The local weld geometry and its variability can significantly affect the fatigue strength of structures, especially for non-load-carrying welds. Standardised definitions, such as sharp transition radii or undercuts, govern stress-raising effects at the weld toe. High-resolution digital tools can nowadays accurately determine these parameters, allowing for studying the impact of geometry variability on fatigue strength. However, real welds rarely exhibit idealised transitions as multiple radii, undercuts, and ripple lines introduce uncertainty in geometry estimations. Numerical simulations of the actual weld geometry, with all its variations, in combination with probabilistic evaluations, have shown great potential for studying the influence of competing notches in the weld. This study compares probabilistic evaluations of 3D scanned welds with analytical relations for stress concentration factors. Results reveal no clear trend between the analytical expressions and the ratio of simulated sectional stress to nominal stress. This highlights the challenge of directly applying existing analytical equations to idealised measurement data from real welds in their as-welded condition for fatigue strength estimations.