Buyer Support
Seller Support
Consumer subscription apps are one of the most compelling business models out there — low overhead, high margins, and global distribution right out of the box. But even the best apps lose more than half their subscribers after year one. So what separates the ones that thrive from the ones that stall?
In this episode of Growth Stage, we speak with Phil Carter, founder and CEO of Elemental Growth and advisor to companies like Perplexity, Gamma, and ElevenLabs, about how AI is changing the game for consumer subscription growth. Learn how top apps are using AI to create better experiences, acquire users more cheaply, and rethink their pricing models — and what you can do to apply these lessons to your own business.
Podcast Full Interview: Audio
Podcast Full Interview: Video
Transcript
Jesse (00:04)
Hello everyone. Welcome to Growth Stage, a podcast by FastSpring where we discuss how digital product companies can grow revenue, build good products, and build the value of their business. I’m your host, Jesse Paliotto. I love being part of the community as part of what I do at FastSpring, and I love being able to hang out with folks like Phil here today. We’re gonna talk about how AI is changing subscription growth.
Phil Carter is here with us from Elemental Growth. Consumer subscription apps are one the most compelling business models out there. Of course, they’ve got low overhead, high margins, global distribution, right out of the box, very appealing. But even the best apps lose what is an astonishing number of subscribers every year. so what separates ones that like thrive and grow from the ones that stall? And so in this episode, we’re gonna talk about that. We’re gonna talk with Phil, who’s an expert in this, an advisor to companies like Perplexity, Gamma, ElevenLabs, companies we’ve all heard of, and about how AI is changing the game for consumer subscription growth. And then we’re also gonna look at how top apps are using AI
to create better experiences, get users more cheaply, think about pricing. And then the application for all of us is gonna be, of course, how do we apply these to what any of us are doing in our own business? So Phil, thanks for being here, man. Really appreciate it.
Phil (01:12)
Yeah, thanks for having me, Jesse. I’m excited for the conversation.
Jesse (01:15)
Hopefully I didn’t overpromise. Can we deliver? That’s gonna be the question.
Phil (01:19)
I’ll do my best. I’m here for it.
Jesse (01:21)
Right on. Right on. just a fun random question. what was the first subscription app that you ever paid for? Do you remember or was it so long ago that it’s buried in the you know the history?
Phil (01:32)
Sure, I’m not gonna get this right. I don’t remember with any precision, but I will say I ru I do remember this. My parents and my brother, I have a younger brother, all got in on digital meditation very early on, and specifically Headspace. And so they recommended Headspace to me. This was all the way back in I don’t know, twenty eleven, twenty twelve. and so I ended up subscribing for Headspace relatively early on.
And and that was great. And you know, continue to meditate to this day, maybe not as much as I should, but when when I can find the time. and then I I mean, since then I’ve paid for many, many, many subscription apps, but that was one of the first.
Jesse (02:07)
Yeah. Do you still pay for Headspace?
Phil (02:10)
I do. I do. ⁓ right on. I’m a loyal user. And I’ve and I at times I have supplemented it with Calm as well, but I would say Headspace is still my favorite.
Jesse (02:17)
Yeah, I’m a fan too. I’ve used it on and off at different times and when I do it helps. So I guess this is a shameless and unpaid plug for Headspace today. Take it as you will. can you tell me a bit about Elemental Growth and what you do there? Maybe ⁓ you can talk about clients, maybe, or just what you guys focus on. I’d be really keen to hear that.
Phil (02:35)
Sure. Yeah. So I founded Elemental Growth almost exactly three years ago now as a growth advising and consulting firm. Initially I thought it would just be me as a solopreneur, but it’s gone well enough that I’ve now hired a small team of like minded individuals who love working on consumer subscription apps and and helping companies grow faster. we’ve worked with close to forty clients since inception. That includes traditional category leaders like Chess.com and GameChanger, as well as some of the fastest growing AI companies in the world right now.
what you mentioned. So Perplexity, Gamma, ElevenLabs, Wispr Flow, Tolan, all clients of ours over the last couple years. you can think of us as sort of a growth team for hire. We come in and we help with primarily product like growth. So we’re not a marketing agency. We’re not we’re not helping you
maximize your meta ads. I mean we can do that, but that’s not our primary value prop. I would say where we’re best in class is on product like growth and monetization. So helping with things like optimizing onboarding flows, dialing in paywall conversion, figuring out the right price point and the right s subscription packaging strategies, and then helping drive more organic growth, whether that’s through share and referrals, word of mouth, ⁓ SEO, AEO, or App Store Optimization.
