Scaling Your Subscription App with Meta Ads – Marcus Burke, Independent Consultant

David Barnard:

Welcome to the Sub Club Podcast, a show dedicated to the best practices for building and growing app businesses. We sit down with the entrepreneurs, investors, and builders behind the most successful apps in the world to learn from their successes and failures.

Sub Club is brought to you by RevenueCat. Thousands of the world's best apps trust RevenueCat to power in-app purchases, manage customers, and grow revenue across iOS, Android, and the web. You can learn more at revenuecat.com. Let's get into the show.

Hello, I'm your host, David Barnard. And my guest today is Marcus Burke, a subscription growth consultant who specializes in unlocking Meta as the primary growth channel for your app. On the podcast, I talk with Marcus about the past, current, and future of Meta ads, tactics to scale subscription apps on Meta, and why you should probably exclude younger audiences in your targeting. Hey Marcus, thanks so much for joining me on the podcast today.

Marcus Burke:

Hi, David. Super excited to be here. I listened to almost every episode of this since 2020, one of the few podcasts I listened to very regularly. So yeah, really, really excited for this one.

David Barnard:

Awesome. Well, it's so good to have you on and long time coming. I've been following your work as well. You've done a lot of great posts on LinkedIn. Anybody who doesn't follow Marcus on LinkedIn, go to the show notes or search his name Marcus Burke, and find him on LinkedIn because he's constantly sharing all sorts of great stuff.

So I've been meaning to have you on the podcast a while, and I'm especially excited about the topic today. So we are going to talk about Meta ads. So Facebook, Instagram. You probably know all the different, I don't even know all the different placements that you can have with Meta ads.

But a lot of people have used Meta ads over the years successfully, unsuccessfully. It's one of the more talked about channels in subscription apps. And so the first thing I wanted to discuss is why, why Meta ads? Why is it such a talked about channel? Why should subscription apps consider it? Should all subscription apps consider it? So let's just start with that. Why Meta ads?

Marcus Burke:

You should definitely consider it. I don't think there are many apps out there that can just say Meta isn't for me. One of the big things of course is that Meta's family of apps is just gigantic. I think they recently shared again that they have around about 4 billion monthly active users across all of their apps. So as you already mentioned, Facebook, Instagram, but then also all the sub placements.

David Barnard:

Do you know if they dedupe that? Is that actually 4 billion unique humans, or is that 4 billion, maybe there's a billion overlap of people who use both Facebook and Instagram? Do you know?

Marcus Burke:

I think it's deduped, but it also includes WhatsApp of course, which isn't all... I think there's ads by now on WhatsApp, but you don't run them through Meta Ads Manager. So that of course kind of blows it up a bit. But still, even if it's just 3 or 2 billion, you can pretty surely find your target audience on this channel.

The other important thing is that Meta has a very, very strong algorithm. They're good at finding your audience for you. It got a bit more complicated with [inaudible 00:03:22] and scan, and I'm sure we'll talk about that later on. But still, Meta's algorithm is really unchallenged. They had their SDK in many, many of the apps out there for years. They had their pixel on almost every website for the last 10 years. So they sit just on a ton of data, and they really know how to build these customer profiles and use the signal that you're sending them to find your perfect customer. And that's really where the magic happens so that you don't have to care that much about how do I target and who is my target customer, but you can really find them in collaboration with Meta.

David Barnard:

Yeah, that's one of the interesting things mean if you think about... And this is why I think it was important to start with their scale, and the reach, and why they matter. Because when you think about those two things combined, three, 4 billion people, whatever it is fully deduped or people who actually see ads, it's in the billions. And so then you combine most people that would ever subscribe to your app are probably using one of Meta's products that they do see ads in. And then Meta is really good at targeting your specific product to the niches that it would appeal to. So maybe a super-duper, duper duper niche app would just not have the scale to work on Meta ads. And then there's all sorts of other reasons with funnel and other reasons that Meta ads might not be performant for an individual app.

But broadly speaking, any human that might subscribe to your app is probably on Meta, and Meta is really good at finding them, so it kind of makes sense. This is why they make tens of billions of dollars of profit a year, right?

Marcus Burke:

Yeah, it's a strong combination. And what I also like about the channel is that in the end, all of it is native advertisement. And you can really go onto these placements, you can see what those ads look like, you can see what other content is on these placements. So you get a pretty good understanding of what can I do here? How can I be creative to really insert my product onto them? While with a lot of other channels they are display networks, you don't know what's going on. There's fraud, which doesn't happen on Meta. So really it's a safe environment. You have a good understanding of what your final ad will look like, and I think that's just a very big advantage as you are trying to validate your app in the beginning. Because if you don't know the context in which your ads were shown, then you can never be sure if the data that you gathered was in the end even valid.

David Barnard:

And then another thing too is that the ads on Meta are actually kind of good. I mean people kind of talk about it a little bit, and then there's of course privacy circles of all ads are bad or whatever. Some of the Twitter noise in the communities I'm in, we'll talk about ads being bad and all this kind of stuff.

But when I get on Instagram, I kind of like the ads. I actually had to remove Instagram from my phone finally, because I was getting enticed by too many products, because it's almost like ads as content that I actually enjoyed the ads and would almost go to Instagram not just to see the content from the people I follow. It's a great product discovery engine in a way because they're so good at targeting you, the specific things that you would be interested in.

I think that's another thing. It's like the two biggest players in ads in this context actually make a ton of sense, because Google is search intent. When you go search something, you're telling a text field on the internet that you have a problem and that you're looking for a solution. And so it makes sense why Google can monetize ads at that level.

And then with Facebook, it kind of makes sense how well they do with their own monetization and for helping apps monetize, because the ads are actually really good.

Marcus Burke:

Glad to hear that marketers, we're doing a good job here. Yeah, I think looking into many ad accounts every month, I think ads could still be better. I still see a lot of things where I feel like, okay, and this is more of a very classic banner ad or something and it's not really well thought through. But definitely, it seems to work for them that they have been increasing ad load more and more and users are still enjoying the experience. So it feels like people get value out of the ads and not just out of the native content, because by now I would say 50% of what you see on those feeds is ads or then influencers posting native content, but then they do product placements. So it's really much of a product discovery engine as you mentioned.

