The splitting of SK(K)S phases is an important observational constraint to study past and present geodynamic processes in the Earth based on seismic anisotropy. The uniqueness of the derived models is unclear in most cases, because the azimuthal data coverage is often limited due to recordings from only a few backazimuthal directions. Here, we analyze an exceptional dataset from the permanent broadband seismological recording station Black Forest Observatory (BFO) in SW Germany with a very good backazimuthal coverage. This dataset well represents the potential teleseismic ray paths, which can be observed at Central European stations. Our results indicate that averaging splitting parameters over a wide or the whole backazimuthal range can blur both vertical and lateral variations of anisotropy. Within the narrow backazimuthal interval of 30–100°, we observe a complete flip of the fast polarization direction. Such a splitting pattern can be caused by two layers with about NW–SE (lower layer) and NE-SW (upper layer) fast polarization directions for shear wave propagation. However, the possible model parameters have quite a large scatter and represent only the structure to the northeast of BFO. In contrast, within the wide backazimuthal range 155–335°, we prevailingly determine null splits, hence, no signs for anisotropy. This null anomaly cannot be explained satisfactorily yet and is partly different to published regional anisotropy models. Our findings demonstrate that there is significant small-scale lateral variation of upper mantle anisotropy below SW Germany. Furthermore, even low-noise long-term recording over 25 years cannot properly resolve these anisotropic structural variations.