Jesse (03:42)
excellent. Yeah, this will be so relevant to so many folks who are trying to drive that PLG loop and try and, you know, drive subscription in a way that’s sort of efficient and also baked into their product in a way that feels like value creating. And I think we’re going to talk about that. actually that kind of gets me right into my next question was you wrote sort of a framework article. I saw the version of 2024, maybe you wrote about it previously before, you called it the subscription value loop.
for people that would have missed that or or wouldn’t have caught that, can you talk about that for a second and what is the subscription value loop?
Phil (04:11)
Sure.
Yeah, this is like my equivalent of the conjoined triangles of success from Silicon Valley. So I’ll be the first to admit that frameworks can be overused sometimes. But in this case, I think it’s really valuable. And the the origin behind the subscription value loop actually came from the time I spent building Quizlet’s product growth team from the ground up, which which happened between about 2018 and 2022. And what I noticed was you have different parts of the organization that are working on very different things, but they all need to fit together into this coherent system that drives growth for the business.
And so ultimately where I got to with this was at the center of any any business, but certainly any subscription business, needs to be a unique and enduring core value promise. It needs to be unique because the app stores have gotten increasingly crowded. This is even more true now with AI, right? Like people are vibe coding apps in weekends. So there are hundreds of thousands of apps across the Apple and Google App Stores. And your your app needs to have a unique enough value promise to stand out.
It also needs to have an enduring value promise because even if you deliver a lot of value to users in their first month or two, if they hit diminishing returns and they’re longer longer getting value over time, then that doesn’t really work for a subscription business model that’s relying on those recurring subscription payments. so at the center is this is the core value promise, needs to be unique and enduring. And then around that you have these three steps: value creation, value delivery, and value capture. ⁓ value creation is typically led by core product teams. These are the people who are actually building the products and features that
ultimately create value for your customers. So these are the product managers, the designers, the engineers, the data scientists. then you have value delivery. And so value delivery is typically led by some combination of marketing, growth product, or if you’re a B2B business, you may also have a sales team that’s helping with value delivery. And that’s all about cost efficiently delivering the value that you’ve created with your product into the hands of your customers and doing it as cost efficiently as possible. And then finally you have value capture. And value capture is about making sure that
You’re converting enough free users into subscribers and you’re generating enough revenue from your subscribers to have a sustainable business and to have revenue that you can reinvest into product innovation and into paid user acquisition. that is the subscription value loop in a nutshell, and I can go into more details, but that’s that’s how the framework works.
Jesse (06:20)
Yeah, that’s thank you for the kind of TLDR version. I know we’re gonna kinda Chat through a couple facets of it’s a really clean way to think about just the whole process and the the loop as you describe it. some of the data I think like I’ve seen you share is sort of brutal. That even the best apps the stat I’ve got down, and you can correct me if I get this wrong, is that even the best apps lose more than half their subscribers after year one.
if you’re somebody listening to this right now who has a subscription app, you might be like, Yeah, I know that. for folks that aren’t in that kind of seeing that data, that may be a shock. ⁓ why is it such a hard business to scale and maintain like that?
Phil (06:53)
Yeah, well and just ⁓
Put a little bit of a finer point on some those numbers. And obviously these numbers are changing all the time. RevenueCat puts out a great annual app growth report that I consider to be the best out there on this. But based on the latest data I’ve seen, so even 75th percentile subscription apps, those that are in the top quartile, are losing more than 50% of their annual subscribers after one year, and they’re losing more than two-thirds of their monthly subscribers after just six months. And so that’s just not really good enough, right? If you want to build a venture-scale business that gets to a billion dollar plus outcome.
Outcome, those numbers aren’t going to work because subscriber retention is the foundation for everything else you do. Another data point I’ll throw out is less than 6% of installs convert into subscriptions, even for those top quartile apps. And so really you have to be in the top decile or even the top 5% to make the numbers work. Adventure scale. Now you talked about at the beginning of the podcast. This is a really compelling business model, low overhead, turnkey support tools from the app stores, doesn’t require a lot of funding to get off the ground. So if you’re a solopreneur,
Or you want to bootstrap a subscription app and not take outside funding and get to five or 10 million in ARR, I think there’s a very clear path to that. But if you want to be a venture scale subscription app, you really have to be in the top decile. And at that point, you get subscriber conversion rates as high as eight to ten plus percent. you get subscriber retention rates that look much better, closer to 60% one-year subscriber retention and closer to 50% six month retention on your monthly subscribers. And that’s that’s a foundation you can build a venture scale business off of.
