Audience Testing in Theory

Let's talk about audience targeting and testing. This is more of a theory thing, so we'll power through this.

As we know from our little previous discussion, testing is always happening. And so, we need to have a process to deal with that necessity.

What is Audience Testing?

So let's talk about audience testing first.

So for audience targeting and testing, we have a couple of basics.

So first and foremost, we have a few buckets of top-of-funnel audiences that we can play with.

We don't test middle funnel audiences.

We don't really test bottom-funnel audiences. In that sense, we try cold traffic. The top, middle, and bottom funnel audiences are almost exceptionally well defined. And I would say 90 to 95% of cases. There are minimal variations. Whenever I'm saying audience testing, it always means top funnel.

So if there are a few buckets of top-of-funnel audiences we can play with.

We can talk about lookalike audiences.
We can talk about interests, AKA detailed targeting.
We can also speak about broad targeting
then talk about different countries and regions, which, in theory, will be different audiences.

So here's the thing.

And this is just from a lot of ad spend working with many other accounts, niches, etcetera.

Audience testing is pretty much anything but common sense.

In theory, a 1% lookalike of purchasers will be the best possible audience you would ever test in practice.

It's pretty much rarely the case.

What would end up happening instead is that it might do well in some accounts and other audiences do better in different accounts.

So it's just kind of the way it is.

If it knows that wasn't the case, then there would be no such thing as audience testing.

And right now, you'd be lying on a beach because we would know the best audience, and there would be no need for audience testing, which would be awesome.

We haven't solved that problem just yet.

One important thing to remember about why this is the case is that Facebook will often use creative and account data to target just as much as the audience definition we provide Facebook.

So that's why no targeting audiences can work well at scale MIS parenthesis there at the end, but that's why at high scaled accounts, 5k 10 K plus per day, usually just using no targeting and only just using ages, for example, can work well.

So Facebook does not just target based on our audience definition. It will do that, but then it will further subset based on that, based on the creatives, the copy, the video image, the account history, etcetera, etcetera.

So more targeted slash and more narrowed is not always better. Typically, the Facebook algorithm will like bigger, broader audiences, audiences with many people in it, 10, 15, 20 million people plus now. That doesn't always align with what we want to be doing. Sometimes our total market size might only be 500,000 or a million. The problem is that the more we try to restrict, the more we try to narrow down, and the more we try to have a targeted audience; Facebook hates that, and they drive the CPM up and charge us a lot more to do that.

And often, the economics of it don't factor out. So typically, more target is not always better for you in theory. What you want to do is keep everything wide and broad there, of course, is a nice balance to strike, which we're going to talk about. But in most cases, more like lasered in a smaller audience is not always better. It's just going to be more expensive. A quick little rule is that all potential audience sizes should be above a million for the top of the funnel to be scaled into again. You know, are there edge cases for this? Yeah, sure. So if you're, you know, if you're watching this and you're saying, and you're saying yourself, okay, but there are only 250,000 people in my audience, or there are only a million people in my audience is total.

So how will I do a bunch of these different tests completely get it? Don't worry about it. What you want to do is you won't have large audiences, and then again, your creative copy, etcetera, will do most of the work there. Okay. So let's talk about, look likes in general high level, I, this, I don't want to get too, too far into the weeds of how Facebook does this, but because it doesn't matter, to be honest with you. It does, to an extent, but more so, it's just, how do we use this rather than how are the deep intricacies of this working? So what happens is Facebook will take a seed audience that we give it, whether it's a customer list, an email list, or pixel data, but Facebook will take this audience.

And we upload that to Facebook and say, Hey, Facebook, find us a bunch of other people who are most similar to this group of people I gave you. That's a look at its base form. 1% look like purchasers is often again the best audience in theory. Still, it is almost never the best audience in practice after multiple, multiple, millions of dollars in testing. Okay. And so, again, it, it's one of those things where even though I know that this is the case that we always want to incorporate some at least, 1% look like of purchaser testing, it is often a good audience, but it's usually not the best, even though it should be. So don't sleep on lookalikes of leads, ads to cart, video viewers, site visitors, page engages, et cetera, et cetera. Again, if you ever start to run out of ideas for look-like ideas, just check out our custom audience list.

So like that massive document that we have with a bunch of different custom audiences and different look likes that you can create from those. So, in general, with a lookalike, you want to start at 1%, and if you find gold within one of those audience tests, then you would continue testing with higher percentages. So if 1% lookalike Add to Cart does well, your next audience test could be 1% to 2% lookalike ads. You can also play with date ranges here. So a lookalike created from the purchasers of the past 14 days can be way different than all-time customers. And why is that? While longer date ranges will take into account holidays, sales events, etc. And we may or may not want this. So, for example, imagine that you had an offer where you were selling your product, which is typically priced at $20, but you were selling it for $1 for two weeks straight. Most likely not happening, but as an example, if you did that, and then you created a lookalike based on the past 14 days of purchasers, you know, raised your prices back to the normal price.

After those 14 days, Facebook will essentially be finding a bunch of people who want to buy a $1 product or another $20 product. So that's where it isn't good. So think about it before you create your day range, typically again, 180 day or all-time customer list, 180 days, 90 days, etc., those are going to be completely fine to do. But again, think about this. So the seed audience of 100 is a minimum. So it needs at least a hundred people in size for Facebook to create a lookalike audience from 500 to a thousand probably is better. So if you do have a hundred, you can try it. But it's going to be 50-50, whether it produces a quality audience or not, okay, following interests in detailed targeting. So this is what I like to think of in three different layers.

So three different orders of interest, and I call this tangential targeting.

So we have first-order interest, second-order interest, and third-order interest.

Let's take an example here.

At a high level, interests are people who have expressed interest in this thing in Facebook's detail targeting section.

It lets us show ads to people who have expressed interest in this thing, whatever it may be. As an example, let's take a higher-end hair care brand.

So the first order of interest is going to be directly related, related to the product. So competitors, in this case, dove and head and shoulders. People who are interested in hair care, people who are interested in shampoo, that's our first order. So highly, directly related to the product. The second order of interest would be things that people are closely tangentially interested in the product.

So skincare, beauty, right? Long hair, beautiful hair, hair, etc., All of these things can, of course, be good, related interests. Now the third-order interests are where things get interesting, which is tangential. So, the framework from this is what else is my perfect customers interested in. Questions can you ask to tease this out? What do they do on weekends? Where do they shop? What stores do they shop at? What do they buy for gifts? What TV shows do they watch? What clothes do they wear? All these things, what they sufficiently large brand are going to have an interest for you to target on Facebook, which is exactly what you want. So, in this case of higher-end hair care, whole foods, right? Whole foods, of course, being a little bit more of an upmarket grocery brand, will probably attract higher quality customers.

Speaker 1: (