The 2005 break in retention
I really appreciate how you've documented the edge cases here — the distinction between deleted records and deactivated ones matters more than most people realize in retention calculations.
Should headcount be as a share of total for this comparison to mean anything?