Every rule works in one of three ways, and knowing which is which explains why some rules have sliders and others have switches.
The three types
Adaptive rules learn what's normal for each account from its own recent history, then flag meaningful departures. You don't set a fixed number like "alert at 2% CTR" - you set how sensitive the rule is. A 2% CTR might be terrible for one client and great for another, so a fixed figure across all your accounts would be both noisy and blind.
State rules watch for a yes/no change rather than a number crossing a line - an ad going from approved to disapproved, a new user appearing, a campaign flipping to "not eligible". There's no threshold slider because there's no number; instead the Adjust panel gives you switches for what counts as a trigger.
Logical rules fire on a specific combination of facts. Negative keyword conflict is the classic case: a term converted and a negative in the account now blocks it, tested with proper match-type logic.
What is a baseline?
A baseline is the rule's picture of "normal" that it measures against.
For adaptive rules it's learned automatically - each account's typical daily spend, CTR, CPA and so on, over a trailing window of complete days.
For the security rules it's a list you approve - your known users, your allowed manager accounts, your allowed ad networks - so anything outside the list stands out.