David Barnard:

And then the web is the opposite, to your point earlier. Banner ads, Taboola ads on news sites, it's like the ad experience on the web broadly is actually the exact opposite of that, including Google's display networks and things like that, that they do. And then especially the way Google... And we're not going to talk much about Google today, but the ways that Google forces you to do these broad campaigns that might end up in these really poor display ad networks, versus the kind of higher intent. With Meta, it's like you know your ad is going to show up in these quality higher intent places even versus Google. That actually brings me to the next question I wanted to ask you is that, how do you think about Facebook in the broader channel mix for a subscription app as they're scaling up?

Marcus Burke:

To me, it would usually be one of the first channels to launch, because it's a discovery channel, you know you can get some volume there. And it's a bit more of an audience you will target in the future than for example, search. Search is super interesting because you know the quality behind it. So if you are trying to validate your funnel, your pricing strategy, then Apple search ads might be an interesting one just so that you know you're buying qualitative traffic and don't have to worry about, "Was my ad good? Did I reach the right audience?" Because if someone searches for a weather app and you have a weather app, then you can be sure it's the right user.

But it's not as scalable, it's super competitive, so it's going to be hard for you to compete with the big players that might have a better funnel, higher pricing. So Meta really stands out because you can have a competitive advantage through good creative and through mastering the channel. And that's why I like to start it early on.

And most apps care about iOS a little bit more than Android, which also gives Meta a competitive advantage over Google Ads because Google tends to be very strong on Android where they own the ecosystem, it's much easier for them to track performance. While on iOS, they've rather been struggling in the recent years.

David Barnard:

All right, so with the broader context out of the way, I wanted to dive into history, which may seem odd but I do think is important in the context of Meta ads, because there has been a lot that's changed over the last four to five years. And of course everybody listening to this podcast will know what I'm alluding to and that's app tracking transparency.

And so before app tracking transparency, Meta was the go-to channel. Because with a level of deterministic measurement with the IDFA, it was just like magic for so many apps. And I think it's maybe even lost on folks who don't follow the industry super close, but a lot of the apps that are big today are big specifically because of the arbitrage of them having scaled so quickly, so efficiently on Meta ads pre ATT, and then ATT happened. So tell me, I mean maybe I've already given enough of the history, but what's your perspective on the pre ATT? And then we'll get into how ATT changed and then how things are evolving.

Marcus Burke:

Definitely it's a bit unfair of how much easier things were back in the days compared to today. So Meta really had one of the strongest device graphs out there, so they really built super strong customer profiles based on these IDFAs, and as I already mentioned, they just tracked everything. Everyone had the Meta SDK, everyone had a pixel on their website from them. So basically, they were tracking the whole internet and were able to create these super strong customer profiles.

With ATT that changed making things a bit harder. Maybe a small anecdote like most of my pre ATT time I spent in the gaming industry working for InnoGames, a big German strategy games developer. And I followed the whole development from browser games into mobile games, and we started Meta as a channel somewhere around 2012, 2013 when I joined the company. And we scaled it super rapidly, always jumping on the newest products when they launched [inaudible 00:12:00] event optimization for the first time, when they launched value optimization for the first time. Those were always magic moments in the end. It was really a gold rush. We turned those campaigns on, we didn't have to care about targeting anymore. They were just looking for the right customers for us, and we were within a month doubling our budget again and again while overachieving our return on investment targets. And it was crazy.

We had times where management told us, "Please don't spend so much money. We know it looks amazing, but we're scared." That's how it worked back then. And that just really changed with ATT, and you have to do your homework a lot more closely these days. You have to dive deeper into your funnel, into creative strategy, and do a lot more qualitative work. While back in the days it was really data driven. You could push out just a ton of creatives, Meta would find the right audience for them, and you didn't even have to care why did this work. It's like, "Okay, it works. We're just going to do it over and over again." And that's kind of the biggest change that happened with the implementation of ATT.

David Barnard:

So as ATT rolled out, it feels like there was an especially tumultuous year, 18 months, two years where things just weren't working. What were you seeing during those times or what lessons can you take from that time? And then it does seem like, and we'll get to this, that things have started improving. But tell me about for history's sake, what was that 18 months to two years like as somebody spending a ton of money on behalf of clients and publishers trying to make it work?

Marcus Burke:

Thankfully actually around that time, I moved from a UA focus into growth focus. So I had some time where I could dig into the topic before I was spending bigger budgets again, which was great. But the main issue was there was a lot of uncertainty really around what's going to happen and what this is going to look like. I think the whole communication by Apple wasn't great. And for the longest time, it wasn't clear what this would do to your data models, what this would do to Meta's algorithm. And those things really only were figured out as it was live. And then we saw a massive drop in performance, everyone scaled back, Meta's stock took a hit, and their advertising income took a hit. And only from there on, things recovered.

So I think probably there would've been ways to make this transition a bit smoother, but this was just what we had to deal with. And I think advertisers really had to relearn how to make predictions on how valuable this channel actually is because you don't have down funnel tracking anymore. Most apps only track a trial start within their ads manager, and that's all they see, but they don't know what's happening afterwards. Do these people converge, do they renew? How long do they stick around? Which puts Meta in a bit of a tough position, because it's always been a more expensive channel in the end on the upper funnel. CPMs are high, there's a lot of pressure in the auction, very competitive. But the value came with that great targeting and the good down funnel performance. So conversion rates were often higher for this traffic, which is something you now weren't able to see and you also were unsure if Meta is still doing a good job, and that of course creates a lot of uncertainty. And you don't want to spend a million a month without knowing is this going to be profitable in the end?

And that was the main issue that took one to two years. I would say now that advertisers are in a more comfortable position, again, they have to find you their data models so that they can predict value based on that trial event and all the other information that they're gathering from users in onboarding now.

