We explore strategies to extract cosmological constraints from a joint analysis of cosmic shear, galaxy–galaxy lensing, galaxy clustering, cluster number counts and cluster weak lensing. We utilize the cosmolike software to simulate results from a Large Synoptic Survey Telescope (LSST) like data set, specifically, we (1) compare individual and joint analyses of the different probes, (2) vary the selection criteria for lens and source galaxies, (3) investigate the impact of blending, (4) investigate the impact of the assumed cosmological model in multiprobe covariances, (6) quantify information content as a function of scales and (7) explore the impact of intrinsic galaxy alignment in a multiprobe context. Our analyses account for all cross-correlations within and across probes and include the higher-order (non-Gaussian) terms in the multiprobe covariance matrix. We simultaneously model cosmological parameters and a variety of systematics, e.g. uncertainties arising from shear and photo-z calibration, cluster mass-observable relation, galaxy intrinsic alignment and galaxy bias (up to 54 parameters altogether). We highlight two results: first, increasing the number density of source galaxies by ∼30�per�cent, which corresponds to solving blending for LSST, only gains little information. Secondly, including small scales in clustering and galaxy–galaxy lensing, by utilizing halo occupation distribution models, can substantially boost cosmological constraining power.