You may know that I’ve largely come to the conclusion that my targeting inputs don’t matter a whole lot, and I’ve adopted using Advantage+ Audience. But, I know that many advertisers resist it, and I don’t ever want to blindly take a path without testing it thoroughly. So, my goal was to prove one way or the other whether this was the right approach for me.
I wanted to see whether targeting impacts conversion results, and how much. I also wanted to be sure to measure the quality of those results, since low quality is often the complaint related to Advantage+ Audience.
The Test
I created a campaign with three ad sets. Each one used the following settings:
1. Optimized for conversions using the Complete Registration event
2. Attribution setting of 1-day click
3. Advantage+ Placements
Each ad set used a different targeting approach:
- Advantage+ Audience without suggestions
- Original audiences using detailed targeting
- Original audiences using lookalikes
Since these ad sets optimize for conversions, detailed targeting and lookalikes are automatically expanded using Advantage Detailed Targeting and Advantage Lookalike (expansion cannot be turned off).
The ads for all three were identical. One of the ads used a single static image and one used Flexible Ad Format.
Each ad sent people to a landing page for a lead magnet that was unique to that targeting approach (there were three different landing pages, and they all looked the same). Each landing page used a different form so that I could track registrations in my CRM.
Results
Surface level results (CPM, reach, impressions, and frequency) were nearly the same, though CPM for Advantage+ Audience was the lowest.
In the end, it was close, but Advantage+ Audience generated the most registrations. Advantage+ Audience and detailed targeting were at the top, and lookalikes were far behind.
But, what about quality? Since I used unique forms, I could track that. Advantage+ Audience also generated the most quality registrations. So, no, using Advantage+ Audience did not lead to more low-quality results in this case.
To learn more about this test, how I set it up, and the results, check out my blog post here.