Sure, but that's not a rollback. That's just figuring out whether a rollout / which rollout caused a revenue drop. (It's basically as if they'd not done a 100%

Sure, but that's not a rollback. That's just figuring out whether a rollout / which rollout caused a revenue drop. (It's basically as if they'd not done a 100% rollout in the first place, but kept a 1% holdout group. The point isn't to get more revenue from that 1%, it's to be able to measure long-term impact from the original change.)