David Barnard:

At what point did you start seeing a shift? Because I feel like maybe it's you specifically that I've been following on LinkedIn for a while. It feels like it's not just you, but that I've just been hearing more and more folks say in the last, I don't know, 12 months or so, and then especially again specifically for subscription apps that it's working again.

So how much of that is Facebook having invested so much into these targeting algorithms and non-deterministic measurement, and those things? How much of it's like publishers getting smarter, like you were talking about? What are all the contributing factors to it starting to get better again? And then when did you start seeing that shift happen?

Marcus Burke:

To me, it also feels like there was a shift definitely in the last year again, while also already prior to that, the best advertisers definitely didn't take until a year ago. They were already ramping this up again much quicker. And I think it's a combination of these factors really. I mean, Meta of course keeps improving the algorithm and gets better at modeling performance again based on the signals that they're getting from scan, but also from much more videos on the platform now. I mean, they're scaling reels like crazy. Most of my accounts is actually one of the biggest placements by now, and it functions a lot more like TikTok. So they have a lot of engagement signals, watch rates, which they can feed into the algorithm. And based on that, I think they got pretty good at modeling performance even from these very upper funnel metrics.

And on the other hand, also advertisers just got better and better. And I think by now, everyone has covered the groundwork and they know we need, "Where did you hear about a survey in onboarding," so that we can create sample data on channel level and look into differences of organic, versus Meta, versus TikTok. Most advertisers are asking for age and gender of users, which is super helpful because Meta always reports you their data on and gender level. And if you then and your product also apply that level again, then you can basically crosscheck and look into... I don't know, I'm seeing in my app that age group 45 to 54 is 5X more valuable than 18 to 24, so is my CPI of 18 to 24 actually 5X lower or is there a mismatch? And these are the kind of things that I think first really needed to be dealt in, and then data models needed to be improved so that advertisers really gain a lot more trust again in what they're spending on these platforms.

David Barnard:

Yeah, I think that's a really good summary. And again, I think this is all going to be super helpful to folks who've heard, "Things are changing." Is it time to come back to the platform? Or if we didn't find success in 2021, what can we do today differently that we'll find success? And I think understanding this kind of context of why it was so good pre ATT, that kind of tumultuous time, and then why it's starting to get better I think is super helpful for people to kind of frame their decisions around when and how to be ramping up Meta ads.

So the last thing on this topic though before we move on is scan three, and we'll talk about scan four because they haven't fully ramped that up yet. But how much do you think scan three has helped, and then how has it helped with Meta getting better at finding these users again?

Marcus Burke:

Most advertisers are really not happy with that scan three can do still compared to what things look like beforehand. You really have a very limited timeframe that you can track users. You have privacy thresholds, meaning you need a certain level of installs per day so that you can actually get a postback with your [inaudible 00:19:15] attached to it. And also, there is postback delay, so it usually takes two to three days for your postbacks to arrive, and then you need to match that to what you've been doing two to three days ago. So your spend levels from three days ago need to match the [inaudible 00:19:30] that you're receiving three days after, which isn't great. I mean, you want your data to feel like it's scientific and you can rely on it, and this always feels a bit scrappy. And it's not great based on what you did before.

But still scan three I think has allowed a lot of advertisers to spend a lot more money again, and especially sub apps were a bit in a comfortable position because their payment usually happens in onboarding already. So you start a trial event there, meaning you have a pretty good value indicator early on in the product experience that you can feed back to that model, so that Meta at least gets a postback within these two to three days. While other advertisers were hit a lot more strongly like talking about gaming before. For games, usually monetization happened not as fast. Still good payers would already purchase something early on, but a lot of, especially in the strategy games genre, the whole business model was based on some people paying a lot. And finding these people got a lot harder because signal equality isn't as strong. They might not pay that big amount of money on the first day, meaning you can't track it, and hence Meta can't find you these people anymore.

So any product that was a bit more niche with purchase patterns that were later in the product basically had a hard time, and hence also product development and how you treat your funnel played a big role here. So really, if you want to succeed on scan, you have to work a bit deeper into product and think about, what are my conversion events going to be, what can I feed to these platforms so they still make smart decisions for me? And luckily sub apps just were in a good position for having that trial and onboarding.

David Barnard:

I want to dive into best practices. I'm sure folks listening are like, "All right, now tell me what to do." But before we transition to that, I did want to get any other thoughts around how to think about subscription apps, why they work so well on Meta. And then a lot of the noise around ATT was even around D2C, direct to consumer kind of physical products and things like that. So any other top of mind differences? So when somebody reads, "Meta is not performing," but it's from a D2C perspective, how do they frame their thinking around that compared to running a subscription app and trying to scale a subscription app on Meta?

Marcus Burke:

I mean, I would actually say that D2C has an easier game because most of what they're tracking is still happening on the web, and so you have Facebook conversion API that you can send your pixels through directly your events through directly from Shopify. There's an easy integration, and then you are out of this whole ATT mess. If you're trying to send traffic into your app for a bigger, I don't know, fashion brand or something that is trying to optimize for traffic, that might be a bit tough. But as you are on web, I've seen actually that D2C advertisers had a much easier game.

And also, a lot of the recent uptick I think was actually from D2C and the whole Advantage+ initiatives that were launched from Meta, so they really automated more and more of what's happening in the app account for the better or the worse depending on what you're looking at. I think for D2C, it works nice because they are sending a purchase event, which is their business goal in the end, and the algorithm knows this person has purchased potentially even for how much. While if you send a trial event, the algorithm can't be as strong for you, because it's not directly correlated with business value. A trial only converts in 30 to 50% of cases, so they might be optimizing for the other 70%. Basically, I've seen stronger performance in these D2C accounts,

David Barnard:

D2C is maybe better, which that's interesting. I hadn't heard it framed quite that way. So again, really good to get that perspective. And then games and delayed monetization is maybe where things are still going to struggle. Any other thoughts on subscription apps and how they vary against other types of ads on Meta?

