Normal (NPI) and/or combined Pressure-shear (PSPI) plate impact experiments are often employed for studying the macroscopic response of polycrystalline metals under dynamic loading. However, even for high symmetry metals such as Aluminum, their single-crystalline response is inherently anisotropic leading to significant uncertainty in dynamic measurements involving polycrystalline samples and finite specimen size. In particular, the role of few grains and finite thickness on the measurement uncertainty in PSPI experiments is presently unaddressed in the literature and is the main motivation for this research. In the present study, we perform direct numerical simulations (DNS) of statistically representative microstructures composed of randomly distributed and orientated equiaxed grains within macroscale polycrystalline specimens subjected to dynamic compression and compression-shear loading. The results from DNS are reviewed for the two loading configurations to discuss the effects of grain size on the variability of the macroscale measurements (typically from free surface velocimetry) expected in NPI and PSPI experiments. The main findings from the DNS show that in both cases, the grain size directly correlates with the coefficient of variation in the macroscale measurements, showing a decrease in the coefficient of variation with decreasing grain size. Remarkably, the magnitude of variations in the particle velocity record is shown to be largest where the deviatoric stresses are expected to be most significant. The mechanistic reasoning for the scatter in the particle velocity record is discussed and a power-law description of the magnitude of the scattering versus characteristic length is provided, enabling a statistical framework for assessing the required number of grains for minimizing scatter for these two experimental configurations.