Marketing Insights

Lower Facebook Ad Costs, Improve Results Via Lookalike and Split Testing

Facebook’s ad platform is driven by intelligent algorithms that can deliver highly profitable ad results. Facebook has a ton of native data that it draws on readily to help your ad campaigns get off to a strong start and deliver the results you need. But there are things you can do right now to accelerate … Continued

Facebook’s ad platform is driven by intelligent algorithms

that can deliver highly profitable ad results. Facebook has a ton of native data
that it draws on readily to help your ad campaigns get off to a strong start
and deliver the results you need.

But there are things you can do right now to accelerate and greatly
improve Facebook ad results that lead to the following:

  • Significantly Lower your cost per result
    (whether that’s lower cost per lead, lower cost per click, or less expensive video
  • Extend the mileage of your budget so you
    can reach more people and drive more results
  • Get your ad campaign off to a stronger start
    more quickly

How can you do this?

Facebook Lookalike audiences.

In this post I’ll share the results of four different
split tests
using Lookalikes vs non-Lookalike targeting with a Facebook
ad campaign. Two of these split tests produced a lower cost per lead, and two
did not.

For you as an advertiser, run your own tests and you will
likely uncover improved results. But as always…test, your mileage will vary. The
bottom line is, Facebook makes it very easy to run a statistically valid split
test automatically so you should be leveraging that feature.

What Are Lookalike Audiences?
Lookalike audiences are audiences of Facebook profiles that are similar to and based on an existing audience.

An existing audience could be a customer email list you have
or an audience defined as those that have interacted with a video on Facebook
from a prior ad, or visited your website, for example. Facebook refers to these
lists as a Custom Audiences.

So, your Lookalike audience is based on a Custom Audience
that you create or define. You then tell Facebook to use this Custom Audience
to find potential customers and prospects that match the profiles, interests,
and behaviors from the seeded Custom Audience list you created.

With a Custom Audience, you’re giving Facebook a big hint at
what to go look for when it creates a Lookalike audience.

You can create Custom Audiences based on many different

For example, A Custom Audience can be created based on visits to your website or even just specific page URLs on your website  – here’s how that looks inside Facebook’s ad manager:

With website visits, you can create your audience of all

website traffic or just specific page URLs and specify the duration people will
stay on the list up to 180 days.

Here’s another example of a Custom Audience based on engagement with any videos you’ve promoted – watched 25% of your video, 50%, 75%, etc. The possibilities here for marketers are endless.

Facebook Custom Audience Video Engagement

And here’s the full list inside Facebook of Custom Audiences you can create.

Complete List Facebook Custom Audiences

Once your Custom Audience is created, you then create your Lookalike audience – here’s what that looks like in Facebook. Under “Select Your Lookalike Source” is where you would add your newly created Custom Audience.

Creating Facebook Lookalike Audience

Once your Lookalike has been created you can then utilize it
in your latest ad campaign.

But for this post, I am focusing on using Custom Audiences
to create Lookalikes to help with your lead generation and prospecting Facebook

The quality and source of your Custom Audience is the
most crucial factor
in creating a lookalike audience that will deliver
results. Garbage into the list and garbage out for the Lookalike.

For the case study results below, we used a list of actual
current customers and past Facebook leads that expressed interest in past
offers. This type of a list is perfect for creating a Custom Audience that can
be used to create a Lookalike after all these are people we know who are already
customers and spend money with the brand or have expressed interest in an

The campaigns we ran were Lead Ads designed to capture lead
information (i.e. name, email, and phone number) in response to a limited time
offer (LTO).

The Seed Lists Used To Create The Lookalike

The seed list used to create the Custom Audience that was
used for the Lookalike was comprised of a national list of past customers and
Facebook and Google leads we had recently collected – the list size was 45,853.
This list was loaded into Facebook as a Custom Audience which was used to
create our Lookalike audience.

Creating The Split Test

Once you start building your campaign or at the ad set level,
you’ll have the option to create your split test. Facebook gives you options in
terms of what to test – in this case, we chose Audience.

We split test our Lookalike audience that we created with our normal interest and demographic targeting that had proven to work well for the last 30 ad accounts we used it in.

Creating Facebook Ad Split Test

Results of The Lookalike Split Test

Our campaign objective for the four split tests was using
Facebook’s Lead Generation campaign objective – we ran Lead Ads to capture
leads of potential prospects right in Facebook that were opting into a special
offer. The business was a retail brick-and-mortar service business.

The first two tests showed good improvements with the Tampa
split test lowering the cost per lead the most at close to 50%. The two other
markets – Charlotte and Tulsa – showed higher cost per lead results with the Lookalike

We tried to keep all other variables like creative and messaging the same in the split test to remove variances.

Results Facebook Split Tests Lookalike Audience

The bottom line is that your mileage will vary. These split
tests are easy to run and cost almost nothing in terms of your effort to run
them – so test the results and you likely will deliver far better results with
your Lookalike audience.

Facebook will also tell you the results of the test – below is an example – it let us know that there was an 84% chance we’d get the same winner again with the same settings.

Facebook Split Test Message

Some Common Questions:

How big of a seed list do you need to create a Lookalike?

A Quora
“evangelist” states you need a seed list of between 1,000 – 50,000 but can be
as little as 100. You need the largest list you can create…more data points
will help Facebook’s algorithm work.

How do I know when to stop the split test?

Facebook will tell you – it will give you percentage likelihood that if you ran the test again that you would see the same results. It’s up to you when to stop the test – in our case we stopped the split test when we had 70% or higher that the same results would occur again.

If you have past customer lists or past customer data like website
visits that you have captured using Facebook’s pixel that you know are valuable
past customers, absolutely you need to test these Custom Audience lists with
Lookalike campaigns…those lists are a data goldmine for your company.

If you’re a restaurant and capture emails from free wifi logins, a dentist with a patient list, plumbers, beauty products and retail companies, contractors or home remodelers…anyone with a customer list. Start testing these lists by creating Lookalikes and improve your campaign results.


Google Ads vs Yelp Ads: Top 12 Reasons Why Google Beats Yelp

Back to Marketing Insights

10 Social Media Marketing Mistakes To Avoid