# What it Takes to Succeed with Paid User Acquisition — Thomas Petit

Channel: Sub Club by RevenueCat
Video: https://www.youtube.com/watch?v=i_XoG6I2RiM
Duration: 1 hr 19 min
Language: English
Words: 15580
Transcript page: https://viewrankai.com/tools/youtube-transcript/i_XoG6I2RiM

---

[0:00] Hello, I'm your host, David Barnard, and my guest today is Thomas Petit, an independent consultant focused on subscription app growth. Over the past decade, Thomas has worked with hundreds of clients and helped manage tens of millions of dollars in ad spend. On the podcast, I talk with Thomas about setting sensible goals for paid marketing, how to measure and learn from the results, and why a single ad creative can completely change the

[0:27] trajectory of a company. Hey, Thomas. Thanks so much for joining me on the podcast today. It's a second or third time on, but I'm super excited to chat. We've a lot to talk about. Yeah, thank you very much. I'm really happy to come back. I'm a huge fan of the other episodes, and I have to say you have really a great diversity of guests. So, I hope I can keep the bar really high, but I'm really happy to

[0:47] come back for a new topic today. The reason I wanted to have you back on, one, your past episodes were amazing, but we talked about very timely things. We talked very specifically about ATT, and we didn't talk as much about general principles around growth. But, the main focus of what I want to get at today is that I've been talking to so many developers struggling with paid UA. The funny thing is this is not a new thing, right? Paid UA has always been hard, and there's a little bit of romanticism of like pre-ATT it was so easy, and that's not true. But, it's just a thing that so

[1:20] many apps struggle with. Any marketing struggles with, like, how do I make the unit economics work? And so, that's what I wanted to kick off with today, since you talk to so many developers like I do, but then you actually work as a consultant with a lot of them to like, "Okay, how do we actually make it work?" I don't have to work that side of the equation personally, so I love talking to somebody like you who is day in day out with multiple apps. Your job is to help them make it work. I wanted to kick it off with your thoughts around kind of the minimal viable status, minimal

[1:55] viable app, minimal viable company, whatever that context around when you should start thinking about paid UA. Like, what does it take to actually start making it work? Yeah, that's a tough question, but I think it's good to ask the question, to be honest, especially in the early stage, like when you really go from zero to one and you don't like have a bunch of data behind about how it's going and so on, like when you're starting. I see too many people like just straight jumping in and I want to spend this money on Facebook or whatever without having done a bit of a health check of

[2:29] why I'm doing it, what I'm going to try to achieve, what stack does it take. And a lot of it comes around goals and you have to dig in but why? But why? Again and again. Cuz usually they just tell you I want to accelerate. I want more revenue. I want to accelerate. But why can't you do it otherwise or why do you think paid is the solution? And most importantly, like what are you minimum bar on this? Especially minimum bar of return. Like, what are you expecting from paid for every dollar you put out there? Are you expecting to make three, five, half of one? And all these

[3:00] situation can justify themselves in cases. When you have that clear, you're going to avoid a lot of the bad surprises. Sometimes it will lead to dead end. I have one app I work with and I've been working with them for quite a while. I was chatting with them for years before we started working together. And they come very clear to say, "If anything can't run at at least two weeks return straight from the money

[3:22] we put there, we're not running it. Period." And I say, "Yeah, but given the metrics you have, it's going to shut down a lot of options and channels and so on." I say, "Yeah, but this is what we're willing to accept and anything below that is trash." I say, "Okay, I mean, that's how you take it." And on the one channel I'm helping them with, we do manage to reach that goal, but only up to a certain scale and there's

[3:46] no more than that. Apple search ads, I imagine. In this case, it's Apple search ads, but like It also happens elsewhere, particularly on Android for subscriptions, like even on other channels. I have the same problem with Google Ads for another client. It's good to set the bar, like what are you trying to achieve? And in some case, you start understanding how aggressive people can go. Like in some cases, accelerating fast is actually a priority because there is a first mover advantage or market share matters a lot or there is a network effect or I work with a developer who was putting money on the table that they were not going to

[4:20] recoup directly, but they needed a critical mass of users for the app to have enough value so that then it started create that virtuous loop. And they had their goal really clear, but I think a lot of people who go from zero to one, they're just like Oh, I want to accelerate. That's it. I want more revenue. I'm saying, "Wait, but what are you expecting?" So, I think that that's really one. And very often it's about return goals, but in some cases it's about achieving velocity and network effects, critical mass or market share. Or on your marketplace, it could be I want to rebalance what I have. I

[4:52] had one case of a marketplace where they had a ton of seller for one category, but no buyer. And they had a ton of buyer for another category, but no sellers. And so, we run acquisition specifically aimed at rebalancing a little bit these categories. Where on surface, it looked like it was expensive. But eventually, it enabled the other users not to leave to avoid loss, basically. Basically, know your

[5:18] expectations and know what you're after. And I think a lot of people, they don't ask this question in a deep enough way. Yeah. Or they give too simplistic of an answer. I think if you ask most people, the answer would be, I want to spend a dollar and get a dollar 20 back. Within 30 days or whatever that answer might be. And And that's the expectations going into it. But I think That's already very specific. That's That's pretty good to know it at that point. I I uh, I hear that, I'm like, "Okay, at least these guys know what he can afford and I consider the time frame

[5:50] and he's looking at that." Like, there are already a lot of sub question that are answered in the way you frame it. So, I would qualify this as a fairly advanced way to define the goal. I think it's an advanced way to define the goal, but for most apps, it's a very unrealistic way to define the goal. And that's was the point I was going to get

[6:06] at is that it's easy to say that, right? It's easy to think, "Okay, my goal is to get a return on ad spend." But when you only have, let's say, a thousand people who've ever downloaded your app, you're very early. Like, it's just very unrealistic with a thousand ever downloads that you're going to start doing paid UA. And this is what I hear from people, like, "Oh, we're just going to turn on marketing and like we're

[6:30] going to make money." It's like, "No." Like, the very first step is that your first goal should probably be to get to a certain floor of monthly active users where you understand if you even have product market fit. Like, with a thousand people, you don't actually know what your unit economics are. With a thousand downloads, you don't actually know if your onboarding is working. You don't actually know if you're even in the ballpark of what you should price the app at. There's so much context behind the goals that if you're really early stage, the goal maybe is just we know we're going to lose money, but

[7:07] we're going to lose money doing paid UA in the service of experimentation, dialing product market fit, understanding our audience, just getting live warm bodies into the app and see how it goes. Because if your goal is to get a return on ad spend as an early stage app, that's probably not going to happen. And then I think there's other contextual goals as well. If you're a VC-backed company and you need a big outcome in the long run, that's a very

[7:33] different goal than if you're bootstrap. Like, I was talking to a guy who has amazing ASO, he wants to grow his business, but he can do it in a more methodical way. He's very profitable. He built a model that I can spend on paid UA if my return on that investment is relatively confident at 366 days. So, he takes all the initial conversions and then discounts the renewals. You know, he knows whatever 40% of my annuals are going to renew at year one. So, if you take the revenue on my 7-day free trial conversion and you take 40% assumption of when you renewal, as long as I'm

[8:14] profitable at day 366, I can cash flow a certain amount of paid UA and then that 40% is going to renew the next year at 60 or 70% and then the next year after that. And so, he's going to build this bigger business over the long haul by having a goal that doesn't sound amazing. It sounds scary to a lot of folks on paper, but the context of his app that can work. So, I think you're right that anybody thinking about paid UA needs to be more introspective and realistic about what they are going to be able to achieve with paid UA versus just thinking I'm going to start

[8:56] spending money and generating a profit on this and getting a direct return on ad spend. In some case, you might want to start UA very early and not even build a base before. Maybe you know that's going to be the channel and you want to know from the very beginning where you are at. Or you even want to iterate on the product from this kind of traffic. And that actually happened to me once. Almost like the first app where I got the whole budget for myself was really, okay, what are the goals? And I like we need to have at least 500 new users per day to

