The intent of this study was to investigate how to maximize the chances of success of an educational innovation--specifically one based on the implementation of the educational modeling language called EML. This language is both technically and organizationally demanding. Two different implementation cases were investigated, one situated in an institution for higher professional education that caters for on-campus students; the other in an institution for higher, open distance education, that serves off-campus, home-based students. Diffusion-innovation theory is used as the backdrop for the analysis. It helps us to understand why the implementations failed and what measures might be taken to avoid future failure in similar situations. (Contains 3 tables.)