Best practices for gathering stats on ERPi schema for improvement in ARM data loads

We've observed variation in table stats due to dynamic nature of some volatile tables used in ARM loads e.g. TDATASEG_T. Our auto gather stat window typically start at 10 PM each night & if we run our ARM loads at same time, we observed that optimizer collects stats for TDATASEG_T table when there are few hundrard thousand rows in it, however after completion of ARM load, this table gets truncated & will have zero records.

On next run of ARM load, optimizer considers those few hundred thousand records as per table stats & creates SQL execution plan according to it; which proved to be worse plan & our loads were getting delayed.