Leverage the Power of Lookalike & Retargeting Audiences
Let’s start with Lookalike and Retargeting audiences. People are always excited and eager to get cold interest and behaviour based to convert.
However, if you want to build profitable Facebook® Ad campaigns your audience arsenal will have to include Lookalike and Retargeting audiences.
Suitably sized lookalike and retargeting audiences will almost always outperform cold interest and behaviour based audiences. Yes, the recent Apple iOS 14.5 update is having an impact on our ability to build these audiences but contrary to popular belief, it is not destroying Facebook® ads.
There is a lot of misunderstanding about iOS 14.5 and the impact it is having. In April 2021 Apple released an update to its mobile operating system which powers all iPhones and iPads. Once installed it upgraded the software of the device to version iOS 14.5.
One of the features of this update was to block tracking pixels, including the Facebook Pixel, from reporting on the actions that iPhone and iPad users took on 3rd party sites (ie your website).
This specifically prevents Facebook from adding people with the iOS14.5 update installed to a custom audience for you to retarget with specific ads at a later stage, however, all is not lost.
The fact is it’s not blocking Facebook® from tracking who’s taking those actions, it is only blocking the reporting of these actions. Facebook® still knows who is taking the actions on your website based on the pixel. It still knows who those people are. So your lookalike audiences are not going to be impacted in the same way as your standard retargeting audiences.
In addition, there are several ways to get around the iOS 14.5 update. One method is to use a program like Zapier. In my mind, Zapier is an essential tool for the modern marketer.
If you are not familiar with Zapier you can think of it as a concierge who has a set of operational instructions who looks for certain events in one software program and when a specific outcome occurs performs a defined action in either the same program or one of 3,000 other integrated programs.
In our Facebook example, your pixel might be blocked from reporting any action taken once they visit your opt-in page, but with Zapier, either based on data received through your opt-in page builder, shopping cart, or email platform you can report back to Facebook the event outcome and add that person to a custom audience and follow up with retargeting ads.
This very simple to set up workflow will ensure your retargeting audiences will operate at a similar level to pre-iOS 14.5 times. It’ll also give the option to build your lookalike audiences from the people who’ve taken the action that you want on your website.
While we are talking about iOS 14.5 I do have to mention the impact on reporting within Facebook®. While I have provided a solution to building custom audiences for retargeting one of the big impacts is the fact that your reports on Cost Per Lead or Cost Per Acquisition will be flawed.
Please make sure that you have established a reporting system outside of Facebook that combines data from Facebook®, your opt-in page builder, shopping cart, or email platform to provide you with an active assessment of your true campaign metrics.
When it comes to lookalikes and retargeting, the most important thing is to start your audiences from day one. Even before you start running ads, you need to plan and set up the retargeting audiences that you want to leverage in the future.
I always start broad with an audience of everyone who visits your website. From there I define ever more specific sub audiences based on the pages they visit and what that tells me about their preferences and likelihood of taking follow-up actions.
This means building retargeting audiences based on people who visit specific pages or who have an interest in specific areas of your website.
Based on this you can retarget these audiences with ads for specific lead magnets or offers based on their past behavior and preferences.
Another valuable audience is your blog visitors. If somebody’s visited your blog, that means they’ve engaged with your content. You can now make assumptions about their knowledge of “the problem” you solve, how you solve it, and where they are on their journey.
The first step in setting up your Ads for success is to create custom audiences for each of your landing pages, opt-ins, thank you pages, upsell pages, and down funnel steps. I can’t tell you how often I have undertaken a campaign audit only to find that this simple but essential step has been skipped.
The further down the funnel and closer to the ultimate conversion point the more valuable an audience is. However, this value is often offset by the size of the audience as fewer people graduate to the lower levels of a funnel.
The first level of the audience could be people who interact with your ad on Facebook®. This will be the largest retargeting audience but also the weakest.
Level 2 may be people who visit your opt-in page. Now, something caught these people’s attention on Facebook sufficiently enough that they took action and visited the opt-in page.
However, this audience includes lots of people who decided not to opt-in but also a percentage of people who do go on to take further action and go to the next phase of the funnel.
Level 3 may be people who visit your opt-in thank you page. That is people who have completed the opt-in process and have been directed to the success page.
