It's easy to spend hours deciding which interest audiences you should be targeting. After all, you know your brand's customers best right?
We can guess some attributes of our customers. If we're marketing a new app for example, then chances are that it'll appeal to younger demographics, interested in technology. There are plenty of things we can't guess about our most likely customers; perhaps it turns out that they enjoy ultimate frisbee, or that they're likely to keep birds as pets.
It's hard to manually pull all these insights ourselves. Facebook's lookalike audiences provide a handy way of automating this process though.
Lookalike audiences take some 1st party data, that's data that you have, and it goes and finds the people on Facebook who most closely match those users. Because Facebook knows so much about your customers, and about everyone else on the platform, it can be an extremely effective way of finding people that are likely to engage with your brand.
To create lookalike audiences, you'll first need an audience that you want to create a lookalike of. This is referred to as your source audience. There are two types of source audience that you can create, dynamic and static source audiences.
A dynamic audience is one which updates automatically. Good examples of dynamic source audiences which you can create are:
The first of these is the best source audience if you're looking to drive conversions. Bear in mind though, that the more people in your source audience, the better Facebook will be able to find people that look like those in your source audience. A larger source audience means a higher quality lookalike audience. For this reason, you should only use people who've converted on your site as a source audience if you've got enough of them for Facebook to be able to learn from.
But how many people is enough? 100 is the minimum, but you should aim to have at least 1,000 people in your source audience. If you've tried creating an audience of people who've converted on your site (check out my retargeting guide for how to do this), and it's below 1,000, it could be worth choosing one of the other options on the list above.
If you are keen on using past converters for your source audience, bear in mind that you can select any date range for creating your lookalike audience. Your source audience could be, for example, people who've converted on your site in the past year. Unless your customer profile has changed significantly in that time, providing Facebook with more data points will always be beneficial.
There will be some cases where using a dynamic source audience won't be appropriate. One example would be if your brand has been running for years, and you've collected plenty of customer data, but you've only just installed your Facebook pixel. In this case your pixel won't be able to create any audiences of past converters, so you can use your own customer data instead.
This can be done by creating a custom audience via a customer list. To do this, we'll head to the audience library, and click Create Audience then Custom Audience. Select Customer list as your audience source. You'll see a few options here on how to create your audience:
I'm going to assume you've chosen not to use the MailChimp option (if you have, follow the flow and skip to the next section).
Click to accept the Facebook audience terms, and you'll be taken to this screen:
Here you can upload your customer data, which should be in a csv or txt format. Your data can have any of the identifiers listed at the top of the screen. The most important ones for you to upload are Email and Phone Number, but you should upload as many as you have to hand.
Once you've added your data, name your audience and click Next. You'll get to preview your data, and check that each identifier is correctly mapped. If this looks good, hit Upload & Create. You should see your customer list audience appear in the audience library.
Now that we've created our source audience, creating the actual lookalike is fairly straightforward. In the audience library, click Create Audience and then Lookalike Audience. This will open up the lookalike creation screen:
Once you're here, you should:
Once you've filled in all your fields, your screen should look like this:
If you're happy with your settings, click Create Audience. Your lookalike audience should now appear in your audience library. You can now create new campaigns and ad sets which target your lookalike audiences.
So you've created your lookalike, applied to an ad set, and hopefully it's performing well. Where do we take it from here? How do we expand on this?
One good option is to test whether we can bring more volume by expanding the size of our lookalike audience. To do this we'll create another lookalike audience with all the same settings (same source, same location), but this time we'll set it to target the next 1%.
To give an example, if our original lookalike was a 2% lookalike audience of purchasers in Italy, we'll create a 2-3% lookalike of purchasers in Italy. We can do this by adjusting our audience size like this:
This audience will contain the 3% of the Italian population that most resembles our past purchases, but excludes the 2% who most resemble past purchasers. Doing this means that it won't overlap at all with our original 2% audience.
Once we've created this audience, we can apply it to a new ad set in the same campaign as our original lookalike audience. This means we'll have a campaign with ad sets targeting:
We'll let them run alongside each other, and see how they perform. We can expect the 2% lookalike to perform best, as it contains people most like our past customers. The real question is if the 2-3% will perform well enough to justify keeping it going.
If it does perform well enough, we can add the 2-3% lookalike audience to the ad set which contains the 2% lookalike audience. The benefit of this is that we'll have one ad set effectively targeting a 3% lookalike. Fewer ad sets is always better, as Facebook's optimisation algorithms learn at the ad set level.
We can keep iterating on this process, next by testing a 3-4% lookalike audience, until we find that creating larger lookalike audiences doesn't lead to improved results. By testing lookalikes in this way, we can ensure we're making the most of the power of lookalike audiences.