[9:29] iterate on this onboarding, on this paywall and get retention data that is we're not going to achieve it organically. We don't have the time to build the organic. We're going to do it in parallel, but we need to accelerate. We can afford to lose that amount of money, but we need this amount of users, and we need to learn within 3 months how much is coming back, and we basically need to measure all of this to prepare for what was coming next. I think here it was clearly a VC-backed startup, but they knew from the start that this is how they would grow, and so for them it

[9:58] did make sense to start early, but not with a I'm going to return to X on that, but rather to like let's build the app towards that. Let's build the foundation, and let's also learn in the process not just to iterate on product, but let's learn the earliest possible what messaging is working. This is what we're going to put in the store, what kind of creative is working, because we're going to build the team and all the infrastructure to buy in terms of also in metrics and tools. So, it was really like learning goal, like bring the user, but also within 3 months we need to have all of this list checked,

[10:30] because we want to accelerate, and this is the goal in this case. So, there are many kind of goals. I think there's another check that really needs to be done. I call it cash availability. It is what can you really afford? Because return is never guaranteed. Like sometimes it's a bit of a roller coaster, you don't really know. You open a new channel, you scale up, you start from zero. Even when you're stable, it change. So, don't assume that you're going to be at this goal. There might be some loss, maybe there's a time of learning period for the algorithm as well, and so on. And so you need to

[10:59] check a little bit what's the cash availability. In the case of the example you gave about the developer who could afford to have like over a year of payback on the renewal, he also knew that he had enough cash in the bank to finance this acquisition before the money was coming back. So, that's something that's quite important to know. If I had to frame it in a bit of a trivial way, I'd say only gamble money you can afford to lose. Maybe it's going to come back, but maybe not. And even if it comes back, it may not come back in a week. So, can you actually afford to

[11:27] finance this? Or maybe you need to change a bunch of things before you can actually get there. So, cash availability is important. I have another case of somebody from Subclub actually to whom I'm talking to where they're fairly at this stage they've got a bit of a base of organic, but the cash availability they have for paid UA is is fairly limited. It's in the very low

[11:49] five digits total, not even per month. And then raise the question of should you actually do any of it? And some channel can be operated at at low scale to some extent. Maybe Subclub, maybe influencer. But most of the bigger platform that use algorithmic optimization, so that's valid for Meta, TikTok, Google, but that's also valid for all DSP and networks and so on, do not go there with $20 a day. It's money lost. Like you're not going to learn anything, but you can't really make the machine work at all, not going to pass threshold, there's a bunch of like criteria. And that minimum bar of the amount you can

[12:25] put to actually start seeing something, whether you like what you see or not is another question, but to actually and learn from it as reason over time. 10 years ago it was a lot more manual and there was less events, and now it's more and more, let's say, automated, and that can work really well in some case, but if you invest too little, it's doomed to fail, even if your product is

[12:48] fantastic. With Facebook, I know so much of this is automated with the happy path, you know, where they want you to spend your money is through the algorithmic ad buying. But are there still ways on Facebook, so if somebody does just have $20 a day to spend, can you manually go in and if you really know who your audience is, actually set up more manual campaigns and spend that $20 a day with like super highly targeted, picking your audiences and stuff, or is that just not even

[13:16] possible anymore? It's technically possible. I would strongly recommend against it. Okay. There are many reasons. First is the targeting, like people say, is never as narrow precise as you think it is. So, that's one. Even if you use advanced things like lookalike 1%, like kind of a very narrow audience, but the rest is kind of even if you limit the demographics and the interest and the lookalike and maybe a specific area and you reduce the So, there's a bunch of problem that comes. First, when you do that, Facebook tends to penalize you. If the audience is not big enough, you may not get any delivery at all. That's a

[13:54] possibility. And that's most likely you're going to get it extremely expensive. Like if you want to go so narrow focus, you're going to pay such a premium to do that that it's probably better to let Facebook figure it out on itself and actually go for these users at a cheaper price by itself. Like it's probably more efficient. There is an

[14:16] additional factor or two even. One is if you're not optimizing like for downstream events, it's most likely you're going to be served garbage because Facebook and other platform, they are going to preserve the inventory that is most likely to convert to those who buy for this. Like when you're saying to Facebook, "Oh, I'm trying to get free trials." Let's say. What is going to happen in the background is their whole machinery is going to calculate in real time that that particular user who's done this on Facebook, this on Instagram, also interacted with this advertiser before has converted on a bunch of other apps and seeing that creative, what's the

[14:55] likelihood of that impression converting into actually this event? And so, they preserve this high-value inventory for whoever optimize for that. If you run at very low scale, you obviously can't optimize for event. There's not even enough data. So, you're going to start buying on install or on clicks or even on impression. What happened in this case is Facebook serves you the inventory that is not going to convert. And so, as you buy on CPI, the install to trial is going to drop. If you buy on click, people are not going to install. There's a bunch of people who actually don't install apps on Facebook and Instagram. And obviously,

[15:31] if I'm buying on install on trials, Facebook is going to show my ads to less of these people. But if you tell them you have to click, they say, "Oh, I've got a bunch of people who are likely to click and do nothing else. So, that's the one I'm going to give you." And so, if you're a big brand and you go on full CPM and you say, "Okay, my goal is to be seen. I want this campaign to reach eyeballs. I don't need them to take an action." Facebook is going to do it, but they're going to actually serve you people who are never going to do

[15:58] anything because they preserve the other inventory. So, that's the second thing is if you're not optimizing for downstream event, it's likely that the quality of traffic is going to be less. And the third reason So, there's a premium. You're going to be served to people who don't convert. And the third one, specific to iOS, but it's going to be the same on Android at some point, is that through the aggregation of data, so I don't want to be too technical at this point of the podcast, but basically with aggregated data and the sampling of whoever has consented, you're going to be blind basically on who converts or

[16:30] not. Facebook is not even able to aggregate enough data to even report to you how many install or conversion have happened. So, you're going to hit, let's say, minimum thresholds that Facebook is going to be blind, you're going to be blind at some point. Either it's going to stop delivering completely or you're going to cut it because they say, "Okay, nothing is coming back." or something

[16:49] like this. There's multiple reasons not to operate at very, very low scale, at least on algorithmic platforms. So, I know this is hard to pin down, but we were talking about cash availability as we were going down that rabbit hole, which I think was a really good one to go down so that people really understand. I mean, I didn't understand that. I thought maybe there was a chance to like spend $100 a day and make it work through manual targeting or whatever. But what then is the kind of minimally viable cash setup. I mean, there's layers to this, right? That people who already have a high average

[17:24] revenue per user, they've got good retention, they've got good conversion to trial, conversion to paying, if all of those things are really well dialed in and the onboarding is really well dialed in, you can start to get closer to return on ad spend profitable quicker, where maybe you don't need a million dollars in the bank to do this big experimentation. But, what would you say are guardrails? Like, you really need to have six figures in the bank that you can and even six figures in the bank you could lose before you try and

[17:57] scale up on these algorithmic platforms? Or, what are the general guardrails around that? I usually think about it more in terms of volume of data to be able to optimize than in volume of cash, but then you can relatively convert one into another to give a bit of a ballpark. To give a simple answer, I think in my case, I do try to aim at five digits monthly budget. And, if you go per platform and assuming you don't have too much distribution of country and platform and so on, I'd say somewhere between 10-20 grand a month, I wouldn't start anything below

[18:30] that. Otherwise, it's doomed to fail. There's also within this budget, who is going to actually execute it? And, if you don't have a person in the house, maybe you can do it yourself and learn. And, actually think that's the best in the early stage. But, as soon as you want somebody to manage it and if you need to hire, you're not going to hire somebody who cost you two, three, five grand if your budget is 10 grand. If you want to go with a freelancer, nobody's going to accept you if the fee is not, let's say, 1K a month or 3K a month or whatever, depends the country as well