And so what the what the subscription value loop is attempting to do is distill down some of the specific framework strategies and tactics that top subscription apps like Headspace or Calm or Duolingo or Strava or Spotify or Netflix are using to just consistently and cost efficiently grow their businesses and get past the launch phase, which is easy, but successfully navigate through the scaling phase, which is where consumer subscription.
gets really hard. And then I’ve also, in partnership with RevenueCat, I’ve developed a tool called the subscription value loop calculator that lets you actually plug in your real numbers. So you can compare the performance of your growth metrics like subscriber retention or your prices or your gross margins or your cost per install, cost per trial, cost per subscriber acquisition. You can plug all those numbers in, compare them to benchmarks, not just overall, but in your specific app category and for your specific tier, 25th, 50th, 75th, 95th percentile.
And then you can see where the biggest gaps are, which can help inform your growth strategy.
Jesse (09:15)
Yeah, that feels like amazing thing to be able to plug that in and immediately see if I can just move this number, I can overall lift my entire like resulting financial, you know, ⁓ picture significantly. I would guess the you know, it’s it’s classic process theory where you get in and there’s bottlenecks. And if you can improve those, you open up throughput for the entire system. ⁓ you mentioned some companies right now, so maybe this those one of those is is the answer here, but I’m curious because I know like in that article you talk about
being good enough across all three steps, but really being elite on at least one in order to really drive this value creation or subscription value, excuse me. But can you talk for a second? What does like an elite look like for this? Is it one of the ones you just mentioned? And I I know you could probably go way into the weeds on but I’m just curious, like what is it what does it look to be just cranking on this thing?
Phil (10:01)
Yeah. Yeah, maybe I’ll give you I’ll give you a few examples, maybe one for each step. So value creation, I think everybody in this space has talked about Duolingo ad nauseum because they just do such a good job with so many things. But I think when you think about value creation specifically, Duolingo is a company that one
largely created a new category in terms of mobile first language learning, right? You already had digital language learning businesses like Rosetta Stone that were out there, but they were sort of this tired legacy brand and there wasn’t a lot of personalization. There wasn’t a lot of di dynamism to how you learn a language on Rosetta Stone. So Duolingo comes a along around, I think it was 2011 when they first got off the ground. They create this new way of learning a language that’s on a mobile form factor primarily.
And then they supplemented that with all of these gamification mechanics that made it much more fun and ⁓ easy and approachable for somebody to learn a new language. And then they augmented that with streaks and badges and leaderboards and point systems and just various types of rewards that really keep people motivated. One other thing they did that people don’t know as much of is that they have their own machine learning algorithm called Birdbrain that they use to really help ensure that every lesson
is fine-tuned to your individual needs as a language learner. Meaning which questions should you be asked, at what level of proficiency, and how do we make sure you’re in the sweet spot where you’re getting enough questions right to stay motivated, but you’re not getting them all right, because if you’re getting them all right, then you’re not actually learning anything. Yeah. And so I think they’re just a perfect example of a company that’s knocked knocked it out of the park in terms of value creation. ⁓ value delivery, I could give many examples here, but I think a good one is Strava. And what I mean by that is
Strava at this point is spending more and more money acquiring users through paid acquisition channels. But for a very long time, they were able to acquire users for free through word of mouth, whether that was offline at real events like marathons or 10Ks or online through people wanting to post their runs or their cycling trips and get kudos. And so they built multiple organic PLG driven loops that allowed them to acquire users at very low cost. And that was a very ⁓ key to their success.
And then finally, value capture. example I’ll use here is Tinder because the dating category is uniquely difficult when it comes to monetization. The reason being if you do a great job, then chances are a lot of your users are going to churn at some point because they’ll find somebody they like and they’ll go off the platform. And then maybe it doesn’t work out and they come back later. But the point is you you need to capture a lot of value from your users early in their customer journey as a as a dating app, because otherwise you may not get another chance. And t companies like Tinder and Bumble have done this really well by
One offering multiple subscription tiers at different price points, just means they’re able to capture more consumer surplus under the demand curve because they’re targeting different user populations with different needs and different willingness to pay with these different subscription tiers. And then on top of that, they’ve layered in these in-app purchases like Boosts and Super Likes, where you can pay additional money to buy consumables, one one-time consumables that you can use in the app on top of those subscriptions.
