stars to phase-space quantities like energy and angular momentum which remain largely conserved. In addition, these models can include the observed ages and chemical properties of stars which are also conserved. The resulting population-dynamical models allow us then to uncover even those accretion events which are now fully dispersed. At the same time, these models also accurately constrain the total mass distribution, including a central black hole and dark matter halo.
I will illustrate how these models make optimally use of observations to unveil the dark side and colour past of galaxies: from accurate measurements of their dark halo, to unveiling the formation history of their disks, to uncovering ancient massive mergers and accreted satellite galaxies. By the end, I aim to have demonstrated that these models provide a unique bridge between the studies of resolved stars in the Milky Way and integrated-light of high(er)-redshift galaxies. Together with direct coupling to state-of-the-art galaxy formation simulations, these population-dynamical models enable us to uncover the hierarchical build-up of galaxies in a cosmological context.