Expanded Interests

If you’ve used Ads Manager, you’ve certainly run into this section under the detailed targeting section.

It’s available for ad sets that are part of campaigns that have any of the following objectives: conversions, app installs, lead generation, engagement, or traffic – and the functionality is fairly straightforward.

If you’ve created lookalike audiences, you know that lookalikes can’t be created from interest based audiences.

The “expand interests” box somewhat fills that void.

By checking this box, Facebook will not just target those who are in the interest based audience but also those who are most similar to that audience.

However, Facebook claims that if the algorithm does not believe that expanding the audience will yield better results, it will not deliver impressions beyond the target audience.

From this, we can assume that if audience expansion is selected, the initial period includes at least some (and maybe a lot of) exploration of potentially inefficient audiences to compare against the core target audience.

My hypothesis was that initially, expanding the audience will cause worse performance, but that over time, it’ll perform better.

Here’s how I put this to the test.

I created a campaign with a web conversion objective. Then I nested two ad sets with the same targeting and set a daily budget of $300. To do a true A/B, I used “User Segments” to divide the two ad sets.

Here’s what the targeting looks like:

Regular audience:

Expanded audience:

Here were the results!

“Starz” regular, non-expanded audience:

182 conversions, $10,070 spend, $55.33 CPA
3.7M audience size, overall campaign frequency of 2.27

  • Week 1: $50.55 CPA
  • Week 2: $29.40 / cumulative after week 2: $36.45
  • Week 3: $63.14 / cumulative after week 3: $42.70
  • Week 4: $59.72 / cumulative after week 4: $46.08
  • Week 5: $326.54 / cumulative after week 5: $55.33

“Starz” expanded audience:

220 conversions, $10,088 spend, $45.85 CPA
100M audience size, overall campaign frequency of 1.79

* The 100M audience size is estimated as if the interests were excluded from targeting

  • Week 1: $55.69 CPA
  • Week 2: $46.73 / cumulative after week 2: $50.52
  • Week 3: $27.84 / cumulative after week 3: $39.41
  • Week 4: $45.18 / cumulative after week 4: $40.74
  • Week 5: $94.32 / cumulative after week 5: $45.85

The hypothesis was corroborated in this case. The regular audience performed better out of the gate, and maintained stronger performance in the first two weeks.

Then, in week three, the expanded audience was much more efficient, and overtook the regular audience in overall efficiency.

In weeks 4 and 5, the expanded ad set was again more efficient.

However, it wasn’t until each ad set spent ~$7,000 that we realized the expanded audience was clearly starting to outperform the regular audience.

Even if you look at the end numbers, there’s only 74% confidence that the expanded audience performs better than the regular audience.

It’s likely that this is the case since both audiences were huge, as even the addressable audience for regular, non-expanded ad set was 3.7 million.

Because of this, I decided to test this with smaller audiences. The hypothesis here is similar to the previous hypothesis, that initially the regular, non-expanded ad set will outperform the expanded ad set, but over time, the expanded ad set will actually perform better.

The second hypothesis is that with a smaller audience, the expanded ad set will perform much better later on, as the non-expanded ad set will start to reach much higher frequency and cause fatigue within the audience.

“AXS TV” regular, non-expanded audience:

370,000 audience size, overall campaign frequency of 2.39

  • Week 1: 9 conversions, $2,099 spend, $233.27 CPA

“AXS TV” expanded audience:

100M audience size, overall campaign frequency of 1.32

  • Week 1: 39 conversions, $2,119 spend, $54.33 CPA

The expanded audience ad set performed miles better right out of the gate, so test was ended after just a week.

Since the difference was so drastic, I wanted to repeat this test with another smaller audience, similar in size to the AXS audience.

“IFC” fans regular, non-expanded audience:

390,000 audience size, overall campaign frequency of 2.26

  • Week 1: 12 conversions, $2,020 spend, $168.30 CPA

“IFC” fans expanded audience:

100M audience size, overall campaign frequency of 1.31

  1. Week 1: 40 conversions, $2,005 spend, $50.11 CPA

Again, the expanded audience performed miles better right out of the gate.

Overall, the AXS expanded audience had 4.3x as many conversions as the regular audience, and the IFC expanded audience had 3.3x as many conversions as the regular audience.

That’s a huge difference!

One thing to note is that while it looks like expanded audiences appears to be a silver bullet in these examples, it’s definitely more effective under two circumstances:

  1. If your pixel has a lot of conversion activity
  2. If your product is a mass market product or service

Even though expanded targeting delivers impressions to users outside of the interest targeting, the delivery algorithm is still the same. This means that impressions will still be served to those who are most likely to convert, based on your pixel’s historical data.

So if your pixel has tracked significant historical conversions, it’ll be easier for Facebook to tap into a much wider audience to find efficient conversions.

Finally, from my experience, mass market products (like the one promoted in these tests) perform better when using expanded audiences.

This is because it’s likely that there’s efficiency outside of the bullseye targeting. That being said, I’ve tried expanding interests with niche, B2B products and it can still work.

As always, test for yourself! This is one of the simplest tests to set up and is a difference maker.

Leave a Reply

Your email address will not be published. Required fields are marked *