When you build a Lookalike audience you ask Facebook to go out and find a percentage (1-10%) of the population of a specific country or countries that look most like your base audience.
The closer the audience is to taking the ultimate action you desire the more specific the data Facebook® will have to base their Lookalike audience on. However, the fewer the people in the audience the less data Facebook® has to work with and the more difficult it will be for Facebook® to identify the defining traits that will make people more likely to take the action that you want.
As a result Lookalike audiences work best on a base audience of at least 1,000 people. When you are starting it might take you some time to get to a point where 1,000 people have opted into your lead magnet. That is the reason why you might start with a higher level audience to start with and then move to lower levels as the size of your audience grow.
Like most things Facebook® Ad related, it is always best to test, however. I have had great results with Lookalike audiences of under a hundred people. In one specific campaign, a 5% Lookalike of a lead magnet thank you page which had 86 visits outperformed the original interest-based audience by a factor of two meaning that the leads from the Lookalike audience cost half the price of the interest-based audience.
Larger Audiences Are Outperforming Smaller Audiences
In a slightly counterintuitive twist, larger audiences are trending toward outperforming smaller audiences. The old metric was that your perfect audience size was somewhere in the region of 1.5 to 2 million people. And if you had multiple audiences that were three, five, six, seven, 10 million, you’d split those up into smaller segments, because that was the optimum size for ad set budget optimization.
But now what I’m seeing is that the bigger the audience’s size, the better results they’re getting.
Facebook® has recently made target expansion on audiences mandatory. Up until recently, you could choose whether to switch target expansion on or not.
Target expansion allows Facebook® to go outside the audience you define while respecting any specific demographics like age, sex, and location if it thinks it can get you more results.
In announcing this Facebook® said that with detailed targeting expansion on, especially in interest-based audiences, there is a 37% decrease in cost.
Not only that but they found in website custom audiences that lookalike expansion gave a 17.3% decrease in costs. And with custom lists, a decrease in costs by 10.1%.
What Facebook® is doing in the background is, it’s spending X% of the budget you specified on your ad sets, but it’s also siphoning off a little bit of money and spending it in other places where it thinks it could potentially get better results for you. And then if it gets good results, it spends more and more money on these audiences.
There’s absolutely no question or doubt that the Facebook® AI is not in a position to 100% select our audiences for us, but this development is pointing to a real statement of intent for the future.
All this to comfort you as to why it is a good idea to make your audiences as big as possible, feed the Facebook® AI as much data as possible, and let it do its thing.
Personally, I always keep the AI honest by testing multiple versions of the Lookalike audience.
Let’s say you’ve got a thank you page from a lead magnet. You build the audience of people who visited the thank you page, and then you ask Facebook® to build a Lookalike audience of the 1% of the population of a specific country that looks most like the people who have visited this page.
Intuitively you may think that the 1% of the population that looks most like the people who have visited your thank you page will perform better than a 3% or 4% or a 5% audience, but actually, in many cases, it doesn’t and the wider audiences outperform the smaller ones.
When it comes to audiences, remember ABT – Always Be Testing and the most important part of testing is making sure that you’re collating and analyzing the best possible data and that every decision you make is backed up by that data.
Split Audiences Into Dedicated Ad Sets For Testing
Often when auditing an underperforming ad campaign, I see multiple interests and behaviors and lookalike audiences, all mixed into one Ad Set. While this is certainly meeting the requirement of bigger is better from an audience perspective it is far from helpful when it comes to optimizing your Facebook® Ads.
Essentially what that means is, you don’t know which audience is working, which audience isn’t working, which audience you should drop, and which audience you should put more budget into.
Always separate your primary interest, and behavior, retargeting, and Lookalike audiences into separate ad sets.
Initially, you’re going to have to control how many Ad Sets you have. If you are only starting, your overall campaign budget may be lower.
This will pose a problem because to exit learning each Ad Set must hit 50 instances of your nominated objective or conversion events that you set.
When launching a Facebook® Ad campaign your first goal is to ensure each Ad Set exit learning. It is only after exiting learning that your costs will even out and you can begin to optimize your campaign.
So let’s say you have five Ad Sets, and each one of those Ad Sets has to hit 50 conversions each week, which means 250 conversions in that week. Now that’s fine on bigger budgets, but let’s say if it’s costing you $5 per conversion, now we can see that’s a total weekly spend of $1,250 per week or $150 per day.
