It is more efficient to aggregate the data in the hub and then export it, rather than accumulate it through the import to the downstream models