And so I could I could give many other examples, but I think Duolingo, Strava, and Tinder are are good ones across value creation, value delivery, and value capture, respectively.
Jesse (13:12)
You’re making me think Tinder and the dating apps are taking a page from the gaming apps and doing live op gamification of the dating space. So like buy a three day boost or whatever. You’re like, I think this just turned into a video game. That’s funny. Maybe not. ⁓ I wanted to dig into the value creation, like maybe we can take a couple minutes and look a little bit deeper, like kind of double click as they say, into some of the specific parts of the with value creation.
By the way, the the Duolingo one is such a great example. They have completely hooked my kids. ⁓ both my kids are learning multiple languages because they just think it’s fun and just don’t want to break their streaks. So if you’re talking about delivering something, especially with the use of AI, automagical in the first 30 seconds, I think that’s something you call out. What does that look like? I I feel like those that are creatively bent probably like will come up with a million ideas, but you know, some folks may be like, what actually is meaningful?
to create in an automagical way.
Phil (14:09)
Well, maybe I’ll start with Duolingo because like you, my kids are hooked on it. ⁓ we try to do an international trip every summer for a week or two. and so my daughters, Kenley and Liana, have been learning Spanish. And the this summer we’re going to Europe, so you know, they’re learning Czech as a new language. And ⁓ so they’re hooked on it. And part of the reason that people get hooked on Duolingo so quickly is because that first 30 seconds is just such a magical experience, right? Like
If you think about the way people have traditionally learned languages, you’re opening a textbook or you’re sitting down in a classroom or you’re installing Rosetta Stone on your desktop device. And there’s just a lot of like drudgery to it. But with Duolingo, you open the app, it asks you a few very simple questions about what language you want to learn, what your current proficiency level is.
Or if you want to, you can take a baseline exam and it will tell you what your proficiency level is, how much time per day or per week you have to devote to learning this language. And then they just instantly drop you into the actual experience of learning a language on Duolingo. And so within the first minute or two of installing the app, you’ve taken your first language learning lesson and you’ve experienced firsthand what makes Duolingo different. The fact that you are getting a personalized experience. It’s adapting to you based off whether you get questions right or wrong. It’s giving you these streaks and badges and various types of rewards.
So I think that’s a great example. Another one, a company we’ve worked with over the last year is called Tolan They’re one of the first and and fastest growing AI companion apps in the world right now. and they have this very novel concept of this AI alien friend where you come in, you meet this character called the Oracle, who’s this sort of like godlike figure that asks you a handful of questions about yourself. You know, it’s all cartoonish and it’s like very whimsical.
But then after it’s asked you some questions about yourself, you get matched to a Tolan whose personality profile has all these dimensions and it tells you why you got matched with this Tolan and and how your personality is unique and how it pairs well with this tollit’s personality. ⁓ and and then you’re dropped straight into an experience where you can get into a conversation with with that Tolan. And it’s such a it’s such a emotionally resonant experience that far exceeds anything I’ve experienced anyways, working with other
LLMs or or other AI products. And so I think that’s a good example. And then one other one I’ll throw out there just because I know some people will say, well, like my product just isn’t that interesting. Like I have a prosumer SaaS tool. So like can I really provide a magical first-time experience? There’s another company we work with called Wispr Flow, and their product is very simple on its face. Like they’re a voice to text dictation tool that just allows you to produce content much faster because you could speak at a much faster rate than you type.
And so especially with all the LLM work that prosumers are doing now, it’s just become a very efficient way to do your work. And they’re an example of a tool where if you check out their onboarding experience, you you would they’d be the last product you would think of in terms of being able to provide a magical first-down experience because it’s just such a such a like prosumer tool. But they’ve really injected so much personality into it. And and they they give you a few examples of like speak this line and and it intentionally has errors in it, but when you speak it,
Wispr Flow’s software is smart enough to get rid of the errors. And so as you go through the onboarding flow, you see firsthand how their tool is better than any other voice to text dictation tool out there. And by the end of it, you’ve like really formed a bit of an emotional connection with this product in a way that very few B2B products can achieve.
