The downward revisions were overall a lot bigger than the upward revisions. BLS methodology causes it to get data from larger companies first, and this tends to systemically bias it towards downward revisions