In this paper, strains are monitored in-situ at the surface of a polycrystalline 316L sample loaded quasi-statically in uniaxial tension. Initial orientation data collected over thousands of grains is compared with the strain data evolution in order to investigate, with statistical significance, the relationship between the microstructural features and the strain localization patterns emerging at the grain scale. It is shown quantitatively that high values of strain appear early and primarily around grain boundaries, before growing into a network spanning several grains and grain boundaries. Strains are also heterogeneous within a grain, with the heterogeneity tending to be more pronounced in larger grains. Statistical analyses demonstrate that the usual microstructural descriptors-such as Taylor factors, average intragranular misorientation angles or intergranular misorientation angles-do not display any correlation with the localization network. Furthermore, higher Schmid factors, albeit exhibiting some weak correlation with higher strains, fail to systematically identify the grains that experience the largest strains. It is thus proposed to analyze three-dimensional descriptors of orientation expressed in the Rodrigues space and they are shown to help readily identify orientations that exhibit high strains early in the loading. Additionally, the study of the full three-dimensional misorientation information, expressed in the reduced Rodrigues space, clearly highlights the fact that some misorientations oppose localization at certain boundaries in the microstructure at hand. Hence, considering the full three-dimensional orientation (and misorientation) data instead of just scalar descriptors is demonstrated to be of great interest when investigating the relationship between microstructural features and strain localization.