Jesse (17:34)
Yeah, that’s interesting. I mean, a couple of things that come to me while while you’re talking is one is there’s this
What does everybody love to do, which is talk about themselves. And if you get somebody to talk about your tell me about you, and I really want to know about you, that feels magical to me that somebody cares that much. Of course, converting it immediately into value that you drop me into the Duolingo thing that actually has the language I want, at the level I want, you know. The other thing ⁓ that that comes to mind is that the ⁓ you immediately show the specific value of the apps, like the Wispr flow one, where you’re saying they kind of highlight this is exactly the type of stuff we can fit.
Fix for you. Like, that’s cool. So like showing the real like aha value right up front. that that comes through in what those examples you were describing. ⁓ everyone says the app is personalized. So speaking of personalization, ⁓ what separates the ones that actually are doing this, especially that are using AI to do this, versus, you know, just sort of using it as a buzzword and saying, yeah, we personalize stuff, but aren’t really doing anything? Like, what’s what’s the difference there?
Phil (18:35)
Yeah, it’s a great question. I guess the first thing I’d say is personalization is very much a moving target. I mean, if you think about how the app store has evolved over the last decade, ten years ago, if apps were doing any personalization at all, it was like they ask you a handful of questions in your onboarding experience. Then you’ve got the loader screen where it says like personalizing your experience. And then maybe it’s personalized, maybe it’s not. Like I I won’t name names, but I I know of plenty of products that
claim to like personalize your experience and actually under the hood there’s there’s literally nothing happening. It’s just it’s just a way to sort of make the make the user feel like they’re being heard. ⁓ but over the last 10 years I think more and more companies began to take the Netflix playbook of like we’re gonna ask the user a few more questions and if we’re a content company we’re allowed to make some content selections and then we’re gonna use a machine learning or a data science algorithm to be able to connect them to the content or the features that are most relevant to them. So that was sort of like stage two personalization. But even there, I think
You’re you’re largely falling into a set of discrete buckets. It’s not like they have an infinite number of personalized experiences they can deliver to you. There’s there’s more of like a a set of buckets that you could fall into tied to user segments or personas. I think over the last couple of years since ChatGPT really came into the mainstream and LLMs have allowed for what I call hyperpersonalization, which is like personalizing the product experience down to the individual user, you’re starting to see the best apps take full advantage.
of that capability. And so I already mentioned Tolan. I think they do a tremendous job of this where your Tolan is literally like built just for you based off of your personality profile. And they they show, don’t tell. Like they actually show you which personality characteristics they identified in you and then how that was used to match to your toll and I think another good example is Runna, which is the health and fitness app focused on runners who are training for their next race. It could be a 5K, 10K, half marathon, marathon
But it’s typically a more serious demographic than your average Strava Strava user. ⁓ and so they have similarly taken an approach where based off of answers to onboard and quiz questions, they are hyper-personalizing the experience for you, figuring out exactly what your training plan should look like for your next race, and then it adapts to you. So every time you run, they’re looking at your race time, your splits, your performance, and then they’re adapting the rest of your training plan up into your run based off of
what they’re locking your performance at for for each of your training runs.
Jesse (21:03)
Yeah. All right. I want to put you on the spot because you’re making me think about fitness apps. And I feel like I don’t know if it’s still the case, but I feel like there’s a period in time where a lot of times the personalization would dump you right into a paywall. So it was ask all the questions and then now that we know about you, if you want to get access to your personalized workout plan, pay ten dollars. And like what’s your opinion on that? Because I know for my personal experience was like that was a turnoff. But you know, maybe I’m an outlier.
Phil (21:07)
Uh-huh.
Yeah, I mean I think it’s part of this broader question of when is the right time to insert the first paywall in your product. And there are a whole different variety of ways you could look at that. Jake Moore, who’s the founder and CEO of Superwall, talks a lot about different options for where to place the paywall. You can place it right at the beginning of the onboarding experience, which maximizes paywall view rate, but could also turn the user off if they haven’t seen any value yet.
You can put it at the end of the onboarding experience like Duolingo does, which has the benefit of allowing you to demonstrate value and ideally get the user to the aha moment first, but then you’re gonna sacrifice some paywall view rate because some users will have dropped off before they even get there. You could do it on subsequent app opens. So every time you open the app, you get a paywall view, or maybe every three or five or ten times you open the app, you get a paywall. Or you can tie it to specific feature usage. So if you have premium features that require payment, then
Every time the user attempts to interact with one of those features, you trigger a paywall. All four of those you can think of as like tools in your toolkit for when and how to position a paywall to a user. And they have different pros and cons and different strengths and weaknesses. ⁓ I guess to use your specific example, you know, you go through an onboarding quiz, which has become table stakes for almost any subscription app, particularly in webbed app flows, but even in native app flows, you get a lot of these onboarding quizzes. They ask you a bunch of questions.
nominally to be able to personalize your experience, but then the first thing you see is a paywall. And it very much can be a turnoff. I think it really just depends on the product. Like if it’s a product that requires a little bit more firsthand experience for the user to get to what Reforge would call the aha moment, where they really understand the unique and enduring value promise of the product, then I would argue you should push the paywall further back because otherwise you’re asking a user to pay before they really know what they’re even paying for.
