Learning phase: the number, and what it hides
The period after a change while the platform's delivery system is still calibrating.
Why learning phase matters
Results during it are unstable and unrepresentative. Judging an ad before it exits is how good creative gets killed and bad creative gets scaled.
What this looks like with real numbers
An ad set needs roughly 50 optimisation events to stabilise. At a $50 CPA that is $2,500 of spend before the numbers mean anything. Judging on day two, at $400 spent, is reading noise and calling it a result.
What it does not tell you
Editing an ad set restarts it. Teams that 'optimise' daily can keep an ad set permanently in learning and never see what it would have done.
The mistake people actually make
Editing budget or targeting mid-phase. Each meaningful edit restarts calibration, and an account edited daily never leaves learning at all.
Questions, answered.
How do you calculate learning phase?
Typically measured in conversions, not in days. The period after a change while the platform's delivery system is still calibrating.
What goes wrong with learning phase?
Editing budget or targeting mid-phase. Each meaningful edit restarts calibration, and an account edited daily never leaves learning at all.
What does learning phase not tell me?
Editing an ad set restarts it. Teams that 'optimise' daily can keep an ad set permanently in learning and never see what it would have done.
Change the creative, not the dashboard
The period after a change while the platform's delivery system is still calibrating.