Monitoring
Activity log
Notifications say what happened. The log says who did it.
Every change that matters is recorded against the experiment it affected, with who made it and when: targeting rules added or removed, weights changed, a variant created, the experiment started, ended or deployed.
It is captured centrally rather than by each handler remembering to write a line, which is what keeps it complete. A new action added to the platform is logged by describing what it did, not by wiring up a second code path that somebody can forget.
The division of labour with notifications is deliberate. A notification tells you an experiment was deployed. The log tells you who deployed it, alongside everything else that was done to that experiment in order. That is the difference between an alert and an audit trail, and on an account with several people in it you want both.
One thing it deliberately does not do is travel. Duplicating an experiment copies its variants, changes, screenshots and targeting, but not its history: the log is the record of what was done to the original, and the copy starts with its own first entry saying it was duplicated.