Translation Memory systems have become established as a key technology to address the challenges faced by the language services industry today. Despite flourishing research on how TM technology could reach the next level of technical sophistication, there has been a lack of systematic research on the usability aspects of TM systems, leaving many questions unanswered as to how translators' actual needs could be met more effectively by future TM solutions. The goal of this thesis has been to bridge the gap between current R&D efforts and translators' work reality by identifying and investigating the needs of modern translation professionals with reference to TM technology. A survey was used to elicit information from TM users, with 874 translation professionals giving their views on what they liked, disliked and wished for in TM systems. The information was analysed using a system of coding and analytical elements of the grounded theory approach, as well as user profiling, in order to reveal the needs of various distinct groups of translation professionals, in relation also to particular TM use situations. The identified needs relate to the functionality of TM systems, specifically to the TM translation editor, its facilities for resource building and management, project management and quality control, its search and translation assembly capabilities and its collaboration enabling features. In addition, several other needs concern non-functional aspects of TM systems, namely the systems' usability. The thesis also explores those needs that cannot be derived immediately from articulated user requests or problems, but can be anticipated from a number of significant implicit clues. A number of such clues reveal a growing need for technologies that can access vast lexical and factual resources, or integrate such resources in a TM-enabled workflow. Such a subtle need is believed to hold the key to innovative TM systems.