How Facebook Machine Learning Prioritizes Users In Audiences

Facebook’s machine learning is meant to do 1 job. The job is serve your ads in the cheapest price possible.

What this means is that if you run an engagement ad and select a worldwide audience for it, Facebook will get you the engagement from the cheapest geo-location which could often be Philippines or India or another country like that.

Similarly, if you run a conversion ad and put in a large amount of countries in your audience section, Facebook will start prioritizing your traffic towards countries that are likely to convert but also where cheaper impressions can be served. So you’d see a large ad spend being done in Brazil or Mexico or Kuwait etc.

Many times, marketers want to sell their products or services worldwide but also have a preferred or prime geo-location which they want to serve first or primarily or in majority.

If you want to sell your goods or services in US and your goal is to have 50% buyers from US but also have some buyers across the world, you should ideally segment your ad-sets. What this means is you should create a separate ad-set for US and another one for rest of the world.

As a media buyer, it is no longer your job to find buyers. That is something Facebook does for you. Your job is to understand how Facebook’s machine learning works, which does it’s job well, but is far from perfect. As a media buyer you’re supposed to identify its shortcomings, and create your ads in a manner to get the best our of their machine learning.

Leave a Reply