In this study non-invasive neutron scattering techniques are used on soft condensed matter, probing colloidal length scales. Neutrons penetrate deeply into matter and have a different interaction with hydrogen and deuterium, allowing for tunable contrast using light and heavy water as solvents. The mesoscopic structure of materials is determined by measuring the elastic scattering of neutrons over small angles. Spin-echo small angle neutron scattering (SESANS) and the reciprocal-space equivalent ultra small angle neutron scattering (USANS) have been used to investigate the structure of colloidal suspensions and gels by measuring the projections of the density-density correlation function and of the scattering function. A hollow sphere model is developed and used to investigate liposomes and E. coli bacteria. The sizes of liposomes and E. coli and their hollow sphere nature were confirmed. Particle size and size distribution have been measured for milk, and the change in typical length scale during gelation into yoghurt was measured kinetically. 3D Monte Carlo simulations of colloidal aggregation have been performed using a varying reactivity, ranging from reaction limited (RLCA) to diffusion limited cluster-cluster aggregation (DLCA), to study the effect on structure and formation. A comparison is made between simulated structures and SESANS measurements by calculating the density correlation function. A relaxation time is introduced into the simulated reactivity to control aggregation speed. The increase of typical sizes during gelation at low reactivity was consistently observed in measurements and simulations, but longest length scales could not be accurately simulated. A transition from RLCA to DLCA occurs if the typical reaction time is below typical monomer diffusion time. The increase in coordination number becomes larger when aggregation remains reaction limited. The optimal sample thickness for a SESANS measurement is derived as a function of scattering cross-section and linear attenuation coefficient. The optimal sample thickness is at about 0.8 scattering events for neutron transparent samples, whereas it is the 1/e-length when samples are neutron opaque.