The report the budget was decided on showed 12,000 sessions out of 540,000: two weeks after go-live
Two reports on the same property over the same period, two different numbers. It wasn't our server-side release: it was an operator that filtered instead of classifying.