Learning in Advertising

What are the Learning Stages?  

In the Facebook advertising system, as in other advertising platforms, there is machine learning. With machine learning, the optimization of the ads provided is more accurately ensured. After the ads are published, factors that change the efficiency of the ad, such as target audience, ad space, and broadcast time, are tested with each impression. The more ads are shown, the easier it is to optimize ad performance. However, as the name suggests, the ad performance may not be stable during the learning and trial process. CPAs are usually high. Advertisers may panic in this situation and make changes to ad sets, but this is a move that will negatively affect the learning process. We will discuss in detail which changes fundamentally affect the learning process in the later sections of our article.  

Ad sets exit the learning stage when their performance becomes stable. Performance usually stabilizes after ad sets achieve about 50 optimization events within 7 days. Ad sets that fail to obtain sufficient optimization events or are expected not to obtain them become “Limited by Learning.” The Limited by Learning warning is an indicator that your budget is not being spent efficiently because the ad set cannot perform optimization. The main reasons for not being able to optimize are situations such as small target audience size, low budget, low bids, high auction overlap, or running too many ads simultaneously. To get your ads out of the Limited by Learning state, you can make the following adjustments:  

  • Increase Your Budget. This allows your ad to be displayed more and increases the likelihood of obtaining optimization events.
  • Combine your ad sets and campaigns. This way, you combine your target audiences and budgets, reaching more people and increasing the likelihood of obtaining optimization events.
  • Expand your target audience. You may want to target your ads to a specific audience, but the probability of a small target audience bringing optimization events is low. By expanding the target audiences of the ad sets you create related to your business or the product/service you offer, you increase the likelihood of obtaining optimization events.
  • Increase your bid or cost control. Your bid may not be high enough to obtain a sufficient number of optimization events, which is why you may encounter ad sets remaining as Limited by Learning. If you increase your bid or cost control, you increase the likelihood of obtaining optimization events.
  • Change your optimization event. Your optimization event may not occur enough. For example, you can choose events like starting shopping or adding to cart instead of a shopping event.
  • Check your existing campaigns. Your active campaigns may hinder the efficiency of your new campaign.  

In addition to these issues, there are other factors that prevent your ad sets from exiting the learning process. These factors are performed by advertisers. Some campaign, ad set, and ad edits reset the learning stage. Edits that cause the ad set to re-enter the learning stage:  

  • At the Campaign Level; Budget change, bid change, bid strategy.
  • At the Ad Set Level; Targeting, ad placement, optimization event, adding new creatives, bid amount, budget, stopping for 7 days.
  • At the Ad Level; All changes in ad creatives.  

In addition, when using Campaign Budget Optimization, changing your bid strategy can cause multiple ad sets within the campaign to re-enter the learning stage.  

Some changes may not affect the learning process depending on the size of the change:  

  • Ad set spending limit amount
  • Bid control, cost control, ad spend return-related control amount
  • Budget amount; If you change your budget by less than 50%, the probability of entering the learning stage is not high.  

Also, adding new ad sets to your campaigns does not affect the learning process of your existing ad sets. In campaigns with multiple ad sets, changes you make in one ad set do not affect the learning process of your other ad sets.  


Tips Related to the Learning:

  1. Spend around 20% of your budget on the learning phase. This will allow the algorithm to gather enough data to optimize your ads effectively.  
  2. Avoid making too many changes to your ad sets during the learning phase. If you need to make changes, try to do them in one go. 
  3. Avoid creating too many ad sets as it can make it difficult to exit the learning phase. 
  4. Use Breakdowns to separate your audience and get separate reports instead of using different ad sets. 
  5. Use Automatic Placements or make changes in the Ads Manager instead of creating separate ad sets for different placements. 
  6. Use a single ad set with multiple language options instead of creating separate ad sets for each language. 
  7. Don’t be afraid to experiment with new strategies during the learning phase as it is crucial to improving your performance. 


Perfist Blog

Similar Articles

Other Articles