Marcus Burke:

Yeah, I think the conversion event is really the most important one. So as I mentioned, your trial isn't a real business value for you in the end, and trial conversion can vary massively on an audience level. As I already mentioned on each group, it's usually big differences. Older people will just convert better because they have deeper pockets.

So that's where it's special on the subscription front, because you need to be much more aware of what's happening after the trial and guide the algorithm in the right direction while a D2C advertiser will probably get along with using more of the automation because the algorithm is just much closer to their actual business goal. I mean they also have recurring purchases and people that sign up for the email list and don't, so I don't want to say it's easy game for them or they're less skilled. But definitely there are special complications with advertising and subscription app and optimizing for trial.

David Barnard:

All right. Well let's move into best practices then. So we've talked about all the challenges. How do you as a subscription app best use Meta to find those payers that are actually going to become subscribers?

Marcus Burke:

I mean, I think it's pretty commonly known that you want to use a consolidated account structure so that you really try to maximize the amount of signal you get on each of your ad sets. So in the end, as the algorithm is supposed to help you find your users, you want to maximize the signals that it has to basically do that. And in the account, you have a campaign level, ad set level, ad level. And if you set things up very granularly and you might be running, I don't know, 10 campaigns with 50 ad sets in them, you would be scattering data across all of these.

So the algorithm will have a very tough time in actually using those signals to find your users, while if you just run one campaign with one ad set in it, all that data will be used to basically fine tune targeting for you. So you always want to find the right balance of where do I need granularity to guide the algorithm, and maybe you want a country split in your campaigns because you know the US is going to be much more profitable than Brazil. So you want to have these splits but then still don't add too many so that there's a lot of signal under each of your campaigns.

David Barnard:

That would then maybe be something that as you scale up... So if you're just starting out and you're spending that first 40K to start figuring out if you can get the algorithm trained, that's when you want to be almost completely consolidated. And then as you're spending more and more and more, then maybe you do have more flexibility to branch that.

Marcus Burke:

Pretty much. So try to start, very consolidated. And then as you're adding spend, you can add additional campaigns and ad sets based on the levels that you want to target separately. And this is also going to really help you pass privacy thresholds which are in scan three, still a big pain for especially smaller advertisers because you need to reach 88 installs per day per scan campaign ID, and actually Meta doesn't even tell you where they apply a scan campaign ID. It's definitely on ad set level, but it might even be more narrower because they receive better data on each of these IDs. So they might be creating in the background two IDs for Facebook and for Instagram. Even though you target these in one campaign, it might be two IDs. You don't know. So basically, you want to make sure a lot of installs are coming through on each of your campaigns so that privacy thresholds are passed and then you actually get postbacks, so that Meta see your trials and can optimize for you.

So start consolidated. Don't go in there, set up different interest targetings, and think that you know better than the algorithm, because in the end you don't.

David Barnard:

So are there any scan specific best practices, or at this point does me handle most of that for you? When you install the Facebook SDK, is there customization and best practices on scan, or is that mostly just handled by the SDK and by default?

Marcus Burke:

I mean, I would say one best practice is to optimize for your trial event and make sure that's basically the highest value one in your scan schema. Other than that, there is not much you need to handle in terms of how you set up your campaigns. You have a choice between tracking through scan and through AEM. It's really not that you need to take special care of a scan campaign by setting it up differently.

What you need to consider is how you evaluate your data. Because as I mentioned before, your events are going to come in later. So actually, when you kick off your campaign for two days, you're not going to see any results. And that can be scary, especially if you're spending a lot, and it says you've spent 5K and you have zero installs and zero trials tracked. But that's just due to the fact that postbacks are coming in delayed, and you need to make sure that you are basically evaluating campaigns that way and always match data from two days back with the right spend level.

Additionally, I would say also give Meta a little bit more time, because they don't get these events super quickly, so you shouldn't be evaluating results after three, four days because then only one day of data came in, meaning the algorithm wouldn't have been able to really do a good job for you yet. So give it some time, at least a week, I would say before you start evaluating performance. That's kind of the biggest thing to take care of when you're handling a scan campaign.

David Barnard:

Because of that, is a three-day free trial a better tactic even if maybe you're not going to convert quite as well but you're going to get better data on a three-day trial than a seven-day free trial, or do both kind of work well as long as you know that and are factoring that into your data measurement?

Marcus Burke:

In the end right now in scan three it's not going to make a big difference. I mean yeah, for conversion, there might be one or the other will be better for your app. But you're still only going to track your trial start event and then you're going to have to look into trial conversion in your product data. And of course, having a three-day window means you're getting a conversion quicker, meaning you can more quickly evaluate your results. But that's really the only benefit here. And I would say if a seven-day trial is actually converting 5, 10% better for you, then rather go with that and don't reduce performance just so that you have results quicker, because over time your modeling and over that data will become strong enough that you can also predict results with a seven-day trial.

David Barnard:

What are your other best practices on the basic level, and then we can get to more advanced stuff?

Marcus Burke:

With the subscription app, you're always going to optimize for a trial start. Generally, try to get you conversion event as close to a business value that you can. For the subscription app, it's mainly going to be that.

Don't target too narrow. These algorithms usually need a lot of reach so that they can use data to then find the right audience for you. If you apply a ton of targeting restrictions on top, then usually you pay a premium for targeting more granularly, while performance not necessarily is better. So I would usually with a new app rather start with a broad targeting, choose your core markets. And I do usually exclude younger users below 25, sometimes even below 30 depending on app, because of what I mentioned earlier, that the conversion rate from trial to paid is so different for older and younger users. And while you are still early and you don't have a good understanding yet of how this is going to convert, I'd rather start a bit more conservatively, use only audiences that I know have a good chance of converting. And only once I've gathered data and get a better understanding, I might broaden that to 25 plus.