[19:02] and you might change country. Or, if it's an agency, they also have minimums. So, I'd say you can operate on 10-20K, but then it's only also one channel and I think that's kind of a bit of a risk. You might want to start with just one and so on. But I'd say for me like the area that I find really hard to nail is the 0 to 50k because 50k months is where you can start having two three channels running in parallel, or even I can split the budget between campaigns and start making a bit more of experiments. And let's say this kind of budget obviously

[19:32] depends the vertical and a bunch of things, but it's basically down to the fact that this kind of amount, so let's say 15k months is roughly $500 a day. That might convert into a couple hundreds install a day, or if I take the example of the US, it actually is the same the other way because the installs are cheaper, but they convert lower. So, let's say you you put $500 on the table, and maybe you're going to get 1 200 installs, and those are going to do 10 or 20 trials. So, depends a lot. Maybe your trial rate is higher, maybe you get cheaper install because your creative is

[20:07] awesome or whatever. But then you start having these 10 20 events per day in the campaign that I believe are a bit of a minimum, otherwise it's going to really struggle. If you're sending Facebook two events a day, it's never going to learn who your users are, the sampling is too hard, and the fluctuation is going to be extremely hard to read even for you. So, that's how 10 20 events per campaign per day backfills into that's going to be 10

[20:33] or 20k months on this channel. Makes sense. And then ideally more. That's the thing. There's also a ceiling where you've got an efficiency of scale, but on the other side, there's some point that you want so much of it that you go to broader audience, they convert less, or Facebook start ah yeah, if you want 100 install, maybe I'm going to serve them at $2, but if you want 10,000 install, then you're going to have to pay $4. Like, it's a bit caricatural, they're not the numbers, but the logic is there. And I think at the beginning, if you're able to provide 50 events instead of 10 every day, it's obvious

[21:04] that the company is going to run better. But at some point, I mean, this is not going to an infinity number. There are also ceilings that you start hitting. I think a lot of the folks I was referring to early in the podcast who are really struggling to make it work, for them, I think this is probably the issue is that they're trying to make it

[21:22] work on 5 to 10k a month. That's really hard zone to operate in. And if you take more than 5 years ago, and I'm not talking even pre-ATT, we're not saying UA was easy, but you could operate at a lower scale. Basically, because the platforms were a lot more primitive and they didn't have this kind of level of optimization and machine learning and what not that they have today. So, you could narrow it down to I'm going to do this lookalike here and I'm just going to bring the traffic and it's going to convert. And because there was not the tools that they have today,

[21:53] but there was also no privacy threshold. The zero to 50k a month is definitely the harder part. I see more and more people, agencies, independent that are like, "Oh, no, I don't want these kind of clients anymore." Because it's double the work and half the reward. So, I'll wait for them to have some kind of

[22:06] validation. And then I am them scale. It's more profitable for me. It's really hard to go from zero to 50k a month, that's what I believe. It doesn't mean it's not possible. And one advice, maybe if you've got 5 or 10k to test, probably would be to avoid algorithmic platform. So, search ads is I mean, the nature of the inventory is very different, but the mere fact that they're not optimizing for anything after the install. And to be honest, they're not even optimizing

[22:32] for anything after the click. Means that you can operate with a much smaller budget. Influencers is also one where you might want to try because you can find influencer of the size that you are and you can perfectly do a $100 operation or what not. And so, those are my go-to channels, typically, where I would start. I wish we had keywords on Google that are only buyable. It's not doable because the campaigns are bundled. But back in the day when it was

[22:57] possible, that was how I was starting. Because one, it was operable at very low scale. and two, at this stage you want to learn what's working, and search is a fantastic channel about where is the intent, what is the messaging that works. I see a lot of web business that are putting budgets of 100 or 500 just to understand what makes people tick. Return goal zero. I want to know what is the message that is most attractive to people. And this is something that is quite harder to do now because a lot of the platform basically don't even let you do that. You could do one just to get the creative on a

[23:33] journey to your website or whatever, and then bring this learning to the ads and actually kickstart a bit faster. That's something that I still do. Usually I do it when the app is not even ready. You're still in development or whatnot, and it's a great exercise to do. Like, what are the keywords that I search for, and what is the messaging that actually

[23:51] makes people tick. At a small scale, do Google web search ads, can you not still target to those? Because people do go to the Google search box, and you know, not Play Store, but even iOS and Android, people just go to the Google search box and say best running app, best weight loss app, or whatever. Can you not target those

[24:14] directly in a Google search? Yes, you can do that. On the web, Google still allows to go like very narrow, choose your keywords, run at very low budgets, and so on. So, that's still doable. They're kind of favoring whoever optimize on conversion as well on the web, but it's still a possibility, and I've seen some campaigns work this way. Usually the struggle at this point is that a lot of developers, I mean, it's really hard to maintain a good app and a good website in parallel. You're going to want to iterate on this landing page, and so on. Is that you have to offer some kind of user journey? You cannot

[24:47] send to the store. Like, it's forbidden by Google to use this campaign and have them run straight to the store. So, you have to have them go towards the landing page. What I would recommend in this case is actually to build something really simple. Like take one of these landing page builders like Unbounce or Flow or Instapage or whatnot. Have the market completely independent. Do not do it on your own website because the iteration is going to go slow. And maybe you do something very simple. Simple landing page that explains what you're doing and redirects to the store. This is possible on a very low scale. The

[25:20] thing is if it was that easy to make the unit economic work, everybody would be doing that. There are a few apps, especially in the subscription space, that do manage to get that journey to work at some scale, but it's definitely not the easiest one to do. And so if you're starting and you're already going to one that you know is kind of hard to crack, the thing is that it's full of learning. So I would still recommend it, but knowing that the unit economic might not be the worst there, but you might get a ton of learning for your SO, for imagery, for what people are searching. So I do think

[25:56] that's still doable, but I would say that's really in the earliest earliest stage. So we mentioned what is your goal. We mentioned cash availability. The last I do, even if the cohort is small, is what kind of revenue per user are we getting right now? And in a few cases, I've said, "There's no chance we're going to make it work no matter what, so let's not even try." And usually that's because our pool is really low. I don't know. If you're making 20 cents per install in the state, there is no UA manager on earth that can attract traffic that is going to be profitable. Like period. Like So

[26:27] running a bit of a very simplistic, very basic, what approximately is my revenue per user? I'd be struggling to say from there you go. It's also very specific by country and platform. And the price of install is not like fixed. It can be extremely different by vertical. If your creative is really good across platforms, it's hard to pin a value, but over the years you learn by vertical more or less like, "Okay, if you're making less than a dollar in the states on a subscription app, it's not going to work. For some categories of gaming, $1 per install is actually a fairly

[27:00] reasonable amount and you might make it. If you're talking about countries where the inventory is fairly cheap, like big publishing country on Android, like Pakistan, Indonesia, Egypt, like massive populations where the inventory is relatively cheap, if you get $1 per install on those segments, there's probably a way to make it work. Like so, it's very relative, but so those are the three checks, like goals, cash, and

[27:26] Apple revenue per install. I want to add some caveats to the revenue per install side of things. I think one important thing to understand too is what your funnel looks like currently and why. So, you know, this is something I've talked a lot about over the years, but you know, my apps have been featured quite a bit. And what I notice when my app gets featured is that the average revenue per user drops through the floor. So, if you're looking at an app whose like primary source of installs is the Apple loves the app, that's great. free installs, it doesn't cost you anything, but your average

[28:00] revenue per user is going to inherently be lower because of that. And then on the flip side, if you're looking at the unit economics where the app has really good ASO and it's like the top result for a very specific keyword, and 90% of their traffic is people searching that very specific keyword, the average revenue per user could be $5, and it looks amazing, like, "Oh, we're going to make Pay to Win work. This is perfect, like $5 per install, like we can buy installs cheaper." But then the problem you run into there is that the intent on a paid install is nowhere near the intent on a search install. And so, I

[28:37] think taking into account the source of that average revenue per user is an important factor as well, because I've seen that happen, too, where people are "Our funnel's amazing, our unit economics are amazing, like everything is lined up and we should be able to make paid UA work, and then they go try and it doesn't work. And why is it? And you dig a little deeper, it's like, well, all their numbers are flawed because they're from a very specific targeted audience who has super high intent and they just can't achieve that

