Transitioning from LinkedIn Lookalike to Predictive Audiences: New LinkedIn Ads Targeting

All right, welcome to another episode of LinkedIn Ads Radio. Today we're going to be waving goodbye to lookalike audiences and starting off with new predictive audiences that LinkedIn is now rolling out. So as of February 29, just a couple days ago, LinkedIn is sunsetting lookalike audiences and the ability to create new lookalike audiences. This is being replaced by the newer, newer generation called predictive audiences. So after the 29th, you're not going to be able to create any more lookalike audiences if you, if they're already created in your account and in use, you'll be able to continue using those for a while, but you won't be able to create new ones. So before we dive into this episode, some more updates from LinkedIn. So recently LinkedIn's also been rolling out their new conversions API called CAPI. We'll probably do a whole episode about CAPI at some point in the future, but it's a very useful API so that you can ping back conversion data to LinkedIn. And the, you know, the bigger picture is since all the privacy changes and deprecation of cookies in the digital world over the last couple years, tracking has gotten a lot more difficult. So CAPI is a tool and an API interface that you can use to send conversion events back to LinkedIn as another way to, you know, not doing it with cookies and digital, you know, not doing it as like an online conversion using cookies but instead sending it back end through the, sending it through the backend API as a way to log that event to LinkedIn. There's several platforms other that already leverage CAPI and then you can use tools like zapier or custom built integrations to build a whole lot more. On top of that, I love Personally I love LinkedIn Personally I love HubSpot's built in integration for offline events. It's very helpful to see conversion data. When a deal and HubSpot moves in the pipeline from stage to stage, you can easily set up and trigger events back to LinkedIn. So you can see which campaign was the last touch, which ad was the last touch before that deal move stages in the pipeline you can start to see which what campaigns and ads are creating deals, which campaigns and ads are moving deals farther along in the pipeline. Then you can also build custom integrations. So maybe if you're, if you have an app, whether a web app or a mobile app, you can create an event to ping back to the API. When a user signs up for your app, you might have conversion events for when they convert to a paid user maybe when they purchase add ons. That's all kind of useful tracking information that you want to be tracking in your campaign so that you know which campaigns are really driving those actions there. Again, a lot of that, a lot of that online is getting a lot more difficult these days with all the privacy changes and the cookies and cookies and third party cookies going away. All right, so now let's get into predictive audiences and this new generation of lookalikes. So the differences about predictive audiences is that these are new AI generated audiences. So there's, there's new tech behind it and these are dynamic, always updating audiences. Whereas lookalikes were static one time lists. Now if you've done any like Facebook or meta advertising, lookalikes are very popular on Facebook, on the Facebook platform where usually you'll upload a set, your set of customers or Facebook will, will see conversion events where so many purchases and then you can create lookalikes based off those audiences. You know, for the B2C world where there's not much targeting facets that you use there, you're more going to let Facebook make that, you know, figure out who that person is and find other similar people like that. So on the B2C World in Facebook and meta advertising, lookalikes are very popular. Over here in the LinkedIn world, lookalikes have kind of been hit or miss over the years and more miss than you know, mostly miss. Because the challenges with B2B marketing, usually we want to reach a very specific type of person at a very specific type of company and we don't want the. And when you use a lookalike, you're, you're kind of asking the algorithm to go make assumptions and find similar people out there. Whereas when you have your, you know, your ICP audience of who you know who's really a potential customer for you, that's usually a pretty well defined audience. So when you're asking the algorithm to go find similar people like that, it might be making a few hops and then suddenly your resulting list is a little bit further out and not exactly who you want to be reaching. So lookalikes have been available for a while on LinkedIn. We've seen, sometimes they're useful, sometimes they're not. There's not, there hasn't been a lot of control over them. You'll just, you'll have your original list in Campaign Manager and you'll cl create a lookalike based off this list. What it used to do was it would create A one time static list based off your seed list, but it would just create that list and then that's your list. The nice thing about predictive audiences is you have a little bit more control. Just like how in Meadow you have the slider to say I want a 1% lookalike, 2% lookalike. And then you have some more options there. You can do the same thing now in LinkedIn where you can drag that slider as how tight you want this lookalike to be created based off of or how wide you want it to be expanded to. Of course, the wider that you go, the more assumptions it's going to be making and usually a little bit further out from that initial seed list that's going to be reaching. Predictive audiences are also dynamic and constantly refreshing. You know, lookalikes were a static one time list and if you advertise that list, you might saturate that list and then suddenly that's a saturated list and then you have to go regenerate manually your lookalike list. The nice thing about predictive audiences is that they're always refreshing. So you know, people are always changing interests, they're always entering the market, exiting the market. Maybe they're changing