Fatigue life prediction of scanned weld geometries can be less accurate than prediction using an idealised weld radius. One possible solution is the use of elastic plastic material models. Furthermore, the accuracy can be improved by adjusting the material model retrospectively. However, this cannot be applied to fatigue life prediction in industry. An alternative solution is to introduce additional factors that take into account stress, strain or shear gradients and geometric factors. Scanned weld topologies show radii of up to 0.1mm in the critical areas. These radii can be smaller than the grain size of the base metal. To investigate the difference in accuracy between scanned and idealised weld geometries, we examined the effect of smoothing the scanned surfaces. We also determined different gradients and found the cause of the non-conservative fatigue life prediction in the FDPRM and FDPFFS. The subsequent effect on FDPFS, FDPFGF, FDTSWT was also investigated