Galaxy-galaxy lensing measures the mean excess surface density DS(r) around a sample of lensing galaxies. We develop a method for combining DS(r) with the galaxy correlation function xi_gg(r) to constrain Omega_m and sigma_8, going beyond the linear bias model to reach the level of accuracy demanded by current and future measurements. We adopt the halo occupation distribution (HOD) framework, and we test its applicability to this problem by examining the effects of replacing satellite galaxies in the halos of an SPH simulation with randomly selected dark matter particles from the same halos. The difference between dark matter and satellite galaxy radial profiles has a ~10% effect on DS(r) at r<1 Mpc/h. However, if radial profiles are matched, the remaining impact of individual subhalos around satellite galaxies and environmental dependence of the HOD at fixed halo mass is <5% in DS(r) for 0.12 Mpc/h, but it fails at the 30-50% level on smaller scales. The scaling of DS(r) ~ Omega_m^a(r) sigma_8^b(r) approaches the linear bias expectation a=b=1 at r>10 Mpc/h, but a(r) and b(r) vary from 0.8 to 1.6 at smaller r. We calculate a fiducial DS(r) and scaling indices a(r) and b(r) for two SDSS galaxy samples; galaxy-galaxy lensing measurements for these samples can be combined with our predictions to constrain Omega_m and sigma_8.

18 pages, 10 figures, accepted for publication in The Astrophysical Journal