[29:06] level of intent with paid UA. I don't know, absolutely. I mean, I'm seeing apps where organic is actually much lower converting and our pool than paid and I'm seeing app where it's the other way around. I'd say it's still good to check and yeah, make a check on sources as well. There is variance. It still gives an idea, but it's definitely And that's why I don't want to put a number. Like I don't want you to ask me, "Okay, so what is the minimum output to run UA?" Like because I can't answer this question simply put. Like it's impossible. It's still good to check, but you have to

[29:35] take into account the context of what is this cohort. There's also a scale question that sometimes you manage to find like an extremely niche cohort of users at the very beginning, but then as you broaden the audience, the behavior is completely different and that happens at different stages, not just at the beginning, but like there are points where you start breaking like the core of the audience that is highly converting and maybe the audience is not big enough or so. There's a bunch of caveats, but let's say those are the

[30:04] checks. We've talked through these kind of like minimal viable situation. You need to really understand your goal and then be realistic about your goal. You need to have a solid understanding of your average revenue per user and why and whether it's realistic to get that with paid campaigns. And then you got to have money. You got to have the cash to spend it, but once you do all that, you also then need tooling to figure out what's going to work. So, you know, if you start with App Store search ads, you don't necessarily need any tooling, but what are the kind of stages that you coach people through of

[30:44] what you need to actually understand the results as you do any scaling of paid UA. It's fundamental that you have a way of seeing what's going to happen. No, you put the money out and you want to see something, like some kind of measurement. So you're going to have to have this in place. In some case, I can see from the very beginning because of the ambition or where it's going or the kind of budget or whatever that this app is going to go on multiple channel at

[31:10] the same time and it's going to scale. In this case, I'd say probably not a bad idea to have an MMP, a mobile measurement partner because if you have it from the start, then your data is unified rather than try both of things and at some point 6 months later say, "Oh, now we need it." And then really hard to reconcile the data. So in some case, I start straight on, but it's actually not a requirement. If you're starting just with influencer, basically need almost nothing. I'd rather have one of these magic link systems that the MMPs are giving, but that you can figure out as well otherwise. But

[31:44] basically, there's no particular measurement that is going to change the result. If you're starting with search ads and I mentioned these two channels earlier in the podcast that I'd probably start there when I have little money. The framework for measurement for Apple is entirely different from every other network and it's free. There is a self-attributing API that you can install yourself. The MMPs are putting it out of the box and there are number of tools that enable you to also have it

[32:09] without doing any kind of installation. RevenueCat is one of those. It comes outside of the box with what you're buying. You can also do it yourself. Like the API is open, you read the documentation, you implement it. It's not particularly complex. So here we come to it depends the channel, see? And here the tooling you need first brings you to the question of actually where I'm going to advertise, which brings the question of who's my audience, who are they? And thinking a bit of the kind of user that is most likely to be audience fit right, you probably thought a little bit about the user persona or have already data or

[32:41] whatnot or I've just these preconceived ideas that may or may not be true, but you know that, oh, this is going to go more on paid social and it's a younger audience and I'm going to go for them on Snap or on TikTok or whatnot or oh, this is a very I mean, there's all kind of people on Facebook, so everybody ends up on Facebook. When I say Facebook, that's Meta, so Facebook and Instagram, but you might have an idea where this is going and so you can see what is possible to do. In the case of Facebook, you can have the measurement. So, just to be is

[33:12] through their SDK, which you can also do it this way. There is a custom way without the SDK that I would not recommend. It's a bit clunky, but basically when you go to Facebook, they tell you, "Okay, do you have the Facebook SDK? Do you have an MMP or do you want to do it yourself?" And those are the three methods. A lot of times after search ads and influencer, Meta is going to be the one where a lot of subscription apps are going to start, so that's why I'm mentioning that one. You do not necessarily need the MMP. If you start on Google, for example, because

[33:44] your audience is in a country that is heavily geared toward Android, in this case, like market share of Android user acquisition is massively on Google Ads, like to a point that is really impressive. Depends a lot where, but let's say 50 to 70%. So, this is the channel that is like kind of the obvious way. Plus, it encompass search and display and the Play Store, you can have video and text, so it kind of makes a lot of sense to go there. Let's say if you're an app that operates in India only, for me it's a no-brainer that I'm going to start on Google Ads. And on

[34:15] Google Ads, all you need is to have actually Google system in place and that's actually the recommended setup to have Firebase. It's free to implement and this is the best way to operate on Google Ads. So, that's where I would start. So, for me the question of the tool really lays down on, okay, which channel I'm going to start and where I'm going in the midterm. At some point I

[34:34] usually tend to want the MMP early on. One reason is one to not lose any time with custom integration every time I'm going to have to open a new channel because they have it out of the box, but also because they're going to add on top like this magic links and a bunch of custom integration with ad networks that you can't do any other way. Like a bunch of networks they don't take you if you don't have an MMP like period. When Pinterest was doing app installs they're

[34:58] not doing it anymore it was like this. I'm not 100% sure but I do believe TikTok is still like this. They said that they would open something but I don't know if they did because all of my clients operate through an MMP these days. At some point it's most likely that you're going to want an MMP but it is not needed for the start. If you're going to only operate on Google, Apple, influencer you can do without. Even Facebook you can do without. So especially MMPs are not free so you might want to consider cost saving or just testing because maybe you want to validate that paid is a channel that is

[35:28] going to work for you. And then you can do it like kind of with a minimum stack that is basically you have to install the measurement that depends on which platform you want to advertise. It's not the same for every one of them so that's why that's where you're going to go. I would recommend to also look at Okay, this is what the platform is going to report the results but where am I going to double check that what they say is

[35:53] actually what's happening? And typically what I do is infer the incrementality or the difference from my baseline. One from the consoles. So I assume you have the App Store Connect and the Play Console. They are free and I don't think you can upload an app if you don't have them anyway. So there's an analytic path that you can access.

[36:14] It's not particularly complex to read. It's not particularly good either but it gives you a data point. And the other one is internal analytics. So some kind of product analytics tool that might be a more evolved one like Amplitude or Mixpanel or that might be Firebase because it's more cost-affordable or that might be one of the dozens of other alternatives that I didn't give the name today. But having those product analytics helps you really understand, okay, this is the self-reported results that the platform say, what am I actually seeing on my side? Because you've got all kind of case where the network might over-report or actually

[36:50] under-report. But there's not a lot you need to have in place because I would say you have the console and you should have some kind of product analytics in place even if you're not doing any kind of paid UA. But it's important that you're checking them in parallel. So, we've been talking about how to use tooling to actually measure the results, to look at the Play Console, to look at Apple's Console, to look at an MMP. But I think what a lot of people struggle

[37:17] with is what do you actually look for? And on that subject, I'm going to pre-prompt, give you some answers, but then I want kind of your your feedback on these answers. So, I just did a talk in Vegas. I'll put a link to it in the show notes, but I spent a lot of time talking about blended subscriber acquisition costs. And, you know, essentially your blended customer acquisition cost since we mostly talk about subscription apps on this podcast and that's what my talk was about, I use subscription. So, you should be looking at your customer acquisition cost and you should be trying to figure out your

[37:52] exact ROAS on the paid spend, but you're not really going to figure that out anyway. But the goal more broadly is to acquire subscribers profitably. And you're going to be doing that through SEO, through ASO, through influencers, through PR, through Apple featuring you. Like there's so many ways that your app is going to get attention, and then some of those are actually benefited by the paid UA. So, as you spend a little bit, you're getting more reviews, which helps with your ASO, which helps with your word of mouth. And so, ideally, like there's multiple kind

