Focused ion beam nanotomography (FIB-nt) is a novel method for high resolution three-dimensional (3D) imaging. In this investigation we assess the methodological parameters related to image acquisition and data processing that are critical for obtaining reproducible microstructural results from granular materials and from complex suspensions. For this purpose three case studies are performed: (1) The precision of FIB-nt is evaluated by analyzing a reference sample with nanospheres. Due to the implementation of an automated correction procedure, drift phenomena can be removed largely from the FIB data. However, at high magnifications remaining drift components can induce problems for 3D-shape reconstructions. (2) Correct object recognition from densely packed microstructures requires specific algorithms for splitting of agglomerated particles. To establish quantitative criteria for the correct degree of splitting, a parametric study with dry portland cement is performed. It is shown that splitting with a k-value of 0.6 leads to accurate results. (3) Finally, the reproducibility of the entire cryo-FIB analysis is investigated for high pressure frozen cement suspensions. Reproducible analyses can be obtained if the magnification is adapted to the particle size. At low magnifications the small particles and their surface area are underestimated. At high magnifications representativity is questioned because local inhomogeneities can become dominant.