The present study researches the implementation of a web based intelligent tutoring system for mathematics at lower secondary schools. In recent years, there is growing concern about the worrying situation at German lower secondary schools. Data from large scale educational assessments in the county of North Rhine-Westphalia (NRW) show that children at lower secondary schools have an embarrassing paucity of basic mathematical skills (Leutner et al., Lernstandserhebungen 9. Klasse 2004 in NRW: Erster Kurzbericht zur wissenschaftlichen Begleitung, 2004). In order to improve these basic mathematical skills in lower secondary school children, several schools implemented the web based intelligent tutoring system eFit. The aim of the present research was to investigate whether eFit constitutes an effective intervention of this target group. The results show that compared to a non-treatment control group, children in the eFit group significantly improved their arithmetic performance over a period of 9 months. As will be discussed, the findings have to be treated with cautions because eFit was specifically designed to alleviate mathematical difficulties and therefore “trained for the test” whereas traditional mathematics instruction followed the regular curriculum. The implications of this will be considered in the light of existing theory and research.