As the media landscape grows more complicated, facebook advertisers are becoming more innovative in measuring their campaigns. Machine learning has become an exciting technology that can help marketers like yourself better understand their ads’ performance. 77% of all online users active on at least one Meta platform—which are users you should have access to.

Facebook understands the opportunities and challenges that Machine Learning brings to the table and the benefits it can have for advertisers in particular. Facebook IQ has explored how automated systems can improve campaign planning and execution, but most importantly, the measurement when traditional methods aren’t enough.

With everything, though, there will be a period of adjustment. In the case of Facebook’s Machine Learning tools, that ‘period of adjustment’ can be an outright nightmare. This forces you to try new things, fresh creative, and touch on brand-new audiences all at once. Just because it’s hard, though, doesn’t mean you should give up or ‘leave it to someone else,’ does it? Giving the system a chance to learn, re-route, and calculate results is part of the process – and it doesn’t have to make you tear your hair out. At least, not anymore. 

What Is The Learning Phase? 

The Meta ads system uses AI, or machine learning, to help you deliver and optimize your results. Every time an ad is shown, the system can learn more about how to target your audiences. It begins to learn who they are, what times to display the ad, and placements and creatives that work. The more this ad is shown, the better the delivery system can be at optimizing for performance. 

This is the process where these systems are still learning and getting acquainted with your ad set. It explores the best practices, trying new audiences, placements, and more. Some things can hold it up, like a lack of events or significant changes re-starting the process. However, about 50 optimization events usually mean you get to exit the learning phase. You exit with fine-tuned and optimized ads at your fingertips. 

Why Does It Matter?

In this early time, you’re not dealing with optimized results, or even delivery. The ad sets are less stable, with a higher CPA. Facebook advertisers who use around 20% of spend in the learning phase will see 17% more conversions. They’ll also likely see a 15% lower CPA than advertisers who use ~80% of spend in the same phase. If you spending more than 20% in the learning phase, you might need to optimize some things to move forward.

Optimize Your Spend

Edit Less Frequently (Or Not at All) 

Edits, otherwise known as Manual Optimizations, can trap your ads in the learning phase. The specific campaign, ad set, and ad edits reset the learning phase, meaning you’re starting all over. These include: 

Campaign – Budget, Bid amount, Bid Strategy

Ads – Any change

Ad Sets – Targeting, Placement, Optimization Events, Adding Creative, Bid Strategy, Bid Amount, Budget, and 7+ day Pause

You can reduce spending in your learning phase by avoiding edits in the ad set or the ad until after. This will help ensure your optimization decisions are based on results more indicative of future performance. With multiple edits, try to batch them at once, so it only resets that time.

Avoid Having Too Many Ad Sets 

High ad set volume will trap your ads in the facebook learning phase. When advertisers run too many ad sets at a time, each one will deliver less often. This means fewer ad sets can leave the learning phase, as it keeps resetting. You should be consolidating ad sets, reducing how many learning’s need to be delivered.

Some alternatives to high ad set volume that can help speed up time in the learning phase include: 

Combine similar ad sets into fewer, larger sets, or use the Store Traffic Objective – Great if you’re creating many ad sets to advertise to many smaller geographical areas. 

Use Automatic Placements and Asset Customization – Great for optimizing many different placements among different sets. 

Set Multiple Languages in a Single Set – Better than creating different ads for different languages. This enables you to dynamically show ads in the correct terminology for the audience reached. 

Aim For High Conversion Volume and Avoid Constrained Setups

However, there are other settings that could be holding you back from the conversion volume you need. These can include: 

Too-Low Budget – Since Facebook machine learning ad sets need about 50 optimization events over 7 days to get out of the learning phase, your ad set should have the appropriate budget. If you set a small or inflated budget, you’re providing the delivery system with indicators that may be inaccurate—thus hindering the information it needs to properly optimize. 

Too Low of Cost Cap/Bid – If you’re using target cost, cost cap, bid cap, or value optimization with minimum ROAS that can restrict your ad set from getting the conversions needed in 7 days, then the ad sets may never properly leave the learning phase.

Limited Audience Size – Larger audience sizes will more likely generate the conversions needed for an ad set to exit the learning phase. 

Infrequent Conversion Events – If your conversion event happens less than 50 times weekly, consider optimizing it for more frequent events. For example, if you see less than 50 purchase events in a week, you should consider optimizing for add-to-cart events instead. Ads with longer conversion windows will also need more time to exit the learning phase. 

Testing marketing strategies and creative assets are essential for the constant improvement of your ads. In turn, the learning phase is vital for optimizing the delivery system of your new ads.

The post The Facebook Machine Learning Nightmare: Tips to Avoid Insanity appeared first on Small Screen Producer.