Using lookalike audiences to reverse the marketing funnel and generate quality leads
As marketers, we got used to letting social media platforms (and Facebook in particular, a.k.a. Meta) do our work for us.
We let these platforms follow the customer journey from our ads all the way to conversion. We let them watch. We let them learn and we let the algorithm optimize and target the proper audience.
The algorithm did everything. It was comfortable and easy.
At the very beginning, Facebook used to share that information with us and we could learn at the same time as the algorithm learned. We used to be able to analyze our audience, our followers, what they liked, what age they were, what gender, marital status, what other websites they visited, and what other pages they followed. We knew as much as the algorithm did.
But then that information was no longer available. Yet we didn’t care because the algorithm was doing its thing and we were getting amazing results. So we got comfortable, too comfortable.
Fast forward to April 2021 and the iOS 14.5 release
The world for marketers using Meta imploded a bit.
For some, it imploded a lot.
Users had to be asked for permission to be tracked across apps and websites and 95% of them decided not to give such permission in the U.S. (84% worldwide).
Since then, social media platforms have had terrible visibility into what is happening to people that click on an ad. Once they leave Meta that is pretty much it!
Meta has done some work to provide estimates. But in my experience things like landing page arrivals or even conversion attributions are far from the real numbers (thanks to Google Analytics and UTMs for the backup tracking ability).
Interest-based targeting is one of the few tools we have left.
So the theory is to feed the funnel with cold leads at the brand awareness stage so that they flow through the funnel and convert without barriers.
There is one problem: because algorithms still have trouble determining positive interaction from negative interaction and, for that matter, they have trouble understanding context – engagement and interest with a particular brand may not mean that they want to be approached by that brand.
Interest-based marketing is a good starting point but misses the mark many times.
Researchers analyzed the accuracy of Facebook activity on their interest-based ads and found that almost 30% of interests Facebook listed were not real interests. That means that if your ad is based on the list of interests, you could miss the mark about 30% of the time.
This study is the first of its kind and has a relatively small dataset, but looking at comments and the engagement generated in interest-based ads I have run, I see the biggest percentage of confused and unhappy comments on this ad set, so NC State is onto something here.
If you got to this point of the article, you might be re-thinking your life choices as a paid social media marketer.
However, there is something still very useful in the platforms:
Facebook may not have as much information about your converters as it did before, but you – or your clients – do!
Instead of feeding this theoretical funnel to cold audiences, let’s go to the end of the funnel and find people like the converters.
The process is similar in all platforms:
- Get your seed list of converters.
- Create a custom audience with this list by uploading it to your social media platform of choice.
- The platform will match the information to what they know about each person in the platform (most commonly email or phone number).
- There are minimum matches needed for this list to be valid and each platform has its own rules for this.
- Once the custom audience is created and valid we can generate a lookalike audience where we tell the platform “find people with similar profiles” to the people on this list.
By creating lookalike audiences we are taking the funnel and tipping it upside down. We start at the bottom and generate a list of cold audiences so similar to our current converters that they may be almost considered warm audiences.
We are now using the social media platforms to help us create personas based on data we know is accurate and then targeting them.
Platforms know a lot about our behavior within the platform. They are not perfect, but these platform-generated personas are way more accurate than inferred interests.
Because you are not targeting one interest, one element, that will be irrelevant 30% of the time. You are targeting a group of elements, interests or platform behaviors. That substantially reduces inaccuracy.
After doing A/B tests between interest-based audiences and lookalike audiences I can tell that I have had results improve up to 40% for some lookalike audiences. Sometimes the results are as small as 15% but I will take any improvements and efficiency I can get when optimizing my ads.
Wouldn’t this give too much control back to the algorithms?
Are we setting ourselves up for the same scenario we had pre-iOS 14.5 by letting algorithms run our paid media? Yes and no.
- There is a little bit of trust we are giving back to the algorithms, but now we know not to put all of our eggs in one basket. We know that interests identified by Facebook are still 60-70% accurate, so knowing your audience’s interest is very valid, even if we miss the mark a little bit.
- Audiences shift, their interests change, and we should be moving with them. Can you tell me your audience looks the same now as it did in 2019? My recommendation is to use lookalike audiences as often as possible but complement them with interest-based ads and continuously A/B test their efficiency.
Consider your campaign objective
Sometimes lookalike audiences are good at converting but may not be as good at engagement.
In one A/B split test I run, the interest based audience had 30% higher cost per click but the rate of positive engagement was double. This audience wasn’t converting, they were spreading the message.
We not only need audiences that follow the funnel path to conversion effectively, sometimes we also need audiences that cheer us on and help us spread awareness.
Please consider this before using lookalikes
A lookalike audience is based on a custom list (seed list), and this list should only be created with data you own and have permission to use.
Check each platform’s policies regarding custom lists to understand this better.
If people unsubscribe from your communications, have a plan to update your lookalike audiences.
If people do not want to hear from you, then why would you want to advertise to somebody with the same profile?
Remember: Platforms change over time, so we must evolve with them to stay relevant and sometimes that means going back to basics. Good luck out there.
Watch: Using lookalike audiences to reverse the marketing funnel and generate quality leads
Below is the complete video of my SMX Advanced presentation.
Opinions expressed in this article are those of the guest author and not necessarily Search Engine Land. Staff authors are listed here.
New on Search Engine Land