The Eurocode EN 1997-1 permits three different design approaches, each of which mostly leads to quite different results The design approach, adopted in the German standard DIN 1054, is the one that best enables geotechnical engineers to benefit from the experience gathered with the method commonly used hitherto in Germany, in which global safety factors are applied. As to numerical approaches, for example, the finite element method, the question arises which partial safety factors may apply. The values used in classical earthwork design have been established empirically and are based on experience and certain models. In this situation, the geotechnical expert has to accept the responsibility. Fortunately, the new standards confirm the position of the expert and lay down appropriate rules. The possibilities and limitations of numerical methods in geotechnical design according to standards have been discussed in the context of the various verifications required for a retaining wall. Numerical methods, such as the FE method, enable the boundary conditions to be taken into account to a far greater extent than in classical earthworks design. They also enable failure patterns to be forecast from deformations occurring close to the point of failure, instead of having to assume them a priori for a particular verification. There are only a few exceptions, such as the ultimate limit state in the lower failure plane or hydraulic heave, where the FE method needs additional reasoning concerning deformation and failure processes. The procedure generally specified in the new edition of the standard (Design Approach 2) is suitable for FE calculations; it involves performing the entire calculation with characteristic values first and only then applying safety factors Performing the calculation with design values would result in a distorted picture of the actual material performance and lead to unrealistic results Experience has shown that actions and the effects of actions can be determined reliably by the FE method. The FE method certainly delivers more satisfactory results than other models, particularly in the case of difficult geometries and complex construction processes. In addition, a single FE calculation is able to provide not only the effects of the actions for ultimate limit state design (ULS), but also those needed for serviceability limit state design (SLS). To enable the resistances to be computed, it would have to be ensured that it is possible to calculate the deformations reliably up to a point just before failure occurs and that a false appearance of failure is not caused by numerical instabilities. In addition, a consensus of opinion on how to bring about failure would have to be reached. This can either be done by reducing the shear parameters incrementally or by applying an additional action, although different numerical limit states are then obtained. Therefore, it can be recommended at the present time, that the actions and effects of actions obtained in FE calculations should be included in verifications and the resistances computed in accordance with classical earthworks design. The partial safety factors used to compare the effects of actions and resistances should be the same as those specified in standards for classical earthworks design, as more extensive experience is not available yet. Due to increasing calculating capacity of the computers, the application of the finite element method increases, creating the problem that differing results will be obtained, dependent on the computer program, the design approach and the material models chosen, being the basis of endless discussions. Therefore, the engineering community should strive for example calculations and calibrations to document correlations of and deviation from other calculation methods. A public database of such relations including comparisons with measurements would be an optimum.