In part I of this report the results of a benchmarking exercise are presented. A deep excavation problem in Berlin sand has been specified by the working AK 1.6 of the German Society for Geotechnics and sent to various university institutes and consulting companies known to deal with numerical analysis in practical geotechnical engineering. The results summarized in this report clearly emphasize the need for these types of exercises in order to improve the reliability and validity of numerical models. The wide range of results submitted is by no means acceptable and therefore it is argued that guidelines and recommendations have to be formulated in order to assist numerical analysts in avoiding unrealistic modelling assumptions, which may lead to consequences the user of a particular code may not be aware of.