Mean age theory is introduced to characterize the mixing performance of tundish based on the spatial distribution of tracer's mean age. Conventional residence time distribution theory was widely used in the tundish analysis; however, it contains no information of the local mixing states. Based on mean age distribution, melt change efficiency is defined as a performance index to evaluate how quickly the old melt in the tundish can be replaced by the young melt from the ladle. Case studies, divided into three groups, were carried out to test the applicability of the new theory in a single-strand tundish with flow control devices of weir, dam and turbulence inhibitor. The developed mean age model was well validated by comparison with measurement in water model and computational fluid dynamics (CFD) results using residence time distribution (RTD) model. Mean age model can reduce the computing time to two orders of magnitude less in comparison with conventional transient RTD model, which improves the feasibility of CFD modelling in parameter studies to a broader extent.