But if it’s another meditation app or if it’s another run tracker or it’s another calorie tracker, like if it’s if it’s an app in a well-established category where they kind of know what they’re paying for, and especially if you have a free trial, then in some cases I think it can make more sense to be more aggressive about moving the paywall up because you’ll maximize paywall view rate and and Jake at Superwall says you want at least an eighty percent paywall view rate. And you you’ll probably get more conversion as long as users know what they’re paying for.
Jesse (23:41)
Yeah. No, I for those who are keeping track, I have completely messed up our loop because we should be going value creation, value delivery, value capture. I’ve kind of skipped us right down into the value capture conversation. So ⁓ for anybody following along who’s methodical in their outlines, yes. I just took us around the the shortcut here. ⁓ while we’re on value capture, you know, Bringing AI features into the, you know, are they behind a paywall? Are they up front? Can you talk a little bit about that for a second? Like what’s
How is AI forcing companies to change how they think about where they put, you know, those types of access to those types of features?
Phil (24:17)
Yeah, I mean I think the biggest change that AI is ⁓ imposing on the app ecosystem from a value capture and monetization standpoint is that the marginal cost of serving a subscriber is now non-trivial. It used to be that serving a marginal subscriber for a digital subscription app, a product like Duolingo or Strava, was negligible, close to zero.
⁓ if you were a meal kit company like Blue Apron or HelloFresh, or if you were selling a hardware product like Whoop or Aura, then it’s a little different because you have you have real costs associated with each subscriber. But for most digital subscription apps, that’s not the case. And so it used to be you didn’t need to really worry about cost. And that meant you could give away seven-day trials, fourteen-day trials, thirty-day trials. You could be very generous with your freemium tiers, which Duolingo famously did for years before they even introduced a subscription plan in 2017. So
That was the old way of doing things. Now with LLMs, the underlying costs to support AI powered features are non-trivial. And in a worst case, what that can mean is you could have a user come in, go through your onboarding experience, sign up for a free trial, or just use your product for free if you have a freemium offering. And they could rack up pretty significant amounts of underlying LLM costs before you realize that they’re never going to convert into being a subscriber. And at that point, they’ve lost.
A non-trivial amount of money for your business. And if you have too many of those users, then you don’t have a business at all because you’re upside down on your unit economics. So what that’s translating into is number one, more companies opting for multiple subscription tiers. I mentioned earlier that dating apps like Bumble and Tinder were one of the first subscription app categories to innovate around multiple subscription tiers. And you mentioned gaming. They actually, in many cases, took that out of the gaming industry where whales drive all the monetization. And so you have multiple subscription tiers in a net purchases.
Yeah, 100%. So dating has been doing this for a while, but now you’re seeing more and more other categories offer multiple subscription tiers, often with a basic tier that has non-AI features, and then a more advanced tier that has AI powered features, but is much more expensive. Duolingo did this when they launched Duolingo Max in 2023. Tolan has three subscription tiers. You have you have plenty of other subscription apps that are using multiple tiers. Another change is companies are getting a little bit less generous with their free trial durations and or their free minimum offerings.
So going from seven day trials down to three day trials, or ⁓ just paywalling more features so that you’re not able to use the app for free, or if you are able to use it for free, you have more limitations on how long you can use it for free.
Jesse (26:49)
Yeah, that’s interesting. I’d seen s sort of some of the dynamic towards shorter subscriptions. I didn’t tie it in my own mind at that point to, you know, you need to kind of cap the expenses that are going to get incurred. And this feels like such a timely I we didn’t really kind of plan to get into this. I don’t know specifically, but very timely thing because of all the news in the last few weeks where ⁓ there was some story that was floating around like some unnamed company generated five hundred million dollars in token usage with Claude. And so all of a sudden the conversation in the last two weeks has been, wow, like
It costs a lot to enable in that case it was an employee base on AI access, but in our case a subscriber user base, you know, equally is gonna jack up your costs. So this this is this feels very timely for what people are trying to process.
