WTF are custom and lookalike audiences?

0:01 What custom and lookalike audiences are

What the f**k are custom and lookalike audiences? Really valuable and interesting tools. Audiences are a super powerful component of Meta Ads Manager, well, they're actually outside of Meta Ads Manager but in that whole world, and how they can be used in your ad campaign. We've already gone over the basic audience types, where you're just creating interests and demographics, like where are you, what age range, gender, stuff like that. There's much more you can do.

First off, we've got custom audiences and lookalike audiences, the two main other types of audiences. I should preface all of this by saying this will all change, because AI is just going to take over and start automatically doing a lot of your targeting. When you boost a post and just say let it figure it out for me, that's the automated stuff, and we don't really know what's going on, it's doing a mixture of all kinds of stuff. That stuff's not super effective right now, so you have a little more control, but in time it's going to get better and better, at least we think, and eventually these will probably phase out, though the concepts will still remain. They're still relevant to you as of right now, so we're going to discuss them.

1:35 Custom audiences and retargeting

Custom audiences are audiences of people who have already interacted with you in some way, or are based around that. They're especially effective for retargeting, meaning you're targeting an audience of people who already f**k with your stuff, like they've already liked your video, or been to your website, or bought your merchandise, or listened to your music before. You want to retarget them with advertisements because you've got new stuff coming out, and you're like, hey, we know you liked our stuff, so we want to send you our new stuff. As you can imagine, retargeted audiences convert a lot better because this person already likes your vibe, so they're more likely to be into the music. Very good for new releases, really good if you have a large audience and you're touring.

Say you want to focus your advertising around promoting shows rather than promoting streams. Streams are a great way to make an initial connection, social media and streams, and then it's all about creating a deeper connection and retargeting that audience when it's time to actually go play, create a deeper connection, and then make sales at the shows. That's kind of how that world works.

2:53 Lookalike audiences and collaborative filtering

Lookalike audiences are people who Meta deems similar to people who have already interacted with you. Essentially it's saying, we know all of X, Y, and Z about the people who interacted with you, they also like all this other stuff. It's basically like collaborative filtering that we talked about earlier in the Spotify recommendation system, so it's like, the people who liked your stuff also liked these 15 artists, so we're going to find other people who like those 15 artists and put them in your lookalike audience. Maybe this group likes 12 of those artists, and this one is a little less similar, and likes only 8 of those, so the disparity you can set for your lookalike audience is how far outside of that circle you're willing to go. That's how lookalike audiences are designed, and they're excellent for efficiency when you're still trying to build top-of-funnel cold marketing.

3:56 Warm marketing versus cold marketing

Warm marketing is retargeting people who are already warmed up to the idea of you. Cold marketing is like a cold call, nobody knows what the hell you're about and you're hitting them for the first time. Lookalike audiences can be really great for that, though the way you make them is very different. So let's take a look at this in Ads Manager.

4:16 Building a custom audience from a video

First I'm going to show you how to get here: All Tools, Audiences, and it's going to load that page. Then you're going to go to Create Audience, and you'll see Custom, Lookalike, and Saved Audiences, which is stuff you've already used that gets saved every time you create an audience in your ad. So I'm going to create a custom audience first. You've got options, I can create custom audiences based on people who visited my website, based on an email list, which is a super excellent way to go if you're collecting an email list, app activity, offline activity, a catalog, or Meta sources like videos. Probably most of you are going to be using your Instagram account and videos, some might use websites, and anybody with a larger audience or a built-in customer list, that's a great way to go too. This requires putting the pixel, the same dataset, in your ads on your website, because you want to associate the same audience since they're all fans of your music.

The pixel can be on your website, your landing page, or your shop if you have a Shopify or Squarespace, all kinds of places to use it. For now let's use a video, I think this is a great way to go, watch one of your videos. I'm going to say next, and you can get people who have viewed at least three seconds, at least 10 seconds, or completed or viewed 15 seconds, you can get specific here. I want people who are really, really engaged, so I'm going to say people who have viewed at least 15 seconds, that's considered a through-play and about the best you're going to get. Then it lets you select your videos, and I've got a video pinned to the top of my profile that I'm looking for.

