Read this to learn everything you need to know about Meta Ads’ learning phase:

• What is the learning phase?

• Why is the learning phase important?

• Exactly how is Meta learning and improving the delivery of your ads?

• How you can take advantage of the learning phase to sell more

I wrote this for my own note collection, so it’s deliberately short and concise.

What is the learning phase?

When you first start an ad set, it is in the learning phase. This phase denotes a period when Meta doesn't have enough data or learnings to predict who to show your ads to for the best results. In the learning phase, performance is less stable, and you typically see a worse cost per action.

Why is the learning phase important?

In a study by Meta, advertisers saw a 19% lower cost per result in ad sets that exited the learning phase compared to those that didn't. The study also found a 68% lower cost per result for accounts that spend less than 20% of their ad spending in the learning phase relative to accounts that spend more than 50%.

Meta says an ad set "exits the learning phase" when performance stabilizes. Typically, this happens when you get 50 "optimization events," e.g., purchases, within seven days.

The "Delivery" column shows an ad set's learning status. If it says "Active," it's out of the learning phase. If it says "Limited learning," it's stuck in a purgatory of not getting enough data to exit the learning phase.

How is Meta learning and improving the delivery of your ads?

Each time one of your ads is shown, Meta learns more about whether it's working and who to show it to give you the most value for your money. As you get more data and more learnings, Meta gets better and better at showing the right ad to the right person, so you get better and more stable performance.

Meta learns on the ad set level, which is where you define your target audience. You can see the ad set as the brain that controls which of the ads in the ad set to show to who in the audience for the best results.

It's important to understand that ad sets don't talk to each other. The brain of one ad set doesn't speak with the brains of other ad sets. So, any learnings in one ad set don't help the brain of your ad sets. This makes account structure important because it determines how many ad sets you have in your account. With multiple ad sets, you need more data, more orders, to feed all the brains with enough data for them to exit the learning phase. If you don't, each ad set that doesn't exit the learning phase will be stuck in an eternal purgatory of bad results.

Ad sets re-enter the learning phase when you make significant edits.

Even if an ad set has exited the learning phase, it can reenter the learning phase after what Meta calls a "significant edit":

• Any change to targeting (audience at the ad set level)

• Any change to ad creative (text, image/video)

• Adding a new ad to the ad set

• Pausing an ad set for 7 days or longer

• Changing bid strategy (e.g. from lowest cost to bid caps)

Meta also say that a change might cause an ad set to reenter the learning phase if you do the following, depending on the magnitude of the change:

• Change an ad set's spending limit

• Change anything related to cost controls (bid, cost per result goal, or ROAS)

• Change the budget, whether it's on the ad set or campaign level (CBO'

Important note

Many people think an ad set resets all learnings after a significant edit. It doesn't. It simply re-enters the learning phase to "recalibrate" after the edit— based on both the old learnings and new data from the "new" learning phase. So don’t be afraid of deleting all your learnings.

Reduce time spent in the learning phase, but never avoid it completely

It's important to keep ad sets out of the learning phase most of the time. But it's equally important to test new ads and optimize your account. So, don't be afraid of making changes or adding new ads because of the learning phase. Just be sure to make changes only for a good reason.

Tips to reduce time spent in the learning phase

You can reduce the time spent in the learning phase by:

• Batch your edits and changes to, e.g., one day per week, so you're reentering the learning phase as little as possible.

• Consolidate ad sets so each ad set gets more data, and, consequently, exits the learning phase faster.

• Make sure you can get enough conversions in a seven-day period to exit the learning phase

  • High enough budgets

  • Big enough audiences

  • Cost controls that aren't too restricting

Key learnings

• Eliminate time spent in learning phase as much as possible. You get better results the less time your ad sets spend in the learning phase.

• Account structure is important as it determines how many ad sets you have and, thus, how much data you need for them all to to exit the learning phase.

  • Consolidation is key, if you don’t generate enough orders for all your ad sets to exit the learning phase. And most likely

• Only make changes to an ad set after it has exited the learning phase. Otherwise, you're making changes based on likely-to-change data.