jobs and changing companies, maybe they're getting promotions. So you know, because people are always moving around, you want your audience, your predictive audience to be automatically updating so there'll be a lot more applicable there. And then lastly is that LinkedIn's applying their new AI technology to creating these predictive audiences. So beforehand it was making assumptions just based on the data there. Now they're using new AI technology to power these. Now when you're creating your predictive audience, you have three options. You can create this predictive audience based on three different types of seed lists. You have to give it a seed list to say this is, this is the starting list. Go create the predictive audience based off this list. Your three options are number one, lead gen form submissions. So if you're running any kind of lead gen campaign on LinkedIn, that campaign's collecting leads and then you can use those leads to create a predictive audience based off that. Number two is a contact list. So you could upload a CSV list to LinkedIn and then say that these are the contacts I want to create the predictive audience based off of. You know, ideally this is like your list of customers or qualified leads that you've received. So uploading that to LinkedIn and then creating a predictive audience on top of that list and then the third is based off conversions. So maybe you have a demo request conversion on your website or you've used CAPI to identify the purchase conversions that have happened, can then use those conversion events and the users tied to those. To create your predictive audience, LinkedIn recommends that your data source be at least 300 people so that it has enough data to create that predictive audience. And then the bigger the better, because the more people on your list, the better image you're going to give to LinkedIn as to who that ideal person is, that they can go find similar people to that. And then the max list size is 3, 300,000 people. So if you have a list of a million, then you need to cut that down to 300,000 people so that you can then upload that list and then create a predictable audience based on that. Now, a tip that we've received from LinkedIn, when creating lookalikes, when you have that slider to choose how wide or how narrow you want to go, the recommendation we've received is to go up to max of 10 times the seed list audience size. So when you're dragging that resulting audience size, you want to try to stay within 10 times the seed list. If you go past 10 times, then you might be going pretty far from your initial seed list. It may be, and it might be dragging in people into your list that you don't really want to reach. And so if your seed list is 3,000 people, then the max size you'd want to go is 30,000 people. If your seed list is, you know, a hundred thousand people, then the max predictive audience size you'd want to go is a million people. So keep that in mind is whatever the size of your seed list, try to stay within 10 times for the size of your predictive audience. And then this is also something you can test where maybe you start with your first audience very narrow, and then see the performance of that. See the traffic that's coming through and the leads that you capture based off that. If those are looking good, then maybe you would create another predictive audience and then go even wider from there. So maybe if the first one is, you know, maybe 3x and then maybe try 5x and then maybe try 10x and then if 10x is still working well, maybe you try 20x and seeing, you know, the different levels of traffic and quality that you get from these audiences after, you know, you test it for a little bit. And now, so once again, talking about what you're going to see in LinkedIn Campaign Manager, you'll go to the audience section and then you'll go to matched audiences. In Campaign Manager you click Create Audience and then you'll drag that dropdown to select Predictive Audience. Then it'll have a little dialog that will pop up. You'll be able to give the audience a name. Usually I like to say predictive name it by the seed list you're giving it. So maybe it's like customers and then the target audience size that you're creating it based off of. So if I'm giving it 10,000 people, then I would say a hundred K in that audience size. So I call it Predictive customers a hundred K. And then the next option there is source. You'll then click that dropdown. Your options are contact list or conversion or lead gen form. So say I'm uploading a contact list. I'll then select that contact list dropdown. It'll, it'll show me which contact list I've uploaded already. I can check that list that I want to use. Then I can scroll down to the next section audience size and then it says where is your target audience? So the other thing to note is that these are GEO based. So I'll select either, you know, United States to start or if I was targeting the uk, I would then remove UK and then add United Kingdom as another geo here or maybe I want to add in Canada. So these are geographic based too. And if you're looking to target different geos, you'll probably want to create different predictive audiences for each geo. And then finally is the select a size slider where you can then drag that slider to the size of your resulting audience that you want to create and then that's where you want to keep in mind. If my seed audience is, you know, 110,000 people, then I wouldn't want to go past, you know, a million one in audience size. So then I'll drag this down to about a million one and then I'll click agree and create and then from there that list will then live in my predictive audiences and then I can then pick that up as an audience and use it in a campaign. Now as I jump over to campaign manager here as what this looks like in a campaign, I'll talk to you kind of two. The two main cases I've seen the most success with, we had the chance to test these out during the beta phase in LinkedIn and now that they're rolled out to everyone, we've used them a little bit more in testing. So as I mentioned, you know, in B2B marketing, usually we want to be