Okay, this is an ad that I ran, only 12 seconds, so I'm looking for the right one, 62,000 views, I'm going for this one, I want people who really watched this video. I confirm, and I want people who viewed at least 15 seconds, for the whole 365 days, and I'm going to call this Strange Angels Video Through Play Audience, anybody who viewed at least 15 seconds of that video. That's my new custom audience.

8:20 Audience retention and choosing your source

When you create your custom audience, it needs time to populate with data. I also want to point out audience retention, this is a rolling retention, meaning it's always going to show data from the last 365 days, so it will keep updating as time goes on, it's not a still shot in time, it's a rolling audience. So you can be specific, only the last 90 days, only the last 30 days, or the whole 365, and it'll just keep moving with you, like a live thing. It's best to use things that are still getting some traction, otherwise the audience will shrink over time. You can also do this off an Instagram account, people who engaged with the account or a specific post or ad, or saved a post or ad, that's another way to do it too, so it includes all of your videos. But I chose a very specific video because I want to promote that specific album again in some retargeting campaign.

10:00 Why smaller, more engaged sources make better lookalikes

So then it takes time to populate, and now it's populated, my estimated audience size is 3,600 to 4,200, which kind of makes sense, it was 64,000 views but that video was made in 2022, so only a fraction of those are still relevant now. That's fine, you actually want a smaller targeted audience if you're creating a lookalike from it, in fact I'd go even smaller, the more engaged your source is when you're creating a lookalike, the more valuable it's going to be. So an email list, or a purchase list, a list of emails of people who bought your stuff, that's your ideal source for creating a lookalike, because you're extrapolating from them, you want the most valuable people to you, the people who bought your stuff or came to your shows. If your goal is to make a lookalike audience for more cold marketing, go for the most engaged people you can get, use the highest quality data to create your custom audience source, and then create your lookalike from that.

11:16 Building the lookalike and reading the percentages

So if I go in and say create a lookalike audience, I could create a new source, or use other sources, and we're going to use the through-play audience I just created. Now I need to tell it where I want those people to be, so let's use the United States, since we're running ads in United States, Canada, and I think Australia, we broke those out earlier in some of the audience videos. Let's create three 1% groups, and go all the way out to 3%. What that means is it's going to create three different audiences, the 0 to 1%, that's 3.4 million people, 1 to 2%, that's another 3.4 million, and 2 to 3%, another 3.4 million. The way to think about this is the further out they are, the less like your source they're going to be, and the closer in, the more like your source they're going to be.

Here's another visual for this: this guy interacts with your ad, that's me, we all like this guy, he's a great guy, so you're like great, he's my custom audience, so now I'm creating a 1% lookalike audience based on him. It might look a little something like this, not exactly him, but it looks a hell of a lot like him, except his hand's really weird, but that's great, we love AI, right? So that's your closest to the source, your 1% lookalike. If we go out to the 2% lookalike, it's getting a little weirder, just further from the source, and even the way I'm making these is like Meta is making them, first I'm dropping this image into an AI image generator and saying make me a copy of this image, and it makes this, most similar to the source, then I say make me a lookalike of this guy, and it makes me this, and then for my 3% it looks like this. As you can see, the further away you get, the higher the number, the further it's going to be from the source, so the harder it's most likely going to be to get conversions based on it. But it's still probably going to outperform a completely cold audience, though you have to A/B test these things.

13:36 Recap: custom for retargeting, lookalike for reach

A common move is to create a lookalike audience based off a core source group and then compare it to cold audiences that you define specifically in your audience controls, for example creating a list of artists like Spotify, people who like these artists and these record labels. So now let's go back and quickly finish creating these lookalike audiences, create audience, and you'll see it's creating three new lookalikes, Australia, Canada, US, 2% to 3%, 1% to 2%, and 1%, those are my three lookalike audiences, all populating. So a quick recap on this video: the two types of audiences we talked about are custom audiences and lookalike audiences. Custom audiences are great for reaching people who have already interacted with you, for retargeting, and lookalike audiences are great for reaching people similar to people who have already interacted with you, and they're great for efficiency on running more reach campaigns to grow your audience.