Translating and revising for localisation require specific skills that are distinct from those required for translating and revising for publishing or other purposes. Their distinctiveness is due to several factors, particularly the need to work within a translation memory (TM) editor environment, rather than the usual word-processor. When using TM, translators and revisers are constrained by strict requirements regarding terminology, and leveraging and reuse from memory databases. Recent studies on translating with TM have shown that segmentation imposes a narrow, bottom-up view of the text, which restricts translators in their preferred approach. Other recent studies of revision in general have indicated, moreover, that there is a significant risk that revision may not always improve final text quality. This risk is especially likely to apply to the revision of TM-mediated translation, since revisers are placed in the top-down role of pulling together (sometimes not even sequentially ordered) the segments into a coherent whole. This article argues that the practice of translation and revision for localisation is being shaped by the needs of a new and unprecedented type of user: the TM apparatus itself. Research and discussion will enable participants to better understand their roles and responsibilities within this new scenario.