very specific in who we're targeting. So when we're allowing, when we're asking the algorithm to go, you know, find similar people, often it's like taking, making a couple of hops from the initial targeting we're giving it. So the resulting list might be a little bit more broad of a list and a little bit further. There might be some people in there that are further from the ICP that we want to reach. So the two cases that we've seen find success with this is number one, when we have a set ICP and we've used that same targeting for a while on LinkedIn and maybe we feel like we're getting towards the end, we're fatiguing that initial audience. We've just reached those people a lot and we want to try to find other people similar to these. That's where then considering the predictive audiences as a way to find people who might, might really be part of our icp. But just for whatever reason, the job title they selected wasn't, doesn't fit under the job titles that we're targeting or the functions that we're targeting. You know, maybe those people in an adjacent function or just the way they typed in their job title just doesn't directly match what's available on LinkedIn. Or maybe they work at a company that they didn't, they didn't create the company page. So LinkedIn has a lack of data on what that company is an industry and size if they never created or never selected the right company page when adding that to their profile. So those are some opportunities of where particular audiences might be able to go find that person, even though they don't match technically our targeting criteria that we're setting. Maybe it can go find that person of who they really do match our ICP and they really would be a good customer. But they don't exactly match LinkedIn's targeting options. So that's number one. Number two is when you have a very large addressable market and maybe, you know, into the, in the figure of like hundreds of thousands or millions of people. You know, even when you're using most of the targeting options in LinkedIn, you're still getting hundreds of thousands or millions of people. That's, that's a big audience. And often when I have a really big audience, I'm like, okay, how can we get even closer to that, you know, that sweet spot? I want to start with that sweet spot first before I kind of think about expanding past there. So if my target audience is in the hundreds of thousands or millions, then I could layer on top this predictive audience on top of the ICP targeting that I'm using as a way to get even closer to that sweet spot of people in that audience. You know, if I'm creating a predictive audience of based off customers and then I'm overlaying, you know, that that predictive audience of customers on top of my ICP targeting LinkedIn, then that resulting audience is going to be, you know, a lot smaller than the millions of people that I originally have. And then ideally those millions of people are even better, even closer to who's going to be interested in my offer. So we've seen that both of those use cases have provided better results when, when using them. The first one, we were a lot, we were able to find new qualified prospects outside just the basic targeting that we were originally using. And in the second case, we were able to improve conversion rates and engagement rates because we were, we were using predictive audiences to get even closer to that sweet spot of the audience. So now the times that predictive audiences might not be a good fit for you is if you do have a very narrow target audience, maybe there's only a few companies out there who are prospective customers for and maybe just very few roles at those companies that you want to reach. You know, when you have a very narrow target audience, then you don't want to let the algorithm do some guessing. You, you instead want to, you know, specifically set that targeting that you know you want to reach and then reach those people. Because if you have a very narrow target audience and then the algorithms going to make some guesses and reach up there, then you, you probably, you might be reaching the wrong people. Along with, you know, our case study that we've seen and we've Talked about here, LinkedIn's already, LinkedIn's also posted a couple case studies showing how other companies out there have leveraged predictive audiences. Really, these kind of just talk about the stats, but these showcase how using predictive audiences allowed the advertiser to get higher engagement rates, lower CPLs. And again, the idea is that you're layering this on top to find the sweet spot inside of your audience and that more of these leads are turning into MQLs at a higher rate because you're reaching those right people in the first place. All right, so that's kind of the quick rundown on predictive audiences. So to recap, lookalike audiences are going away as of February 29th. If you've already created them, you can still use those lookalike audiences. But then the new the new feature here is predictive audiences. You can create predictive audiences based off a lead list, a conversion list, or a contact list. Then you have control on these predictive audiences as to how narrow or wide you want to go. And then these predictive audiences refresh daily. So if you have any questions on predictive audiences, always feel free to reach out. And if you find any success cases that you find them work really well, we'd also love to hear about that too, and then feature you in a future episode. All right, that's what we got for you today. Stick around for future episodes. In upcoming episodes, we're going to do some more ad format deep dives. Recently we went deep into document ads and had some good feedback about that. People asked for more formats, so we'll be going over message ads best practices and talking all about those. We'll break down video ad format and then we'll talk about all the best practices there and some examples as well. So those will be kind of future episodes that'll be coming up soon, along with deep diving into some of the new retargeting audience features that are available on LinkedIn. Those are really exciting.