It is understood that rocks do not instantaneously transport along a grinding mill and are consequently not perfectly mixed, but this is conveniently overlooked in mill modelling. A lack of understanding of the transport phenomenon and the difficulty of modelling realistic transport in a complex tumbling system, contribute to this. Visual observation of segregation in AG and SAG mills and measured lag time of mill response to changes in feed provide evidence of the influence of axial transport within mills. Incorporating this lag in dynamic models should underpin dynamic simulation of milling circuits for design and for model-based control. Limitations and errors in predicting the influence of increased power via mill length versus diameter, inability to predict pebble production, and a lack of the differential in residence time across rock sizes and competences, all contribute to inadequate design prediction that can be especially acute for large Autogenous mills. Examples of dramatically under – and over_x0002_loaded pebble regrind circuits and the inability to predict pebble load in AG milling discouraging the application of this energy-efficient option, point to the need for predictive modelling of breakage by competence linked to transport through and out of mills. Incorporating progress in granular flow and mixing, coupled to computational advances enable models to be based on complex breakage and transport models at high spatial resolution. This allows mechanistic models to simulate thousands of times faster than real time, opening the door to more accurate mill design, including dynamics in process simulation and to live simulation for process control – all with advanced mechanistic models. The value of grasping this opportunity is not easily refuted.