But really, the age group below 25 is really neglected by most of the advertisers because they're usually very cheap to buy, because they use social media more often and they don't have high purchasing power. So CPMs are low. Meta really loves spending on them because they're so cheap, and they're going to create a cheap cost per trial, but then the conversion down funnel is going to be very, very poor. So I haven't seen anyone really correct this issue yet of how can you spend towards these audiences and still convert them with, I don't know, a good discount strategy or whatever you might be apply in the back to it. And yeah, 25 plus is kind of the best practice I would say in the industry.

And one other thing is I already mentioned like Meta also pushing for Advantage plus more and more, which is their automation features. You by now we'll see probably 10 different things that are called Advantage+ in the UI, which gets very confusing at times. And it's basically what you mentioned with Google before, that they are trying to make you opt into everything that maximizes reach on their end so that they can sell away any kind of inventory.

For some advertisers, it works very, very well, especially if your conversion event again is close to your business goal. But with a trial just be careful. I wouldn't say it's not a strategy to use, but you want to know the algorithm works before allowing the algorithm to do everything for you. So I usually start on manual campaigns, set up things myself, have separate ad sets per country, and then exclude younger age groups, while only later in the lifetime of an account I might layer on these fully automated campaigns because then I know, okay, my conversion rate for these users is going to be that much lower, so I'm bidding less aggressively on them to have a cost per trial that is cheaper so that the equation works out. Be skeptical of all that automation. Meta likes to kind of append recommended labels to everything, but it's not really recommended for everyone for sure.

David Barnard:

You kind of mentioned you don't want to go super narrow on your targeting, but then you mentioned you do want to target via age. So you're saying there is some level of penalty in the auction in a way. You're going to just have to pay more in the auction to get these older users, but that's the one place where being more granular in your targeting really tends to pay off?

Marcus Burke:

Yeah. With age definitely that's the biggest lever. If you're struggling with trial conversion rate, then target older people and you're eventually going to bring it up. If you need good numbers for your next VC round, then that's definitely a good trick to make those numbers look better.

David Barnard:

Are there any more advanced techniques that you start to get into as you scale that people should be aware of?

Marcus Burke:

There's a few ones I would want to highlight that I tend to see in accounts where people just stay a bit too basic and don't optimize on that deeper level to really fully leverage that channel. One of those I would say is creative diversity, which is linked with placement performance and really optimizing for each of these placements.

As you already mentioned, there's a ton by now. I also don't know all of them by heart anymore. But really, there is totally different audiences behind any of these. With some it's really different apps like Facebook versus Instagram. But then also on Instagram, the people that use the Instagram feed compared to the people that use Instagram Reels are totally different. Reels, it's much closer to a TikTok audience. It's younger, lower intent, while the feed is a bit more qualitative but also usually more expensive.

So you really want to find out, what are the placements that I can find my target audience on? And from there, go really deep into optimizing your creative for these placements. So you want to actually use these apps. If you find Facebook is still working for me, then use Facebook. You probably haven't been on there in a while, but it changed a lot. People still use it, and you want to know what does this placement look like, what's the typical content people are sharing here, and how can I insert an ad here that makes sense that is close to the same experience so it's not just a banner ad like on a website, but you want to really work with this native feeling.

And I feel that's where often people just look at performance on a too high level. They look at, "Okay, this creative has had a cost per trial of 10 euros and the other one has 15, so one has done better than the other." But you really want to dig into, "Okay, where did I deliver with these?"

And Meta shares all that data. You can always break down creative performance on a placement level, on an age level, on a gender level. So you can learn a lot really about where do I need to run these ads to get good performance in my app, and creative strategy is really where it's at then so that you tailor things to the right placements that are going to serve you well.

David Barnard:

Got you. And then of course once, you've run the ad, so much of the potential performance for that ad ends up being what then happens in your app. How do you think about the funnel optimization from onboarding, to payroll placement, to surveys, to other things that are going to happen early in that product journey, in ways to optimize that toward being more successful on Meta?

Marcus Burke:

Yeah, that's definitely one of the changes I would say I've seen that UI managers need to be more and more product focused as well and they need to look into full funnel performance and not just purely on the ad side, because much of that efficiency from the algorithm is gone and you need to find additional levers to improve performance.

And I try to always use Meta as my channel for experimentation, to find angles that I can use for advertising, certain copywriting that works well and styles. And then take my winners and inform my full funnel with it. So if I found something is working very well on Meta, then I can bring these learnings into app store optimization easily.

Oftentimes, it's as easy as reusing your ad as a video or just a screenshot of it in your app store. So if Meta is one of your main channels, then that already creates coherence. People click through, they see I've learned in the right place, and this is the app that I was looking at. That often already creates an uplift, and you can take that then also down the funnel. If you found that, okay, we are seeing best results from an audience on Facebook with this type of creative that is between 35 and 44 and female, of course there's a lot in there that you can hypothesize around to create experiments throughout your funnel and tailor to that audience. If you showcase reviews in your onboarding, then you might want to take a user review from someone that is actually in that demographic or at least make it look like it. I mean you can always just say this is from a woman 44, from the US for example.

Additionally, you also want to make sure that other parts of that onboarding really work in sync with what you're trying to achieve on Meta. So if you are for example, driving a lot of traffic through Reels and you're targeting relatively young users, your pricing needs to fit that strategy.

I often see a mismatch there, where people are using a lot of these UGC videos these days, user generated content styles. So it's someone filming themselves on their phone talking to the camera, why they're loving this product basically. And this video style is what's native to Reels placements. And it's portrait mode, short form, usually below 30 seconds, and that placement drives younger traffic. But if your app actually... I mean maybe your app is already tailored to older users, which means you shouldn't be doing this anyways. But even if it's an app that is a bit younger and your pricing is very high, let's say 80, 90 bucks a year, you're going to have a hard time converting these people. Because they're just not as high intent and don't have as much purchasing power.

So your pricing, what your paywall looks like, and the onboarding needs to be really synced up with the ads you're running and the placements you're on so that things even get a chance to convert.

