A number of issues concerning the verification of ULS using numerical analysis are described by means of three simple benchmark problems. There are pros and cons associated with the two limit state factoring approaches: material factoring approach (MFA) and load and resistance factoring approach (LRFA). Based on these simple examples, for the calculation of factored structural forces, LRFA provides a more consistent level of conservatism. MFA often also provides satisfactory levels of conservatism, but when there is no soil yield, the factoring can have a negligible effect, and where there is a lot of yield, the degree of conservatism can also be too low. For the verification of geotechnical ULS, MFA is more straightforward because it involves the factoring of input parameters and identifies the critical failure mechanism, as well as its margin of safety. LRFA requires particular geotechnical failure forms to be verified but forcing models into particular forms, even in simple cases, is difficult to achieve. Alternatively, outputs of mobilised resistance can be compared with independently calculated limiting values, but this was found to be unworkable for the passive resistance of an embedded retaining wall where the factored output of horizontal stress exceeded the factored passive limit, no matter how deep the wall embedment. Consequently, when using numerical analysis, it is recommended to employ both approaches LRFA and MFA, which together are equivalent to DA1 or DA2+DA3 in Eurocode 7, in order to benefit from the advantages of each.