This article focuses on the use of disposable, ad hoc corpora in translation. The study is based on a workshop experiment using the Web as a corpus resource. The main emphasis is on the compilation of these corpora and their analysis tools. A distinction is made between competence and performance in translation. It is pointed out that disposable corpora can be structurally simple but nevertheless very useful performance-enhancing tools in translation. A simple corpus structure does not, however, imply that there are no design criteria for these corpora. It is claimed that the knowledge of how to compile corpora and use them is an essential part of modern translation competence. A tentative list of corpus management skills is sketched at the end of the article.