I read a lot of conflicting advice first in my own time, so the sample is at least reasonable. What actually decided it was how new accounts are treated in week one, which changes noticeably month to month. It came through clearly consistently enough to trust well beyond what I would call noise.