[38:29] of flywheels in here. And so, looking at a really high-level at blended subscriber acquisition cost, I think aligns incentives so that you're not only focused on your paid UA, but you're still pushing on SEO, you're still pushing on ASO, you're still looking for those organic channels, you're still incentivizing word of mouth. So, you're doing all of those things. But, you do need to be in the MMP dashboard figuring out, okay, am I like just totally losing my shirt on this paid acquisition? So, tell me about how you measure the results and how you look at both these high-level metrics like blended subscriber acquisition cost, but then also still try and get

[39:10] some sense of what your return is on that paid spend. That's a super valid point, and more than what my blended cost per paid subscriber or whatever is, what is the trend of this branded going as I pay more or less? Because if you just take the value and let's say you've got a ton of organic and you start spending not a significant amount, the blended cost per subscriber still's going to look great, even though your paid UA is doing absolutely horrendous. So, it's more

[39:38] about how it's evolving as I move spend. And that's a must-have exercise to do. One, because that's what's going to show up in the P&L of the company. And two, because that's the only thing you can trust. Attribution is not exact science. Like assigning where the traffic is coming from is actually not something that is going to ever be precise. And it wasn't ever. And I think this is something that is kind of interesting that before ATT, about 2 years ago, most marketers tend to trust whatever result was coming indirect and take it for granted. I always like, no, let's look at the blended on top in parallel and

[40:20] one to make sure that we're reaching company goals overall and not just from paid because if I succeed on paid, but we're not making it in total, it's not going anywhere. But two also because as the time go, this conversion that are getting reported are being more and more modeled and sampled. And that creates room for errors, especially when it's the same company that is selling you the ads that is responsible for the modeling. So, it's a really healthy I think the blended was always something that had to be looking at. But it's mostly the trend. I think the cost per blended cost per subscriber is a great

[40:57] one, but actually look at a bunch of blended metrics and how many installs moving, how many trials moving, how many because you start detecting as well like, oh, here I've bring a lot of installs and a lot of trials from this campaign. I can see how when my spend's going up, these metrics are going up, but the pace of paying subscriber is not. And I already know that these trials are the worst conversion than the rest of the traffic, which is something that most of the time the network cannot inform you about because they use proxy event and because they don't know have the full development of the cohort and

[41:30] so on. So, I get a real hint of is this actually profitable? And is this nature of traffic really different? And sometimes it looks in the network that I'm achieving my goal and I've got a ton of your manager who told me to bring these trials at $20 per trial and I did it. I was like, yeah, but your trials are the worst, like they're killing me. I said, yeah, but that's the only metric we have in there, so what can I do? I was like, you look at the blended, you look like me. So, it's a really healthy thing to do. The thing is looking at the blending,

[41:59] there's a lot of things you infer. There might be a lot of noise, seasonality, featuring, other product change that happened in parallel that makes the picture gets a little bit more blurry and that's how it is. And it's still to still look into it. We still use the metrics that are given by the ad networks themselves or the MMP or whatnot because looking at the total doesn't enable you to optimize the campaign at all. You don't get the creative level, you don't get the ad group level, and you have to directionally trust what you're seeing being reported directly to actually make the change. Very often what I do is kind

[42:39] of I make my budget decision and budget allocation based on what I see in the blended, but I'm going to make deep channel specific change based on what the network is reporting. And that's really the normal exercise of any marketer to be juggling between both of, "Okay, what do we do strategically? What do we do tactically?" All the time. That's how the game is played these days. It should have always been played this way, but that's how we eventually

[43:06] got there. So, what I hear you saying and the takeaway is that you should be looking at whatever tools you have at your disposal, whether it's an MMP, whether it's the Apple Search Ads console, whether it's Facebook's console, Google's console, or whatever combination of all of those you have access to, you need to be looking at those to get a sense for cost per install, cost per event, some estimation of return on ad spend, and make as much inference and optimization you can that level, but then you're always looking at the higher end level of blended subscriber acquisition cost, but then also looking at blended funnel

[43:49] analytics, like looking at all of your primary metrics, and then essentially doing incrementality testing. I made this big change adding a channel, how are those users interacting with my onboarding? How are those users converting to paid? How are those users and you look at all those different key metrics to make some inference on whether that campaign is actually performing at the level that maybe the MMP or Google Play console or whatever is telling you they're performing at, you need to like trust but verify. And the way you do that is looking at the trends that as you make changes, what changes are you seeing in your

[44:31] results? So, do you specifically make big changes like dramatically to be able to see that? Like will you ramp up spend for a few days and then ramp it back down to like do these incrementality tests? And then are there any good tools yet for doing this and modeling that kind of incrementality across product and subscription life

[44:56] cycle events and everything else? That's kind of my bread and butter personally. Like one because I like it and two because most marketers either don't like to do it, don't know how to do it, or don't want to do it, or whatever reason. Usually they're the bad ones. Higher the other ones that do it

[45:11] all the time. But you have to do it because attribution is never an exact science. The only truth is the blending. I'll add something and it adds complexity, but it actually un-blurs the picture a lot is that there are very very little interaction between countries and even between platform. So, I usually run these analysis on a country plus platform segment. So, I will only look at Brazil Android and I look at the trends at the spend, at the install,

[45:40] etc. etc. And I'll look at the US iOS. Sometimes I would bundle them into groups so that it got more volume because if I'm looking at Costa Rica alone, maybe it's three installs a day and I'm seeing nothing. So, I'm going to bundle all of Central America or sometimes I don't bundle them regionally, sometimes I bundle them by

[45:58] level of similar conversion or whatever. So, you have to do it in a segmented way. But, this kind of a must-have because that's the only truth. That's the only way you're going to infer anything. I personally do a lot of it manually, but that's also because I like it. That's something that was not so common until a few years back. So, the tooling around it, they're fairly new. And the platforms, like the networks are Facebook and Google specifically, have open-source models to do media mix modeling and try to infer. The one from Facebook is called Robin open-source and you can build on top. The one from

[46:33] Google is called lightweight. So, you can have a team of data analysts actually build on top something that is specific for you. There are also tools that are out of the box. I've used one for over a year that's called Incremental from Mao. Really nice guy. Perfect timing. He was preparing this and then suddenly ATT comes like a

[46:50] visionary. A lot of founders are a bit visionary, but I think he had really a good sense of timing. So, good for him to be there before this happened. There's a couple of tools that appear in that space. Recast is one that was not specifically made for apps and that preexisted. There's another one that's called metric.works. Their founders are publishing very interesting stuff

[47:09] regularly. So, I follow what they say. Those tools, I'd say, if I'm operating at a certain scale, I probably want one of them. If you're small, it's not just to justify the cost, but there's just not enough data for these tools to operate properly. And the thing is you can do it in a quick and dirty way pivoting an Excel spreadsheet or having a Google Studio or like Looker Studio or whatever for very minimal cost and level of reliability that's still interesting. It enables you to filter and clean because you've ramped up your spend only in one country, so you're going to zoom in there and you just pivot the data

[47:47] quickly. And very often you need to clean up a bit because there was interferences from elsewhere that you want to clean. Yeah, I'm going going that day because there was the featuring and I'm going to remove January because what not in my space. There's a lot of cleaning in this data, so actually doing it manually is very often not a bad idea. And if you're operating at a small scale, for me it's obvious that Excel is enough. You can you use

[48:11] Google spreadsheet if you want. Yeah, the other thing too is that I'm hearing parallels from pre-algorithmic Facebook and Google is that five or six years ago part of what worked about manually managing all these campaigns and targeting and look-alikes is that actually gave you a lot of insight into your users and like who was your best potential customer and how do we go find people like that? And now for a lot of the big apps that is outsourced to the models and you have to use other means to understand those customers and then you're not able to like target them more directly. And so I think part of what I

[48:53] hear you saying too is that you say you like it. Yeah, I used quotey fingers there for people not watching the videos. Thomas says he likes it. I would assume part of why you like it is not just that you're a nerd and like to nerd out on all this stuff, but it's that it gives you a really deep understanding of what's actually happening. Is it when you spend the extra $5,000 this week on influencers on TikTok you know if those influencers and whether their TikTok audience is actually working for your opportunity not and then that informs future decision-making. You understand you made a big change in creatives and

