The activity of innovative practice relates to seizing opportunities now for patients in desperate situations and the ethical dilemmas around it. Sadly, the term innovation has multiple meanings, adding confusion and misunderstanding to an already difficult debate. This paper aims to enhance the definition of innovative practice in two ways. Fist, we propose to curb ambiguity by replacing the term "innovative practice" with the more precise of "new non-validated practice" and by distinguishing the specific meaning of innovation at stake. To identify this meaning, we analyse the traditional research ethics’ distinction between research, validated practice, and innovation. Second, we propose the following unified definition of "new non-validated practice", that is, the first or recent use of diagnostic, therapeutic or preventive interventions that introduce a significant change (new); with an insufficient level of evidence of safety or efficacy for regular healthcare (non-validated); and with the main aim to benefit patients (practice). To avoid objections to our definition, we analyse its three core elements (significant novelty, insufficient validation for, and patients' best interests) and we present a 2D classification of medical practice that follows from our definition. If sound, our enhanced definition of "new non-validated practice" allows comparing exceptional activities in different fields of medicine that previously have been considered unconnected (e.g. compassionate use of investigational drugs, humanitarian uses of devices, novel off-label uses, etc.). In turn, it also provides a promising conceptual tool to inform empirical research, discuss responsible access to innovative care and evaluate the regulation of innovative practice (e.g. right-to-try laws).