Facebook ads learning phase: why you should wait
During the Facebook ads learning phase, an ad set’s numbers mean nothing yet. Spot it and you’ll stop killing campaigns that were working.
What happens during learning
When an ad set starts, the algorithm doesn’t yet know who to show your ad to. It explores: it tests profiles, times and placements, and watches what drives the event you optimize for. That exploring is expensive, and results go up and down.
The phase ends once the ad set has collected enough events for the model to settle. Until then, a good day isn’t proof, and neither is a bad one.
What restarts it
Any big change resets the clock: new audience, new optimization event, a sharp budget change, or swapping all the creatives at once.
That’s the most common downward spiral. It feels too expensive, so you change something. Learning restarts, costs go up, so you change something again. The ad set never leaves its unstable phase, and you decide Meta “doesn’t work.”
Break the spiral
Set a rule before you launch: for X days, I don’t touch anything. Write it down. When you get the urge to tweak (and you will), you’ll know the decision was already made with a cool head.
If you really must step in, change the ad, which disrupts less, rather than the audience or budget. And change one thing at a time, or you won’t know what caused the effect.
When an ad set never finishes learning
If an ad set stays stuck in learning week after week, it almost always lacks volume: the budget is too low, split across too many ad sets, or the optimization event is too rare.
Waiting longer won’t fix it. Consolidate: fewer ad sets, more budget in each, and if needed a more frequent optimization event to feed the model.