David Barnard:

Are there any signals that you get in the app from Facebook? I know back before ATT, you could actually get from the Facebook SDK the specific ad, or channel. Or you got a lot of very granular data where you could even personalize your onboarding, and your paywall, and pricing, or anything else based on even the specific ad they saw. Do you get any signals from Facebook now, or do you pretty much have to do that personalization based on a short survey or other context?

Marcus Burke:

You don't get any data anymore really. It's all aggregate. Mostly, you don't even know the amount of traffic coming from Meta, because what's tracked with scan often has a bigger discrepancy. Often, around even 50% of trials are under reported. So what you get is an aggregate number that is not all the users, and you don't know which specific user came from the channel. So that's why it's super important to really have an onboarding that enriches that data again so that you ask questions. "How old are you? What's your gender? Where did you hear about us?" The data you gather here is never going to match what's happening on the UA front one to one. It's not that you can ask users, "Where did you come from?" And then only attribute the ones that said, "I came from Meta," because a user journey usually looks a lot different from just seeing one ad and clicking on something. So they might have heard about your app from a friend than they saw a Meta ad. But in the survey, they're still going to answer, "I heard about you from a friend," because that was the first touch point.

So it's really just meant to create these data samples that allow you to compare. And you're going to see differences in someone that answered, "I came from TikTok," compared to someone that answers, "I came from Meta," for sure. And same for age groups, which is then the data that you can use to personalize and to inform your data modeling, to actually know how valuable your user acquisition efforts are to you.

David Barnard:

And then how are you doing this measurement? You've talked several times about modeling. And I guess maybe we should talk about this at different levels. When you're first starting out, how do you think about that measurement? And probably not even fully modeling. You might not even have a data warehouse that you're dumping product data, and analytics data, and everything else into. How do you think about then measuring this?

Marcus Burke:

Yeah. Usually, as soon as people have a data warehouse and analytics team, then I don't need to do that work anymore. So what I do is usually a bit more MVP, which is really using a few data sources to build trust in the channel. And one of them is incrementality. So you want to switch ads on, switch ads off, and see what's happening to your baseline, which is always of course a good position for smaller developers. If you don't have a ton of organic yet, if you don't have 10 channels running, it's going to be much easier to see, "Okay, I'm spending now 2K a day on Meta and my baseline has increased by 50%. That's probably the effect from the ads." The more advanced you are, the more you will need a data science team for this, because the fluctuations on your baseline are going to be much smaller and it's going to be harder to see what's going on.

Two other data points going into this are your ATT opt-in audience. So there's still a smaller set of users that you are able to match to a channel as you were used to back then. So that again, creates a sample where you can see, I don't know, 5% of people I was able to match on Meta, 6% I was able to match on TikTok, and I can compare how valuable are these.

So just to inform, should my cost per trial on Meta be 3X, 4X higher on TikTok, or should it be a lot cheaper because traffic quality is less than that? The other one is this, where did you hear about a survey? So that users actually give you a qualitative answer on, "This is the channel I came from," which can be a third input into this model. And from there, it's really a bit of setting your cost per trial goals accordingly. So that you know Meta is always 2X as valuable as TikTok, so we need to adjust our cost per trials based on that level.

David Barnard:

Got you. And then as folks scale up, what are you seeing with these more sophisticated apps working, and tooling too? I mean, at what point do you layer on an MMP? Or if Facebook's your only channel, are you just dumping everything into a data warehouse and having your internal analytics team do aspects of what an MMP would be doing for you anyway?

Marcus Burke:

Yeah, I mean you don't need an MMP to run Meta. They have their own SDK that you can implement. And you can insert your scan model there, track your trial event, and then it's doing the job. Depending on what your ambitions are, how quick you want to scale, I would say if you're a small developer, start out with that and don't add on additional tooling. Because it creates another breaking point. Something can always go wrong, and also it's going to cost you money at some point. I mean, usually there's a bit of a free plan. In the beginning if you have very few tracked events, but then it quickly gets to be another cost center.

And usually, RevenueCat is one of the tools that is used for these early devs to look into then trial performance. How do my trials convert into page, how's that different on a country level? And that would even inform modeling. So really, I think most of the developers I'm working with are 10 to 20 people teams, and they don't have very sophisticated tooling for this yet. And it's not really needed at this early stage.

David Barnard:

When you say modeling, this is spreadsheets and intuitive "modeling," not necessarily having to build an algorithm to be creating some sophisticated matching and all that kind of stuff. It's more looking at your data, half of the model's in your head as you're running the ads, looking at the data, and making decisions based on that, right?

Marcus Burke:

Yeah, it's not super sophisticated as I said. It's very scrappy. It's really just getting an understanding of the audience that you're driving, how valuable it is on a country level, age level, and a channel level, and then setting your cost per trial goals accordingly.

And of course, don't forget about also rechecking if your assumptions are right, because it's always just assumptions. So if you're assuming, "I can spend 50K a month on Meta and well get my money back within," I don't know, maybe even instant payback, so on day eight after the trial converts, then check your actuals. Don't rely on just the modeling, but make sure that you're actually recouping money in the timeframe that you predicted it would happen. But often in the early days, it's quite scrappy. And then as teams grow, they would collect all of this in a data warehouse. Usually, have someone that is a bit more skilled in all of this. So a data engineer analytics person who then looks into how can we create a predictive model and take in all our UA data, what we have on scan, what we get from the MMP, and feed a UA manager then with a estimated ROAS pretty quickly so that they can take decisions.

But with most of the teams I've seen even bigger ones, it was still at that scrappy level. Because as I said, it's not scientific anyways, it's all just predictions. So I haven't seen anyone of the small to midsize teams getting super advanced on it.

David Barnard:

Yeah, one of the things I think a lot about too, kind of seeing behind the scenes at RevenueCat as we roll out things like our experiments feature and even our charts and so many other things, we have 30,000 developers now, and they're all checking our math on everything. We find bugs that we have to fix, but we're finding all the bugs because we're working with so many different developers with so many different needs, and find so many different edge cases.

