Stuck in the Learning Phase

If you see “Learning” in the delivery column of your ad set, the algorithm is still learning how to deliver your ads. Makes sense, right? Performance will fluctuate and you may see higher costs during this time.

Learning Phase

Ideally, you want at least 50 optimized events per week to exit the learning phase. It can happen before that, but this is the target number. If you get there, performance should become more stable — and even improve.

So, does it matter if you’re stuck in the learning phase or you get that dreaded “Learning Limited” label?

Well, to a point…

If you have the resources to spend more to exit the learning phase, it can help get optimal results. But this may not be reasonable for most advertisers, especially when optimizing for purchases.

There is one step that almost any advertiser can take. In many cases, this message could be avoided. If you create multiple ad sets for cold audience targeting, each ad set will need to get 50 optimized actions to exit learning. Instead, you can also combine ad sets — and budgets — to get there.

Otherwise, know that you’re at a disadvantage. The system has limited data to base delivery changes on. It’s reasonable that if the algorithm struggles to learn and make accurate adjustments, delivery may not be efficient.

The inability to exit the learning phase isn’t a death sentence for your ad set, though. Bottom line: Are you getting acceptable results? If so, don’t worry too much about this.

You could probably get better results if you were able to spend more, but there’s nothing wrong with letting it run if the results are good enough.

Let performance be your guide, not a warning message.