[49:30] you realize, oh wow, like that creative did not land. Sure, it drove clicks, but that's not the goal. So it sounds like you actually learn a lot by doing this kind of analysis as well that then informs the rest of product decisions of future marketing efforts and everything else. Absolutely. I mean there's a bit of both. I'm a bit of a nerd, but there's also a lot of learnings that I had that I had to get otherwise, and they're very complimentary. Like, it doesn't replace making user interview and so on. It comes on top. Like, it's kind of another

[50:00] vision. It's true that the networks, partly because it wasn't their interest, but partly for privacy reason as well, provide a lot less detailed granularity on the audience where we're showing up. But, actually there's a lot that you can infer also from the creative themselves. Like, because you see how people react on the video, and it's still first-party and so on. And this is fascinating to see how people actually act with different message outside of your product. So, also extremely complimentary. One thing about the nerdiness about it, like I've always checking funnel revolution and segments all the time for like every day for weeks, months across many segments

[50:40] is at some point you develop some kind of instinct about the pulse of the product that is going. And as soon as something happens, and it can be that it's a campaign that is going sideways. It can be that it's a product change that has completely unplanned consequences. Or it can be a change in seniority, but you see it instantly. And very often it's kind of I can act on it also as well. Or, oh, last year I concluded that this was seasonality. I need to prepare this year ahead. And when you're doing it all the time, I believe it gives you a unfair advantage or fair advantage in this case just

[51:14] because your level of knowledge is better of I'm going to anticipate. I'm going to make the better decision. And I'm going to detect stuff that happens earlier instead of let them run for a while and suddenly, oh, what has this going on here for weeks where we're losing money on that particular segment, but it wasn't showing up overall because it was too small to move the overall needle, but we're still losing money there. And I love getting to that point where you've done it so often that it become really instinctive, and you stop feeling it like, oh, I can sense that something is happening there, so I'm

[51:46] going to I had a bunch of discussion sometimes like either we start a list or PMs or whatever say, "How did you thought about getting there? Why did you split iPads there? Or why did you look at Latin America that day?" I was like, "One, I'm looking at those all the time and you're not and you should. So, that's the first answer. And second is if you do it all the time, you start getting a sense for it." And it's a very exciting point. One, because I'm a nerd, but actually from a business sense of view because it basically enables us to

[52:15] improve results faster. So, we've talked about a lot of aspects of what it takes to make paid UA work, but we kind of glossed over so far one of the most important things. Is you actually have to like make ads and you actually have to make I was I was going to say good ads. I will revise my phrasing and say you have to make performant ads. Sometimes the most performant ads are actually the worst looking ads, which is such a funny quirk to me. Yeah, let's talk creatives. Like can you outsource this? Is Midjourney creatives going to be a thing? What's that kind of minimal viable creative

[52:56] department to actually think you're going to be able to be successful at ramping up paid UA? We had to talk about it because I mean there's such a massive part of paid UA success. I remember the early days of it wasn't that important. I mean the pressure on that was less. People were less used to ads and honestly we were putting like random stuff that was working. It was fantastic. I would put the same ads that I was putting eight years ago. I would go in straight to the wall like it's clear. The importance of ads has grown a lot in particular because a lot of other levers have been

[53:28] removed like as everybody gets more and more automated in a bunch of ways. And most of the network these days just okay, tell me how much money you can afford and what is your goal. The only thing you have left is what you input as a creative or creative assets. So, this is your lever. This is how the battle is won or lost. And I think we should all pretty much have it put in the minimum requirements in the sense of what message do you want to put? And it goes deep inside two things, which is understanding your audience and what triggers them in this, what you have different, what is your

[54:01] USP, and so on. Which for me is something you have to think before you spend the first dollar. And then it goes on an iterative mode on what works and keep iterating on it. One thing that is a bit maddening and makes it very not scientific is that not that it's random, but it's unpredictable. I used to have a saying, which is ugly wins, but the truth is you

[54:24] cannot anticipate what's going to win. If you show me 10 ads and you ask me which one is going to be the most efficient, the likelihood of me finding the ads is at best the same as randomly picking, okay, let's say that one. And I realized also for me it was actually an advantage that I saw a lot of people thinking like, oh, I know this one is going to work better because of this and this and Because I knew from the beginning that I was really bad at anticipating which ads was there. I had zero like criteria about I'm going to show them all and

[54:53] made a winner win, no problem. I think here the creatives like they have to anticipate, but they're also really hard to interpret why it's winning, which is even more surprising. But the thing is when a creative starts to win, luckily for us, so it's complicated to design and prep, it's complicated to analyze in terms of why it's been working, but it's not

[55:14] complicated to see what is working. Because the platform they're incredibly good at showing the creative that works better. Insanely good at with just 100 or 200 impression I put 20 ads and they already detected that this one is going to get and it's getting most of the spend. Like that's basically the signal. Like the creative winner is the one that's getting all the spend. And 99% of the time the platform was right. They do have a couple of false positive and false negative, but that's the exception. So here, in terms of how do you deal with that, one is understanding your USP and your audience, one is

[55:46] really understanding the platform, like because kind of the format that can work for a Tik Tok is completely different that what works in the store and search ad and completely different that what works on Pinterest and on Instagram and on Maybe you can recycle a Tik Tok on Reels, but YouTube is going to be another beast. What is the context of this? Oh, people are playing that kind of game. I could put this. And here comes the problem that a lot of designers and app producers do not necessarily have the context of where the ad is going to be placed and at

[56:17] which point the user is going to see it. And that creates a lot of losing ads because this has not been factored in, even though it's still hard to plan, but at least like taking a bit of context. Then it takes a lot of testing. I mean, finding a winner is really tough. The thing is when you find a real winner and testing them cost money, but when you find a real winner, it's game-changing. I mean, but like really game-changing. I can't remember a couple time in history where we find like a It's not a 10x winner, it's a 100x winner. It can change the course of the

[56:45] company to find one. Like it's vastly understated how you could completely change the course of a company just by finding this winner. But for this, it means you deploy and it's a bit of a number game of you have to test a lot of things. One of the most common mistake in creative is people are testing 20 variants of the same concept. Like they're testing over and over the same thing with small variants. You have to test radically different thing. Is it a person or is it an animation? Is it provocative or is it really bland? Is it

[57:15] strong color? Is there a lot of text? Like there's so many elements that you can change that is not even the message you're communicating with the format. Like let's take a longer video, a shorter video. Do we start with the app or do we start with a person or do we start with whatever? And you have to be really radical. For me, the framework is really we're going to test a lot of different concepts even if they're not completely finished and polished and nice. Wherever we get a bit of traction, this is where we're going to add layers of let's do violence of this. Like there

[57:44] is a bit of traction there. Let's dig. Let's drill down. And okay, let's test in another order. Let's test longer. Let's test with another person doing the exact same. Let's take the same scene or let's change elements after elements until we polish this winner. And one thing you said is at the synthesis ugly wins that sometimes like the ads that are most performant they're not the ones that you would want for your brand to communicate. Maybe they're even completely off brand guidelines. I had a ton of problems with other teams say, this is what you're showing to users? Like this is detrimental to company in the long term. Like this is

[58:18] problematic. I say, yeah, but it's bringing twice as much the money. So no-brainer. I'm going to keep showing it. He's actually finding ways of okay, here I've got a winner. How can I realign it a little bit with I know this concept is winning, but also it might have consequences in the long term that I'm not seeing because UA is very short-termist. So not only try to find variants and the elements that work, but actually also realign it with the message that the company wants to give. That's really a complex topic. I personally predicted two or three years ago that we would start seeing the role

[58:53] of creative manager, I call it. There's not really any naming for it. And I'm surprised to see that it's not more standardized because the need for this is massive. Everybody is doing it, but it's completely unstructured. And you ask also a little bit about externalizing and so on. Personally, when I get this question, I'm like, I want both. I want to have a designer in-house. I want to call creators. I want to use agency to produce ads. And I want you to try that. Like shoot a video selfie on your own and let's Let's it works. I know a bunch of companies where the best performing ad is actually the

