Since Kragh and Christie's 2002 seminal paper on repeatability, nrms and predictability (PRED) have been the most widely used metrics for analyzing 4D noise in time-lapse studies. However, their values and behavior are rather counterintuitive, and they invariably end up stagnating above 15% and 0.95, respectively, in repeated marine data of variable quality. Moreover, their respective variations with regards to measurable differences in data, acquisition, and processing are still poorly understood from a quantitative standpoint. With the objective of improving our assessment, interpretation, and understanding of time-lapse repeatability, this paper introduces an analytical formulation of the 4D problem from the perspective of perturbation theory, building on the preliminary works of Calvert and high-quality data from our latest time-lapse campaigns in Gulf of Guinea. This framework brings to light blatant problems with our current repeatability metrics, and proposes a complementary set of QCs and best practices to get more out of 4D data. The cornerstone of this effort is the introduction of the new signal-to-distortion ratio (SDR) attribute as a true, reliable indicator of time-lapse repeatability. It provides the essential common ground on which different acquisition technologies, deployment and processing strategies are to be compared.