For over a hundred years, Charpy impact testing has been performed. It is one of the most frequently used material tests. Due to a very simple test setup, a huge wealth of experience and an enormous database of results, it is still state of the art to evaluate toughness of materials and welds. Modern welding technologies like laser welding or electron beam welding (EBW) are characterized by a low heat input. The high energy density of these technologies results in very narrow welds. These kind of joints are difficult to analyse by means of a standard Charpy test. The combination of weld cross section and properties lead in most cases to a fracture path deviation (FPD). Although the crack starts from a notch, it deviates during propagation into the heat-affected zone or even in the base material. Therefore, the weld itself is not tested and cannot be characterized. FPD is a wellknown issue but so far little attention in standards has been paid to this complex topic. Hence, establishing a valid welding procedure specification for beam welding procedures may imply difficulties. This work focuses on avoiding FPD in electron beam welds in soft-matensitic steel (1.4313) by using standard and side notched Charpy impact specimens. Valid toughness result had to be found for 20-and 100-mm-thick welds at −20 • C. Several test with three different kind of specimen in two different heat-treated conditions were carried out. It was found that beside the narrow seam a major reason for FPD is the significant overmatching of the weld. A post weld heat treatment to reduce overmatching facilitates testing but can decrease toughness. Adequate results for qualification were found and EB welds reached sufficient toughness at −20 • C. Toughness of EB welds was compared to the toughness of conventional gas metal arc welds (GMAW). Finally, a recommendation for adapting the toughness characterization for narrow and overmatching seams is proposed.