[59:28] CEO shooting himself like in a very poor environment, like old phone, whatever, it's blurry, whatnot. That often works. And the idea behind using different ways of production is that everybody's going to come with extremely different concepts and styles and so on. And that's the only way to find winners. If you always rely on the same person to create ads, no matter how good is this person, you're going to hit a ceiling because we're all limited in our creativity range, let's say. So, I basically want different people to do this. And this is why over the course of the last 2-3 years, not just because influencer marketing was growing, that

[1:00:06] advertiser used creators more and more not to reach their audience, but to actually use their ability to understand the channel to produce content. When we talk about user-generated content, like UGC, the ads that are created by influencers, mostly, it's actually creator content. Because they know the rule of the platform, they're really good at making these videos that don't look like ads as well, so people are going to skip less. And on paid social at least, it may not be true for every channel, but it's clear that content created by creators is what has been winning in the last 1 or 2 years. That's what the younger generation wants to

[1:00:43] see. And it's not the only way to do ads, but if you're not doing any of it, you're missing out on something huge. That is fascinating. I haven't heard anyone quite go as deep into the reasons and the variety of things that need to be taken into account. I think there's so many people in the industry still thinking we need to figure this out in-house. Like our designer needs to be the one making the ads. And it's fascinating to understand why you should not try and build intuition because if you think you're building intuition, you're handicapping yourself because you're going to not do as broad of

[1:01:24] experimentation. You're going to say that ad's not going to work, and then it would. The biggest takeaway for me is just like don't rely on intuition. Don't try and make ads you think are going to work. Try and make ads that are ads. There's another argument I think this is interesting here to add a little bit of a closure on creative. But there's another element which is really the creative is the targeting. Like most

[1:01:48] platform you don't target people at all. Like on Facebook you can, but the most efficient strategy is just go broad. You just put the country and I want people to be at least 18, all of the US, and have an iPhone. And then the creative and the events that you're inputting are doing the targeting work. And if you only have one team, whether it's in a house or even an external, they're going to gear towards one concept that is maybe attractive, and you might find one winner this way or a couple. But you want also the diversity in concept because ads are going to work for

[1:02:22] different audience, and you want them at the same time in the account because Facebook is oh, that ads resonate to that kind of user, and that ads resonate to that kind of user, and you can group them differently, and actually expand your audience, and find performance in more groups that you would have find otherwise. If you always keep milking the same winner over and over again, you're also missing out on a ton of potential audience for your app. And so you have to go broad no matter how niche you are, there are always a ton of triggers to different people, and those they're not demographic criteria or

[1:02:58] interest. They're really what you're saying about your app, and how you're saying it in the format. So that's also why you want to never give all the creative to one person. I don't care whether it's in a house, even the creators, you want to test a lot of creators. Because if all your creators are done by one creator, you run into

[1:03:14] the same problem again. Yeah. Yeah. So, the one thing you did mention that is a bit of intuition and maybe something that should have some level of focus in the I creative is the unique selling prop. How do you incorporate that into this variety of testing? Like, in your mind, does an ad need to explicitly say something about the app bringing value? Does it need to say something specific? But, then each app has like multiple layers of unique selling propositions, right? Like, my

[1:03:48] weather app, I always think about this. Like, some people just like to talk about the weather. Some people care about whether they need a an umbrella to work today cuz they're walking. Some people It's like there's so many different kind of features that would resonate with different audiences. So, it's not that you pick one unique selling prop, one unique feature, and like all your creatives are just variations on that. But, how do you think about it? Like, ultimately, you're still selling something. Is there some like overarching principle that you think about that an ad needs to, at a

[1:04:17] minimum, communicate something? The simple answer is no. Like, there's no clear pattern of this kind of USB is definitely most likely to work or like this kind of expressing the USB. There are some verticals where I insist on this because you have to differentiate in markets that are more competitive. And so, it depends the phase you're in, but one typical example would be meditation apps where you could go 5 years ago with a completely bland message about, "Oh, we're doing meditation." It might have worked. But, a lot of people have tested them, and a lot of products have grown, and now I'm seeing founders come to me say, "Yeah,

[1:04:56] I'd like to launch this meditation app. Can you help me grow it?" I say, "Okay, what's the twist? Are you a meditation app for seniors? Are you a meditation app for kids? Are you a meditation app that is inspired by whatever philosophy behind? Or, what's your angle? What are you doing different from the others?" I say, "Oh, no, I just have great content around meditation. They said So you just have the exact same stuff that the other 500 meditation apps that have been running at billion dollars for the last

[1:05:21] 5 years? And I said yes. Okay, we can't work together. So I think in those vertical that extremely competitive, you have to differentiate with something. But it doesn't mean that your twist is what's going to work in the ads. Not necessarily. Sometimes the USP come after, sometimes the creative is just a teasing that is not expressing what makes you different. You're not saying it there, you're going to say it later

[1:05:42] in the app store or even in the product. Maybe you can find success with a creative for your weather app that is just showing a funny situation with umbrella of people losing their umbrella and then the transition is well done and people discover later that your app is better for weather because of XYZ. Like it doesn't mean that the USP needs to be in the creative. But I think you need to know your USP to test after

[1:06:08] what users, see what makes them react. Like it leads you to their interest and very often the best ads there are bit indirect. They're not about oh my app is doing this, you want to download it? Never works. But say okay, people were interested in that, they very often talk about the rain. Find funny stuff about the rain, put it on TikTok. Find hot facts about the rain and put them on Reddit or Twitter. Find, you know, oh that's what they're interested about. This is going to be my hook. The hook is like kind of the first second or three seconds of your ad. It is extremely important

[1:06:41] because if people don't stop in their scrolling, you're dead already. You need them to stop on your ads. This is fundamental. And very often if you talk about what you're doing, there's no way people are going to stop on your ad. So you need to find kind of a hook and then slowly transition and maybe it's not even in the creative, transition into what you're doing actually to solve that problem or to do it differently. And it's kind of tricky because sometimes what the user want to see is not what you're doing, and finding this way of transitioning is not easy at all. So, I don't think the USP needs to be in the

[1:07:13] creative, and I don't think there is a pattern, and I don't think there is a particular recipe. I've seen creative that are so different work for different people. It's amazing to watch. Like, I'm friends with the developer of Motivation app, and developers call Monkey Taps, very successful apps. And its best performing ads of all time, there were a screenshot of a notification. And I've tried this trick with other apps from not the same exact USP, but let's say approaching USP, never managed to make it work. Like, sometimes it's the time. Maybe you had something like really different in there, or maybe it was the first, or

[1:07:49] maybe what. And obviously you could see in the notification it's an app, but it wasn't particularly saying or how the app was better or different, and that ad just worked. Like, sometimes it's surprising to see what works, but it's also amazing to see that different things work for different apps. So, have to try a lot, like literally. It's hard to do any other way than trial and

[1:08:09] error. Is there an example you can think of where there wasn't even, you know, like you're saying, for my weather app I should try a TikTok ad that's just people talking about the rain or rain facts or whatever. If that's an ad, how do you even incent people to then go download the app? Are there examples of like you just don't even need to incent them to download the app? Or you did talk about like transitioning from rain facts to some mention of the app, but how does that actually play out? Cuz it

[1:08:40] just you're blowing my mind right now. About just running ads that are so indirect to the value prop of the app. How do you span that gap from this hook that gets their attention in the first 15 seconds to then they do still have to go download the app? Yeah, fair enough. How does that work? Sounds like you already owe me money all the input and now you want the twist to the secret sauce on top of it. Like

[1:09:06] you're asking for a lot. Now you're getting to the hard part. It's the same. Like there's hardly a recipe. I've seen ads that are completely indirect and actually not transitioning to anything. But just because the ads are cool, people are clicking on it or they're just curious. I don't like to do that because one it may not work, but also very often

