The paper explains the theoretical background and findings of an empirical study of revision policies, using Denmark as a case in point. After an overview of important definitions, types and parameters, the paper explains the methods and data gathered from a questionnaire survey and an interview survey. Results clearly show that most translation companies regard both unilingual and comparative revisions as essential components of professional quality assurance. Data indicate that revision is rarely fully comparative, as the preferred procedure seems to be a unilingual revision followed by a more or less comparative rereading. Though questionnaire data seem to indicate that translation companies use linguistic correctness and presentation as the only revision parameters, interview data reveal that textual and communicative aspects are also considered. Generally speaking, revision is not carried out by specialised revisers, but by staff translators, who revise the work of colleagues and freelancers on an ad hoc basis. Corrections are mostly given in a peer-to-peer fashion, though the work of freelancers and inexperienced in-house translators is often revised in an authoritative (nonnegotiable) way. Most respondents and interviewees are worried about increasing pressures on the translation market, which, combined with customers' general lack of understanding of the translation process, mean that systematic, all-encompassing quality assurance is rarely financially viable.