Now a $150 per day budget is tiny for a lot of advertisers but if you are starting out facing an unknown cost per conversion and an unknown Earning Per Lead this may be a little daunting.
I normally set Course Creators and Membership Site Owners who are new to Facebook® Advertising up with two to three Ad Sets, featuring three Ads per Ad Set in each.
Once the campaign is up and running this gives you enough variance to assess winners and losers, and then introduce new audiences overtime when the losing audiences are dropped.
You always want to give an audience five to seven days to settle down to give Facebook® an opportunity to work out the best people in the audience to display your Ads to.
And remember, as you initially refine your Ad copy, opt-in page and funnel audiences that may not have worked in the past could work better with the revised copy so it is always worth retesting audiences over time. With different audiences, you’re going through a continual process, like refining your ad copy, refining your opt-in page and funnel.
Remember your audience messaging match is hugely important too. Small tweaks to your Ad copy based on the specific audience may have a big impact on the end result. This could be one or two words in the Ad but always be conscious of the people you are talking about and how you could potentially tailor your Ad copy for them.
One of the key metrics that will tell you whether your audience and messaging match is your CPM (Cost Per Thousand Impressions). If you’re getting a high CPM, it means that the message for your audience is off or that the language of your Ad could be approaching a breach of Facebook’s Advertising Policies.
The goal of Facebook’s Advertising Policy is to safeguard its users’ experience. They want their users’ feed to be a safe and enjoyable place to be. It’s what keeps users coming back to the platform. Your Ad copy may not breach a policy but if it isn’t contributing to a positive experience, or if Facebook® doesn’t feel like it is adding value to its users’ feed you will be penalized with higher CPM.
If your goal is a profitable Facebook® Ad Campaign having a solid audience testing strategy is key. To appropriately test audiences it is essential that you split them into separate Ad Sets and track their performance.
When to Use Ad Set Budget Optimization (ABO) and When To Use Campaign Budget Optimization (CBO)
When you are launching a Facebook® Ad Campaign one of the first decisions you are faced with is whether to use Ad Set Budget Optimization (ABO) or Campaign Budget Optimization (CBO).
During the early stages of a campaign when you are testing creative and testing audiences you should always start with ABO.
ABO lets you set an individual budget for each one of your Ad Sets which forces each Ad Set to spend the budget assigned to it. This will give you a solid basis to assess each audience individually and make a decision on your best-performing audiences.
As mentioned earlier in this post, we need each ad set to exit learning as quickly as possible As a result when you are starting, especially with low budgets, you want to limit your campaigns to 2-3 ad sets with a minimum of 3 ads per ad set.
Once you get your initial ad sets up and running you then enter the optimization phase, testing, honing, refining the creative and audiences until you reach the point where you are happy with your conversion rates and your CPL or CPA
At this stage, you’re going to want to scale your campaign. This is when you move from ABO to CBO.
Campaign Budget Optimization hands more of the power and decision-making process back to Facebook®. This is not wise early in the campaign as there are too many variables at play, but once you have figured out your winning combinations it’s time to hand the reins back over to Facebook® and give them the flexibility to find those who will be most likely to take the action that you define.
When you move to CBO, you have to make sure that you give Facebook® even bigger audiences to work with. A common initial CBO campaign will feature one super audience containing all your audiences from ABO combined into one.
This one super audience will be used in 4-5 ad sets with the exact same ads in each. This seems a bit strange I know, but in effect what Facebook will do is venture into different pockets within the audience to figure out who will be most likely to convert.
Once you exit your learning phase, now there are a number of different strategies you can pursue.
You could go with one ad set with all your audiences and let that scale. Or you could possibly go with a super broad audience, where your only definition of the audience is a demographic like age and location. Alternatively, you could go with multiple % lookalike audiences: a 1% lookalike audience, a 2% lookalike audience, and so on with the 3%, 4%, 5% lookalikes.
That gives you plenty of options in terms of scaling because the more finite you make your audience, the higher your CPM will be. By making it super broad and moving the budget to a campaign level, you’re giving Facebook® much more control.
2 Ways to Scaling Your Ad Set Budget Optimization Campaigns
In order to get to the level of ad spend that makes it viable to move to CBO, you are first going to have to scale your ABO campaigns. Let’s say you begin testing with a $50 budget a day, and things are working out. You have identified your winning creative and you are now looking to scale.
