Facebook Lookalike Audiences Best Practices

Key Takeaways

lookalike audiences are one of my favorite prospecting tools on the Facebook network there's a lot of algorithmic learning that goes into how a look-alike audience is put together that can really reach beyond what we as advertisers can use with the behavior interest or demographic targeting that we have available in the platform that said there are a lot of ways that you can make a look-alike audience more impactful more successful by utilizing some pretty key strategies with your route audience so that's what we're gonna talk about today I love space so I'm gonna use a pretty loose space metaphor if you will about trying to find planets there in what's called the Goldilocks zone we're effectively trying to find other planets that are similar to Earth for whatever reason we want to study them or we want to try and inhabit them someday that sort of thing so scientists are going out trying to find different planets that behave really similar to earth they have similar environments they have similar temperatures all that good stuff obviously a lot of them here in the Kepler system and that's basically what we're trying to do with our target audiences on Facebook we're giving Facebook a route audience and then having them go find things that are similar to that audience like I said with the demographic information behaviors interests that sort of thing we can go out and find folks but it's almost just like being an amateur astronomer being in your backyard with your nice little telescope whereas Facebook has access to the equivalent of the Hubble telescope to be able to find new users that behave like the target audience that you want so we want to make sure that we're leveraging all the information that we can get out of the platform look-alike audiences can be created off of pretty much any route audience that you can make in the platform so that can be people based on engagements whether they've engaged with your page engage with your post that sort of thing website audiences any of your retargeting audiences if people came to your website whatever pages they've been to on your site they've converted any of that sort of thing any based retargeting audience you can make same with page audiences again if they've engaged with your page if they follow your page within the Facebook platform and then also custom audience uploads if you've uploaded a list of your customers or subscribers to your newsletter anything like that you can also create a look-alike audience off of those folks that have been match to that custom audience upload process the Facebook look-alike algorithm is very much a data processing unit and the route audience is where we can impact how successful it will be at processing the list we give it and finding new users who behave similar to it it's similar to that route audience that we started with so what we want to do is make sure that whoever we base a look-alike model off of is a very strong patterned list of users they all have a lot of characteristics in common they're very specific but not too small we need to make sure that we have enough data going in but not so much data that a pattern is indiscernible in that audience when we upload an audience that has a strong pattern list of behaviors and markers the Facebook algorithm will have a lot of success of having a strong model come out the other side it'll be able to find good patterns find a lot of users who behave very similar to those folks on the Facebook network and put them all into an audience for you if we upload an audience that doesn't have any patterns in it it can't find any recognizable distinctions between the users in the audience honestly anything can come out of it you might get a really good audience but it also could just be terrible coming out the other side so the biggest thing we want to do is make sure that the audience we upload has strong patterns and we can make a good look like model off of it here's something that I think most people don't think about each of the models that you use can be different even if they are just slight iterations of each other a common list that people tend to upload is all of the people who convert on their website or all of the customers that they have within a platform that will then produce a look-alike of those users which is great that's fine you're gonna get whatever percentage model you put in is going to come out of that but something that I think people don't pay as much attention to is depending on how many users you have within that list it might make sense for you to have a subset of those users who spent let's say over 500 dollars on your e-commerce website right they spent more money than a lot of other people on the website and we want to find users who behave like that now you might think that all of the users that it's going to go find are going to overlap 100% what they look like if you're all converters but that's just not right there could be a lot of different users out there who how have different markers that are more in line with those people who spent over $500 then the entire bucket of people on your website so yeah there might be a short a small portion of it that overlaps with the look-alike of all converters but you also might have a large portion of people that you wouldn't have had if you didn't take that smaller subset list of your converters who only spent over $500 and now you've got a better patterned list the look-alike audience size resulting from it is going to be the same because you always have to do it if a percentage model based on the population of that country that you're targeting but the patterned list should be different based on what you uploaded in that route audience when you're utilizing a route audience think about meaningful segmentations within your customer base or how people engage with your website or any other marker that's in here maybe you want to target the people who hit a thank you page but also maybe target people who just saw your pricing page on certain websites a pricing page visit could be an indicator that that person is really close to wanting to make a purchase because they're starting to compare different pricing models between either your product that you have or maybe comparing you to your competitors if you're uploading a list of users maybe if it's like a b2b solution maybe you want to segment by job title it's not it doesn't necessarily make sense to target everybody at your customers companies maybe you want to target just the people who have a Yorty title that means that they might be the ones who could sign the contract maybe you want a segment returning verse one time purchasers on your e-commerce website because a lot of people probably purchase one time but if you want more people who are purchase multiple times create that segmentation in your list and upload that and see what look-alike model comes out of it the only thing that we need to keep in mind is finding a happy medium with our route audience size this is the only thing that can kind of start to hamper how sigmund did you get with the different models you have so going back to my super fun space metaphor the Goldilocks zone is basically an area that like Goldilocks and the Bears is not too close to a star that it's so hot and inhabitable but it's also not too far that it's too cold and inhabitable it's just right if you will what Facebook suggests for your root audience upload size is for it to be anywhere from 10,000 to 50,000 users to then create a look-alike off of I personally think that that bottom number the 10k is way too high if you have that that's great that that's excellent use that a lot of companies we work with have less than 10,000 users in a list some of them have as low as 500 users in the list and I've seen those perform well as long as the audience has strong patterns that Facebook can latch on to and find similar users to it if you start to have audiences a customer list let's say that you're uploading and there's 50,000 users in it stop and take a second and think how can I naturally segment this audience to be not one list of 50,000 users but how can I make two or three lists of people and now we've got 225 thousand user lists or maybe three lists where a couple of them are fifteen thousand ones twenty thousand something like that and now you've got two or three models that you can start to create a look-alike audience off of see how they compare to each other performance wise and then optimize accordingly so think about that when you start to set up your campaigns and step your route audiences as to what patterns are likely to be found within these user lists and how is Facebook gonna go find those people on the network and make them target it in my account so when you want to use look-alike audiences think about those two things that we talked about right how can I find meaningful segmentation in my route audience and make sure that it's still a large enough size that Facebook has enough users to base a look-alike off of but not so many that the pattern becomes muddied and it can't find any users that actually behave like that and it becomes the equivalent of just targeting the internet because there are too many people in there balance out those two and I guarantee you'll start to see better results with your look-alike audiences thanks for watching our video make sure to subscribe to the paid media pros channel to see more videos