We started scaling Brad's account when Campaign Budget Optimization (CBO) was relatively new. With CBO ad managers are not able to set bids and budgets at the audience level. Instead, campaigns will automatically, in real time, optimize towards the best performing, lowest-cost audiences based on what Facebook deems most efficient and scalable. This is another instance where Facebook is moving towards machine learning based on the pixel data to optimize ad campaigns.
In order to have success with this new format we did a lot of experimentation with two types of campaigns. One type of campaign had a huge budget with 10+ ad sets per campaign. The other was a smaller budget campaign with only 3-5 ad sets. This allowed us to start campaigns as high as $4k a day in some cases, and to know within 24 hours whether to scale or turn it off.
Our goal was to mainly scale horizontally, increasing ad spend with new interest targeting and look alike audiences. This also helped prevent campaigns/ads from burning out all at the same time - we were launching new high spending campaigns every day, so when something stopped working we always had another campaign in the pipeline to build on and replace the struggling one.
We made sure to constantly be testing new creative, with a focus on brightly colored, wacky images. We tried to launch new creative multiple times a week, and run new ads against old ads to see which continued to outperform. With our high ad spend we also built up lots of positive social proof on his ads, which further helped conversions.