Galaxy-galaxy lensing is rapidly becoming one of the most promising means to accurately measure the average relation between galaxy properties and halo mass. In order to obtain a signal of sufficient signal-to-noise, one needs to stack many lens galaxies according to their property of interest, such as luminosity or stellar mass. Since such a stack consists of both central and satellite galaxies, which contribute very different lensing signals, the resulting shear measurements can be difficult to interpret. In the past, galaxy-galaxy lensing studies have either completely ignored this problem, have applied rough isolation criteria in an attempt to preferentially select `central' galaxies, or have tried to model the contribution of satellites explicitely. However, if one is able to {\it a priori} split the galaxy population in central and satellite galaxies, one can measure their lensing signals separately. This not only allows a much cleaner measurement of the relation between halo mass and their galaxy populations, but also allows a direct measurement of the sub-halo masses around satellite galaxies. In this paper, we use a realistic mock galaxy redshift survey to show that galaxy groups, properly selected from large galaxy surveys, can be used to accurately split the galaxy population in centrals and satellites. Stacking the resulting centrals according to their group mass, estimated from the total group luminosity, allows a remarkably accurate recovery of the masses and density profiles of their host haloes. In addition, stacking the corresponding satellite galaxies according to their projected distance from the group center yields a lensing signal that can be used to accurate measure the masses of both sub-haloes and host haloes. (Abridged) Comment: 16 pages, 10 figures, Accepted for publication in MNRAS