Testing of ultra-soft tissue, e.g., brain tissue poses several challenges. The fragility, ephemeral nature, adhesiveness, and ultra-softness of biological tissue impede sample extraction for material testing. Complex sample geometries to elicit complex stress and strain states do not seem viable due to the aforementioned challenges. Therefore, it is generally aimed for geometric primitives. Common sample geometries are cuboids and cylinders. Both geometries are suitable for compression and tension, the former find use for simple shear whereas the latter offer themselves for torsion. Sample extraction, in all cases known to us, is carried out manually, either by free-hand cutting using medical devices, such as scalpels, biopsy punches, and trimming blades, or by using custom-made devices to fix the tissue and guide the blade. Final sample geometries differ mildly to severely from the intended geometric primitives. Nevertheless, those deviations are usually not considered in the parameter identification following the testing, i.e., idealized geometries are used in the simulations. This introduces an error in the identified parameters, which has not been quantified yet. The sample extraction and testing procedure we follow is presented in detail in Reiter et al. [1]. We extract cylindrical samples from brain tissue and test them in a rheometer in tension, compression, and torsion. During the testing procedure, we capture the sample geometry on photos. This allows us to extract the geometry from the images and reconstruct the actual sample geometry. Using finite element simulations, we determine the effect of using the actual geometry to the idealized geometry on the resulting force-displacement response and the material parameters identified from the experimental data. This will aid the design of future experiments, the interpretation of existing data and the improvement of parameter identification procedures resulting in more accurate material parameters for ultra-soft tissues.