Meta is Testing Custom Audience Expansion

When I logged into Ads Manager recently, I noticed a new alert at the top. I had been selected for an advertising study. Based on the description, it seems Meta is testing Custom Audience Expansion.


I know that’s not the best image. And I no longer see the message. But the full alert read as follows:

This ad account has been selected to participate in a 9-week study to help improve campaign performance. A small proportion of ads using Custom Audiences will be delivered to people beyond the Custom Audience. Expanding your audience is often an effective way to improve results, but you can opt out.

Next to the message was a button to opt-out.

First: Kinda cool to be part of a study!

Second: Oh, crap. What is this?

Let’s sort it out…

A Continuation of a Trend

Custom Audience Expansion would be a continuation of a trend. First, there was Detailed Targeting Expansion (which became Advantage Detailed Targeting). Then there was Lookalike Expansion (which became Advantage Lookalikes). In each case, the approach was similar:

  1. You pick an initial audience to target
  2. If Facebook/Meta think you can get better results, the audience will be expanded

That is precisely what’s happening here with CUSTOM audiences now (at least that’s what this looks like). You choose a custom audience or group of custom audiences to target. But, if Meta thinks you can get more or better results by expanding outside of that custom audience, the audience will be expanded.

Lack of Transparency Regarding Reporting

One problem, which I keep making regarding expansion of detailed targeting and lookalikes, is that we just have to take Facebook’s word for it. Technically, the audience will only be expanded if and when you can get more or better results. In theory, it can only be used for your benefit.

But, there’s no way to confirm that it was helpful. There’s no break-down option to see things like:

  • How much your audience was expanded
  • How many conversions happened because your audience was expanded
  • The overall impact of expansion on your results

Because of this, we just have to trust the expansion, and that doesn’t feel right.

The Relevant Message Problem

One of the reasons remarketing is so powerful is that you can reach people with extremely relevant messaging based on things like:

  • What they bought
  • What they viewed on your website
  • A specific action they performed

But, the way that you target these groups is with a custom audience. So, let’s say you create an ad targeting a custom audience with a very relevant ad — an ad that would be irrelevant to anyone else. We obviously would not want to reach someone outside of the custom audience.

I understand that there may be cases when expanding a custom audience may make sense. For example, if you’re targeting your website visitors generally to promote a product and you just don’t get the volume of traffic to exit the learning phase. Some expansion could be helpful.

But not always. In the example I describe, I absolutely do not want the audience expanded. Hopefully, it won’t be on at all times with an inability to turn it off. Of course, expansion always is on in some cases related to detailed targeting and lookalike audiences, so there’s reason to be concerned.

Trust Expansion?

This test is part of a continued trend related to Meta Advantage:

  • More automation
  • Less control
  • More trust in Facebook
  • More hidden behind the curtain

While automation can be helpful, too much automation can be a problem.

Your Turn

Do you have an ad account that is part of this test? What do you think?