Phil (27:31)
I think that’s right. And I also think it’s inevitable that these big LLM labs are gonna start to raise prices at some point. And Claude recently announced that they were increasing prices not on humans typing on their keyboard credits, but if you have an open claw or another device using Claude tokens, that the cost of those tokens was going up.
And I think that you’re gonna start to see more in that over time. ⁓ because as the industry starts to consolidate, like right now, OpenAI and Anthropic and a number of other ⁓ larger LLM players are all in this sort of dogfight to get maximum market share. But at some point their attention is gonna shift from user growth at all costs to profitability. And when that happens, it makes sense that they would start to increase their price per token.
Jesse (28:23)
Yeah, it’s classic Silicon Valley playbook. We’re gonna come in, we’re gonna under we’re gonna fund it all with VC money, and then at some point you gotta pay. Yeah. And then it’s a different a different game. I even sense it, you know, in my own interactions with colleagues and folks that the conversation has turned more to how can I do this efficiently?
Rather than just how can I do this? You know, what safeguards do I do? And that, you know, I suppose and once again I’m kind of putting you on the spot, we didn’t really talk about this ahead of time, but like I’m I’m curious, like, is that part of even workflows for subscription apps? Is like do making our AI features as efficient as possible allows us to give them away as much as possible to private.
Phil (29:04)
Yeah, I mean it’s sort of economics one one, right? The more you the more efficient these companies and developers can be about supporting their AI features at low cost, the more they can pass those savings along to their customers. And that could come in the form of longer trial durations, more generous premium plans, or it could come in the form of just lower subscription prices for the premium subscription tiers. But all of that matters because in any given app category, you’re not going to be alone. Like they’re going to be
Dozens, if not hundreds, if not thousands of other apps competing with you. And so you’ve got to find a way to deliver maximum value at minimum cost. And one of the ways you can do that, which is becoming increasingly important, is using AI tokens efficiently.
Jesse (29:49)
Yeah. Yeah. I’m a guitar player, so I’m just thinking of the amount of guitar tuners out there. If anybody’s trying to do an AI equipped guitar tuner, you got thousands of competitors. So good luck. let me catch us back up into the loop and then we can kind of sort of head for the finish line here. The value delivery. So we went from value creation, I kind of itadly took us right into value capture, but if we go into the value delivery stage, ⁓ I know you talk, I th I believe in that article from a couple of years ago about using AI to build tests.
build and test hundreds of AI creatives. Does it just become like a race to the bottom when everybody’s got the same tools? And you’re that just becomes table stakes to advertise effectively within this environment.
Phil (30:29)
Yeah, I mean, I’ll be careful how I respond to this one because my dad was a lifelong marketer and marketers are very important. Their work is critical for distributing value to users. Having said that I’m I know. Well where where I’m going with this is I would argue that in some ways marketing has always been a race to the bottom, right? It’s like the more competitive any industry becomes, the more you typically tend to see players invest in marketing to try to maximize market share and steal customers from one another.
But in some ways it’s an arms race. And the you know, more money spent doesn’t necessarily lead to better results in in every case because you’re just sort of it’s a game of attrition. And so I think when you look at it through that lens.
AI tools like VO3 for video ads or ElevenLabs for changing ⁓ the voice of a spokesperson into multiple languages, or Suno for putting different soundtracks into your ads. Like all of these are tools that can allow you to generate an order of magnitude more ad creatives every month. But at some point, there is this battle of attrition where if all of your competitors are doing the same thing.
Are you actually gaining ground? And so I think the reason I love this question is because I do think what it means is there’s gonna be more and more of an emphasis placed on authenticity and quality as opposed to just sheer volume. Right. Like to some degree, brand impressions are a good thing because the average consumer needs I I’m not a brand marketer, but I
I heard at one point the anecdote was like at least 10 brand impressions before somebody really remembers your product and seriously considers buying it. So volume does matter, but in a world where people are being bombarded by more and more ads every day in their social feeds, on TV, on radio, you’ve got to get more and more creative to stand out. I think it’s part of why Gen Z in particular is known to gravitate more towards low production value, authentic ad creatives in the form of short-form TikTok and Instagram videos.