But I often think about that. Any one app, if you are a larger app and you've created this sophisticated algorithm, and you have a data team, you still need to be going back and at very high levels and in less sophisticated ways, double checking all of those assumptions. Because one little mismatched SQL query is just going to blow things up in a way that's going to cost you a lot of money. And so while as you grow, you can get more sophisticated, you need to be careful about that sophistication because things can go wrong and do go wrong.

Marcus Burke:

Totally, totally. Yeah. Even back in the days when I was working gaming and everything was still tracked on an IDFA level, we had very advanced models on all of this predicting lifetime value for the next two years. But those were usually built on huge amounts of data and then I'm optimizing on an ad level, and of course the model can go wrong the more granular you go. And a big part of the job was really digging into that data, improving the data team that they're making a wrong assumption, and my ads might actually be doing better than they think they are.

So definitely always go back. Also, check historic data. You had an assumption. Now you have three months worth of data. Look into, how did this perform? Is it better or worse than what I had predicted to fine tune this over time?

David Barnard:

And then by definition, any model breaks as you make big changes. So like you said, three months ago, you have a baseline and you think it was working. And then you roll out a big new feature. Well, that breaks your model by definition, because those assumptions that went into the model that has been working now doesn't work, because you have new features that you're promoting in a way that maybe creates higher intent and drives higher value per user and things like that.

So yeah, I mean it's funny how scrappy teams in some ways are better off because you're just doing the math in simpler ways. And the more sophisticated it gets, the more places it can break. It's just a matter of being smart about it, and triple checking assumptions, and not forgetting the basics when you're doing all this kind of stuff.

Marcus Burke:

If the UA team also handles the modeling themselves, then it's much easier to get these changes. And while once another team is actually responsible for it, then it's company politics and you need to prove them that maybe something needs to be changed, and that just makes things more complicated. Yeah.

David Barnard:

The big elephant in the room we haven't discussed yet is creative, and that's so make or break for these campaigns. So how do you stand out? How do you use these short little interruptions or content in Facebook, and Instagram, and other places, to get people's attention, to drive that intent, to find the right audiences?

Marcus Burke:

I would always recommend using these apps and being close to that content. Many people love to dig through competitor ads to make ads, and I think there is a lot of value in this, especially if you look at big brands and you know they're doing well on the platform. But in the end, what you're trying to achieve is that your content is similar to the content that people engage with when it's not an ad. And for that, you need to be using these placements and understand what makes them tick.

So that would be my biggest advice here, especially talking about Facebook. No one I know my age in their mid-thirties is using Facebook anymore. But for Meta ads, it's still a super valuable placement. It's a bit older audience, a lot higher purchasing power. So you want to know what's going on there and how you can tailor to these audiences.

I would also recommend try to look outside your bubble. So follow some sites that you wouldn't usually, because your target audience will be a lot more diverse than what you're looking at. So I tend to follow, for example, a broad spectrum of news sites on Instagram and on Facebook, because I want to see what are the headlines, how clickbaity are they, what other styles they're using in their images. Because then, I can tailor my content to them. So while I don't agree with anything they write, I still follow Fox News on Instagram and Facebook just to see what's the content some of my target audience might be consuming.

David Barnard:

That makes a lot of sense. In this new paradigm where you do, especially early on, have to run very consolidated campaigns, how do you test creative? And how many creatives are you running and how do you scale that up?

Marcus Burke:

I would say the times are gone where you created thousands of creatives, and you just throw them in there and hope for Meta to do the job because creative testing has gotten a lot more expensive and data quality has decreased. So you want to be a bit more in charge and make more qualitative bets here. So my creative testing volume has gone down quite a bit. In terms of best practice structure, always test your creatives in a separate campaign, not in the one that you're scaling and that is running smoothly, because it will interrupt them and things can go wrong. There's this infamous learning phase. So whenever you make a change to your running campaign, it resets it, meaning the campaign recalibrates. And that can lead to fluctuations in performance. So if you upload a new creative into your scaled campaign, you want to make sure it's a good one. You want to have tested beforehand and know this is actually going to improve performance and not decrease it.

And other than that, I always try to test them in a similar environment that I'm also scaling in. So if my scale campaign is optimizing for trials and I have 10 core markets that I usually advertise in, then my testing campaign is going to run in the cheapest of those markets, but in the ones that I'm also live in. And it's also going to optimize for a trial. It's not going to be a totally different campaign setup, which I often see advertisers do because they want to save money, and they only optimize for an install, and they go into markets like Thailand or Mexico where there's cheap CTMs. But then the audience is going to be a lot different if you're not live in these markets. Usually then, of course people haven't heard of your brand yet, so the effect there is going to be different. So I'd rather try to be a bit closer to my actual audience when testing. It's going to be more expensive, but then results are just going to be easier to port over to skilled campaign.

David Barnard:

The last thing I did want to talk about real quick is the future, and Meta has already made some advancements. Scan four is coming out. Does the future look bright for continuing to improve upon the improvements we've already seen? And then what are the things that are contributing or potentially going to contribute to increasing performance over time?

Marcus Burke:

Yep. I would say yeah, the future does look right. Scan four as you said is on the horizon. There are a lot of things in there that are going to help you with the channel. Two main benefits I see is for one, privacy thresholds are coming down, so it's now called crowd anonymity tiers. And based on some first data that AppsFlyer shared from what they see in their client's data, it's only going to be 20 installs per day down from 88, which is quite the difference. And especially for more niche advertisers that have high CPIs, it's going to make a difference for sure.

And the other big thing is that we're getting these two additional postbacks, and that's also very interesting to again feed that model. So we have the word that you hear about this data, the ATT opt-in audience. And now on top, we can also look at the scan, second and third postback just to feel that model and look into what does my trial to paid conversion look like from first to third postback on Meta, compared to TikTok, compared to Google, which is going to get you even more security in what you're doing there.

