The presently described embodiments relate to improving system productivity where maintenance purge routines are required through use of a digital front end (DFE) job scheduler. This approach utilizes knowledge of future jobs to maximize productivity. So, even if a low coverage area job is being processed, and a purge routine is scheduled, the purge routine may be avoided. This is achieved by projecting the system evolution over a future time horizon and determining the schedule of toner purge events (a non productive dead cycle) to minimize a cost function that penalizes the purge event (dead cycling and material loss should be minimized) and the deviation of average toner resident time in the sump from some desired set point of range. In this regard, knowledge that a high coverage area job is downstream and average toner residient time may be advantageously used to effectively perform the purge itself while in productive mode. The system gains knowledge of whether low coverage area jobs or high coverage area jobs are pending by using information stored within the print job file (e.g., a page description language job file). For example, a page description language (PDL) file typically includes information on the area coverage trajectory over time. This will allow a system to generate a predictive model which can constantly recalculate statistics based on knowledge of currently running jobs, new jobs or a change in customer criteria.