There is a gap in the research literature with respect to explaining why some failing IT projects are allowed to continue instead of being terminated. Part of the problem is a gap with respect to an accepted general definition of IT project failure. This is significant because varying definition of IT project failure gives rise to misunderstandings in research and practice. The use of the derogative term “failure” is in itself problematic. This article gives a) an overview of what is meant by IT project failure in the literature, b) an updated set of IT project performance criteria associated with the attribution of failure, and c) a new argument for the theoretical predictability of IT project termination - the marginal cost trap. The article furthermore presents d) an outline of how these findings can contribute to preventing IT project failure. The research methods are a hermeneutic literature review (content analysis), small-n case studies (plausibility probes), and abductive theory generation.