You have no idea how many times I have seen people bump their ABO budgets from a daily budget of $50 all the way to $200, only for the ads which were doing so well up to this point collapse and die.
You can only scale your ad spend by 20% of the budget every 72 hours. So if you’re starting with a budget of $50 a day, it means that you can only increase the budget to $60 in the next step. If you exceed this daily budget you’re likely to reset learning and there is no guarantee that your ads will exit learning in the same way as they entered learning.
72 hours after this initial change you can then increase the budget again from $60 to $72. Now, most people, if they get their ads up and running, they’re uncomfortable with moving at that slower pace.
For the majority of people in my audience you may be coming up to a launch, a big promotion, or maybe you have a webinar just around the corner and your ad campaign has a specific deadline so the slow and steady budget increases can be frustrating. Well, the good news is that there are several ways around this and there are two approaches to scaling which I like to use.
The first is vertical scaling, which is what we’ve been talking about thus far in this section, and that is increasing the budget of an ad set. The second is horizontal scaling, and that is scaling through the addition of extra audiences.
One of the best ways to scale horizontally is through the addition of larger lookalike audiences. Let’s say you have a lookalike audience of visitors to a thank you page, or lookalike visitors to the opt-in page itself, depending on how long the campaign has been running and the size of the base audiences.
If you are currently using a 1% lookalike of the thank you page, one strategy is to duplicate the ad set and use a 2%-5% lookalike audience and set the new budget to $150 or $200.
If we needed more spend we could undertake a similar process using a 5%-8% lookalike
Another approach is to introduce new interests or behaviours. Bringing in additional interests and behaviours is always tricky because the audience might not necessarily respond in the same way and you are introducing a new variable to the campaign.
In addition to the horizontal scaling strategies we also have a vertical scaling strategy to get around the 20% every 72-hour rule.
This is to simply duplicate the ad set keeping everything the same but increasing the budget of your new ad set to your desired level.
In effect, this will put the new ads in direct competition with the old ads targeting the same audience, but this has a high success rate.
One downside to this strategy is that your new ad set will start the learning phase again but Facebook will be able to leverage the previous data accumulated. A big upside is that the original high-performance ad will continue unaffected.
When this strategy is shared with people, many worry about competing against themselves. What you have to realize is you’re competing with hundreds of thousands of other advertisers who are targeting your audience. So the little increase in your budget is not going to have a discernible impact on your overall CPM.
One word of caution with this strategy is to watch duplicate reporting between the two ad sets. As the same people are in both audiences if someone opts in through one ad having been shown an ad from the other ad set it is likely that they will be double reported in both ad sets.
But remember what I said earlier in this post, you need to have a way to independently verify the Facebook® reporting and a second method for calculating your Cost Per Lead and Cost Per Acquisition.
Never Ever Kill The Golden Goose
Number six is one that most Facebook® Advertisers encounter and they normally experience it at the very worst time possible. And that is that you never, ever kill the golden goose.
If you have an ad that is performing, never make any changes to that ad or to that ad set. It is always best to duplicate your ad into a new ad set, or even into a new campaign if you’re switching objectives.
This rule applies to any edits outside of your 20% budget increases, anything that has to do with changing images, videos, headlines, body copy, or audience changes.
When people optimize their ads, normally they are cutting their losers and introducing new creative. But, never forget that a good optimization approach is the incremental improvement of your best-performing ads.
And that is when it gets dangerous. Never be tempted to make changes to the winning ad or ad set because that’s going to drive it back into learning, and unfortunately, as weird as this may sound it may not come back out of learning in the same way.
So, yes, this might cause some untidiness in your campaign and possibly some headaches from a budgeting perspective, trying to match up the budgets and getting everything to balance, but it protects your CPL until you have other ads that are outperforming it.
Today we’ve presented a “what’s working now” report from the Facebook ® ad trenches.
I know there is a lot to unpack in this post but hopefully what we have outlined above will save you hours of frustration, and lots of $’s as you launch and scale your profitable Facebook® ad campaigns.
So, let me know what challenges you face with your Facebook® ads, whether you’re just getting started or you’re a seasoned pro looking to scale your ad campaigns.