Translation Memory systems are already well established tools among translation services and freelance translators today. They are claimed to be particularly useful in contexts where the texts translated are highly repetitive. This thesis was written in collaboration with the German section of the Central Languages Services at the Swiss Federal Chancellery, a service where most translated texts are not very repetitive and which consequently does not work with a tool yet. The thesis examines different aspects of the context, and in particular the texts, before testing them with two tools, SDL Trados and MultiTrans, to determine whether such a tool would be useful for them. In a more general perspective it also approaches the question of the tools' usefulness for any translation context where repetition in the texts is limited. These questions are answered through systematic testing, and in particular a scenario test which also investigates the user-friendliness of the tools.