I guess you could find out the storage required for data using the data load rate, then double it to account for indexes and then double it again to account for any underestimation you could have made.