[1:09:24] because it crashed my Insta rate hard. This something that is particularly tricky is you need to be a bit fun or provocative or even click-baity or extremely indirect to actually interest people on their feeds. But then eventually you need them to download the app as well. And this kind of relation and transition is really the hard part. Finding something funny to stop them in the feed is and you can clickbait all you want, but if you clickbait too hard when they're going to reach the app store, they'll say, "No way I'm installing this app. I

[1:09:53] was just here for funny cats videos." And that's kind of a bit the secret. I've seen some ads that work without much of a transition just because the ad was so good. But you can think of ways of inserting your product somewhere in the video. In some cases even just the video is there and the end card. So if people look until the end and there would be something that presented. I would try to think of something a little bit more subtle in transitioning into it, but it could just work of adding an end card at the end. Like just say, "Ah, we have a ton of this in the app." Even

[1:10:24] though it was just a funny video or you made a remake that's in the rain or whatever. And then you transition somehow. Like but you can be a bit abrupt in the transition because it's also part of the joke. I think some of the best ads they're actually in how they manage to find this funny moment or fun fact into where actually bringing to this. Sometimes a bit more obvious than others. I don't know. For example, some of the most successful Blinkist ads they're about

[1:10:52] showing famous CEO or whatever. This CEO is reading 500 books a day, and that's already so close to the value prop of what Blinkist is doing that they can transition very easily, and it's fairly natural. But, they find a way for people to be interested like, "Oh, yeah, I want to be as smart and successful as that guy." And suddenly, "Oh, that's the solution. Super cool. Like, that looks simple as hell." And with the rain, I'm sure you can find one. I'm more of a nerd than of a data guy, so I'm not particularly good on the content creation and on these ideas that really works, so I lean on other people to find

[1:11:23] it. So, there's not one recipe. I would recommend to transition somehow, but it doesn't have to be transitioning into this is what we do. Like, the USP, the exact USP. Just to transition into we're doing something around this problem that you're going to like. You can tease a little bit. I would recommend telling people that this is an app. Otherwise, when they get to the App Store, the rejection is hard. Like, they weren't expecting to go to the App Store, or they're not even making the Okay, this video was cool, and then move on. So, I would recommend to find some kind of transition, but you can try the brutal

[1:11:55] method. Might work, but I would definitely try to find a a slight transition and tease a little bit towards. Depends a lot on what you're doing here. So, that's hard to generalize as well. You mentioned briefly in passing that reducing your install rate is potentially a negative thing. Are you talking about that if you're running a very clickbaity ad on Facebook, and your install rate goes to crap, they're going to show it less because you're too clickbaity, and they have to show a million impressions for you to get one trial start, and you're optimizing for trial starts, and so that super clickbaity ad is just not delivering

[1:12:38] enough to train the algorithm, and for them to actually be able to deliver you the impressions. Is that what you're kind of getting at? More or less. There are two things that can happen that are both bad in this case, or that you need to know as a background. The first one is that maybe the ad is very successful because you've pushed it really hard. It can be because you're click-baity and faking it, or it could be because you really find something super funny. And then the engagement on the ad is going to be

[1:13:02] insane. People watch it till the end. They react, they comment, they like, they share, they do whatever. Like all the signals who create this are high. Like people stop, they talk about it. Like all those signals that give the platform, "Oh, people want to engage with this ad. I'm going to show more of it." And then you get more and more of these impressions, which eventually you're going to pay one way or another because no matter what the billing is made on, and you're told that you're paying for a trial or an install or whatever, you end up paying for install

[1:13:29] no matter what. It's just a conversion. But then these impressions do not convert into your action. You're not going to get the raw as you want from this, even if the traction is insane on the creative. Because at the end of the day, as the second thing is, the whole logic of these auctions is that Facebook or TikTok or whoever Okay, you want trials? What I want, besides user engagement, is to monetize attention. And on this segment, the benchmark is going to be $10 per thousand install, like revenue per install or revenue per million install, we say in this case, but I say, "Okay, so if the ratio to which

[1:14:08] you convert an impression into an install and a trial is better, then basically for the same money you're paying, the platform gets more money." So they're obviously going to prioritize you over every other advertiser out there. And they're even going to charge you less for every install and for every trial, and they're still winning on revenue on their side. If everybody clicks on your ad, but nobody installs, eventually your ratio is really bad, which means that the money you're willing to pay is little, which means the platform is not making money, and they're going to punish you for it at some point. Like early on, they move

[1:14:43] this creative a lot, you get traction on the creative because it's engaging, but eventually the unit economics doesn't work for the platform itself. Like it's kind of you want your goal, but one way or another, Tik Tok, Google, everybody, they're going to monetize the audience at the rate that I mean, it's also a balancing system. No, if everybody starts paying less, Facebook is going to accept that and they can not monetize as well. It's not that they fix the price of okay, that impression is $10. Like

[1:15:10] it's a moving dynamics. But you're in competition with like millions of advertisers. If your ratio of converting an impression into a click into an install into a trial is lower than your neighbor, you're either going to pay more for it or you're not going to get any of it. All right, Thomas. So, this was supposed to be like 45 minutes. It's going to be about an hour and 15, hour and 20 minute episode and we're not even halfway through what we wanted to talk about today. So, I've been threatening you that we should be doing quarterly podcast because it's just so insightful and we get so much great feedback on

[1:15:43] your episodes. So, if you're listening to podcast and you want more Thomas episodes, shoot me an email, hit us up on Twitter. There's so much more we could dive into, but we do need to wrap up. I think the biggest question people are going to be asking at the end of this episode is how do I hire Thomas? And unfortunately, I have to tell everybody who asked me that question,

[1:16:01] you can't hire Thomas. He's too busy. He's almost never accepting a new client, but I did hear recently you're working with a group of fellow growth marketers. So, you might have some friends who have availability even if you don't, but you do take new clients here and there depending on circumstances. So, if somebody wants to get in touch, what's the best direction to send them to work with Thomas or somebody like Thomas to figure all this

[1:16:29] kind of stuff out? Yes, thanks David for the nice words. You can try to email me or LinkedIn me, but actually I'm pretty bloated over there, so I would not recommend it. I'm really easily joinable on direct messaging Twitter. I'm in a bunch of Slack. I'm in the Sub Club forum. And I do try to answer to the best of my ability. To add work is complicated, although I do add some regularly. The reason is what we were talking before about the insight we make on the long term. They require that you work in the long term with people. I can't get them in the first week. Like I've digged into

[1:17:01] the data regularly and I also enjoy seeing the development of the ads on the long term. But I do know a bunch of people that work on similar things. Yeah, so ping me up. I try to answer everybody. Email is just not the best way. Yeah, it was really fun. You mentioned maybe quarterly or maybe another part. I'm going to tease a little bit. The topic I did want to talk today but I'm too talkative so we didn't get is the balance between freemium and premium and how hard you get with your paywall. So hopefully in a few months we

[1:17:30] have the chance to talk about that. That's one of the topic that is keeping me busy lately. So if you reach that far in the podcast and you are interested in that, do ping David so that we do prioritize it. Yeah, I know. I I was actually really excited to talk about that as well. We just did an episode with a current

[1:17:47] Headspace PM and a former Headspace PM. It was fascinating to hear Headspace's transition from being a very freemium product to being an almost completely paywalled product. And there's so many apps and so many examples of people experimenting heavily with freemium and premium right now. And so it's going to be a fascinating topic. So I think we just committed ourselves to doing this maybe in September or October. So Thomas, always a pleasure chatting and look forward to talking more in the

[1:18:17] future. Likewise. Always happy to be back. We still did a good one and we've got more topics coming. 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.

---

About this transcript
Read from YouTube's own caption track and laid out by ViewRank AI (https://viewrankai.com).
ViewRank AI finds the videos already beating a creator's own average on Instagram, TikTok and YouTube Shorts, transcribes them from the audio itself in more than 60 languages, and turns what worked into new ideas and scripts.
Free transcript tools, no account needed: https://viewrankai.com/tools
How to read any video this way: https://viewrankai.com/llms.txt