The other things happening we mentioned quickly before was AEM, which is aggregate event measurement. It's basically Meta's form of fingerprinting. It's been a bit of a buzz. I see some people saying, "Hey, just switch everything over to AEM. You're going to have same data as you used to before scan." Others are a lot more cautious. I'm a bit in the middle there. I think it is a great tool, and I've seen advertisers that this really helped, especially the ones that have high CPIs and can pass privacy thresholds.

So basically with aggregate event measurement, they're getting instant postback data. And Meta is basically just doing the magic on the backend. But also in the end it's a black box. Everything is modeled. They're just looking into IP addresses and trying to match these users, and then add their modeling on top. Hence, you cannot really trust it, neither can you scan because there is discrepancies.

But I would say be a bit cautious about it, because Apple might take that away as well. If your whole strategy is based on running AEM campaigns and you don't invest into your scan infrastructure and how you evaluate performance under scan, then at any point, you might be in a similar situation to when ATT launched.

So I'd like to build my account foundation on scan because I know it's future-proof. It's what Apple wants us to do, and it's also getting better with each version. I mean, scan five is already on the horizon while we've been waiting a long time for scan four to be finally rolled out.

And then I use AEM for scaling. If I see performance is good with it, then why not use that opportunity? I wouldn't just not do it because it's unsecure. I'd rather take my chance there. But it shouldn't be the foundation and your pure performance based on it, because you can't be sure what's going to happen.

David Barnard:

When do you expect scan four to be rolled out enough where we're going to fully see the benefit?

Marcus Burke:

I have no idea. I also don't want to take guesses. I would've expected it to happen way faster because of these benefits, but the whole industry has been quite slow. There is this nice singular tracker from where you can see basically the postbacks that are coming on scan four versus scan three for all the different networks. And Meta has ramped up to 40% scan four some in February, I think. Now they actually went back down to 20% again. So I have no idea when it's going to happen. And this is already their second attempt at rolling it out. The last one was somewhere back in September 2023.

So they seem to be taking their time. They seem to not be really sure yet how to make sense of the new postback data and hence I won't make any guesses. But yeah, I'm excited for it because of the benefits that I mentioned. And there are a few more where data granularity is just going to be better on this model.

David Barnard:

Well, that's a good place to wrap up. It's nice to kind of leave on a high note that not only has it gotten better in the last couple of years, but that it's likely to get even better. Not to dive to chase another rabbit as we wrap up, but it is personally frustrating to me that scan four is probably what scan two should have... Because scan two is what Apple rolled out when ATT was rolled out. And had they spent more time, talked to the industry, really understood the problem of measurement better before rolling out ATT, I think we'd be in a way better place today and we could have avoided a lot of the challenges of the last few years.

I'm always frustrated with Apple and how slow they are. I was so frustrated with Apple Pay. It was like, "Why don't you just pay all these merchants to install the new hardware and take Apple Pay? And let's get the show on the road."

Well, they're patient, and it's fully rolled out. And almost nowhere I go doesn't take Apple Pay anymore. And so it's like they're patient and eventually it gets there. And so I think that's at least a hopeful place to end the podcast is that scan 4.0 is coming and it's going to keep improving. And scan 5.0, and scan 6.0, it is going to get better.

And I mean, the cool thing personally, I'm a bit of a privacy zealot myself. And so it is exciting too that as someone who cares about privacy and not wanting to have my user data just hoovered up by a million different people, and selling data, and all the kinds of things that have gone on pre ATT, it's exciting to me that all of this is happening in a privacy-friendly way, that you're not having to violate user privacy to do it. That's all pretty exciting to me.

Marcus Burke:

Totally. And I really like how the job has changed due to this. I really like that you have to go deeper and go into qualitative data. And I don't know, doing user interviews is super helpful because you learn how to do better advertising. And it's not just all sitting in front of your data analytics stack and throwing hundreds of ads at it every week, but you actually have to do your homework a little bit better. And to me, it's a lot of fun and it's kind of moved things in the right direction. And of course for Apple, there's also the benefit that Apple search has 5X the market size of something.

David Barnard:

For those of you who are running Facebook ads, since I'm not, you've probably been shouting into the void asking questions that you hope I would have asked Marcus more advanced questions, more tactical questions. The good news is Marcus actually did an AMA in the sub club community. If you go to chat.subclub.com and look for the AMA and that's public, you have to join to get to some of the private channels, but we've been doing public AMAs there on the sub club community. So he's actually answered a lot of more tactical questions in that AMA.

And then I'll put you on the spot, but I want to have you back and do another AMA. Because those AMAs, it's like when people are flying blind and not working across multiple apps like you do, it's just so helpful to be able to talk to an expert who's doing it.

So yeah, join the community and we'll do more AMAs with Marcus, with Thomas, with other kind of experts in these fields in the future. And then you can check out what people already asked Marcus on the AMA that happened a couple of weeks ago.

So yeah. Anything else you wanted to share as you wrap up? I know you're actually accepting clients. So if people do need to scale, and as much as they've learned from this podcast maybe still need a helping hand, are you open to clients right now?

Marcus Burke:

Yeah, always happy to kind of work on fun challenges. I have a big backlog due to LinkedIn going quite well, but basically I'm always looking for a fun challenge. So feel free to reach out, give me a follow on LinkedIn. I try to post every day also more tactical stuff and feedback has been great, so I think it's useful. And I'm also planning to launch some digital products soon, probably an email course on Meta ads and how to make it work for subscription apps. So hopefully by the time this airs, I might even already have a signup for that on my page. So definitely give that a visit.

David Barnard:

Very cool. And we'll have in the show notes links to his LinkedIn and other places to find Marcus, and make sure we link to the AMA as well. Check the show notes and get in touch with Marcus.

Thank you so much. This was a blast. We could have gone two or three hours, but I do try and keep the podcast around an hour. We've gone a little over, but this was so fun. Thank you so much for joining me.

Marcus Burke:

Thanks a lot David. It was a lot of fun.

David Barnard:

Thanks so much for listening. If you have a minute, please leave a review in your favorite podcast player. You can also stop by chat.subclub.com to join our private community.