Cellulosic depth filters embedded with diatomaceous earth are widely used to remove colloidal cell debris from centrate as a secondary clarification step during the harvest of mammalian cell culture fluid. The high cost associated with process failure in a GMP (Good Manufacturing Practice) environment highlights the need for a robust process scale depth filter sizing that allows for (1) stochastic batch-to-batch variations from filter media, bioreactor feed and operation, and (2) systematic scaling differences in average performance between filter sizes and formats. Matched-lot depth filter media tested at the same conditions with consecutive batches of the same molecule were used to assess the sources and magnitudes of process variability. Depth filter sizing safety factors of 1.2-1.6 allow a filtration process to compensate for random batch-to-batch process variations. Matched-lot depth filter media in four different devices tested simultaneously at the same conditions was used with a common feed to assess scaling effects. All filter devices showed <11% capacity difference and the Pod format devices showed no statistically different capacity differences.