The norm optimal approach, both in its basic form and the extension to predictive action, where the predicted errors on a number of future trials are explicitly included in the cost function for controller design, is now a well established area in iterative learning control in terms of the underlying theory. By the fact that it includes the predicted errors on future trials in the cost function, predictive iterative learning control is clearly a higher order law. Hence it is now appropriate to ask if, in practical situations, predictive norm optimal iterative learning control can deliver significantly improved performance over its norm optimal alternative to merit the extra computational and hardware costs associated with its application. This is the area addressed in this paper using a somewhat new application area in the form of chain conveyor systems.