In spite of the continued importance of an innovation's attributes to research methodologies, and the increasing tendency toward multidimensional conceptualizations, the lack of a theoretically derived and empirically developed classification of innovations, conceived in terms of these perceived characteristics, continues to deter substantive research in the area. The absence of a stable descriptive framework has constrained researchers' facility to develop cross-case and cumulative research. In this paper, in which innovations are conceptualized as complex and multi-dimensional, we report on a mixed-method, exploratory study addressing the question of innovation classification. Data from a rigorous thematic investigation of the literature and four case studies, are synthesized into a descriptive framework incorporating 13 variables (innovation attributes). Following operationalization of the framework, we conduct a cluster analysis of the returns from a post-adoption survey of 310 innovations. Three distinct innovation types are identified: readily-adopted, challenging and under-cover. The attributes disruption, observability, profile and risk were found to be particularly important in distinguishing clusters that offer opportunities for new theoretical development. The UK National Health Service (NHS) forms the context for the study. Implications for theory and practice are examined.