Now, AI tools could still be used to create ads that fit within those form factors, but I think the best companies are finding ways to use AI to accelerate the velocity that they generate at creatives without losing sight of the fact that they still need a unique and authentic brand and voice to be able to stand out in the market.
Jesse (32:48)
Yeah, I I completely resonate with that. Like if you’re speeding up operations, awesome. If you’re trying to replace human authenticity, it just won’t work. ⁓ people people are too smart, they will see it. ⁓ with with AI native apps, ⁓ kind of the punchline question here, although I’ll I’ll I’ll expand a little bit, is it actually cheaper to acquire users for? You know, there’s things that drive user acquisition like PLG loops, word of mouth, user generated content.
AI is a fun thing to talk about in the last twelve to eighteen months. If you are an AI native app, do you actually get slightly cheaper acquisition because people are more willing to post about you, talk about you, recommend you? Or is that is that a figment of my or somebody else’s imagination?
Phil (33:34)
I mean the short answer is it depends. I think on the whole, there has been a period, and I don’t think the period is over yet, although it’s probably coming, where simply by virtue of being an AI product, you’re more likely to get organic user growth because people are talking about you, right? That could come in the form of offline word of mouth. It could come in the form of people posting about you on social media. It comes it could come in the form of PR because
journalists are writing about you or you’re on the news. ⁓ so there’s that bucket, which is like AI is so hot right now. Prosumers don’t want to get left behind in their careers. Consumers want to try the hottest new AI product. And so you’re getting a free tailwind from that. Yeah. On my mind. So there’s that bucket. And I do think that that is valuable, but it’s not really controllable. And my guess is like any tech wave, there’s this window of opportunity where there’s a real tailwind and at some point that tailwind will start to subside.
Jesse (34:17)
That’s
Phil (34:31)
You know, I would argue that the AI era really started in terms of consumer software when Chat GPT broke into the mainstream in late 2022 and early 2023. So we’re like a few years into that wave. And My guess is there’s probably not all that much more time left before you start to see the tailwind subside. It’s it’s not like it’s not going to exist at all, but the impact of it will start to subside over time. I think the second bucket is more controllable, and that is if you can use AI.
And apply it in a way that creates this automagical experience that people haven’t experienced before, similar to the first time you use ChatGPT in 2023 or 2024, or similar to the first time you pick up Tolan and meet your AI companion. Then I think you can not only improve the efficiency of your ad spend because people are clicking through on your ads at higher rates, but you can also then downstream of that.
increase your trial start rates, your trial conversion rates, your subscriber conversion rates, even your subscriber retention rates. And this goes back to how the whole subscription value loop works as a system, because if you’re reducing cost per install on value delivery and you’re also increasing subscriber conversion retention under value capture, then your LTV over cac ratio is higher, your payback period is lower, and you just have a much more efficient business that can grow faster. So I think that second bucket is the one that I would really be encouraging companies to focus on.
And one specific example I’ll give is there’s this company PhotoRoom that is photo editing software, and they ⁓ introduced an AI powered background removal and background alteration feature a couple years ago. And I had their ⁓ head of marketing ⁓ on my podcast at one point and he talked about how it so significantly reduced their ⁓ acquisition costs.
And so increased their subscriber conversion rates that they were able to unlock international markets like Mexico and Brazil and Indonesia that were previously unattainable for them because the unit economics just didn’t work. But because this feature performed so well in ads and converted so well downstream in terms of subscriber conversion, suddenly the unit economics worked in these countries where it previously hadn’t worked. And so I think that’s a great example of where if you lean into compelling AI features, it can really make a difference.
Jesse (36:52)
That’s really powerful. I we’ll grab a link for that episode because that would be, I think, fascinating for a lot of people to see how like such a key feature can unlock markets and unlock ⁓ or unblock subscription ⁓ rates. ⁓ so just to sort of maybe head towards a finish line a little bit, you know, this is this is a good ⁓ thought.
Like where you were at with Chat GPT coming out in twenty twenty two, twenty twenty three, and the kind of the arc of the ⁓ sort of wave of AI. When you wrote value loop, the subscription value loop article that I saw was a twenty twenty four. What do you think’s changed most about it since then? It’s been two years, I think, at this point, or nearly two years, the AI wave has crested and continues to to push. What’s what’s changed do you think since you first thought through this model and this framework?
Phil (37:38)
Yeah, I mean in a way I think we’ve hinted at a lot of these things throughout the conversation today. I I would say on the value creation side, it’s just become faster and easier than ever to create software. It used to be you needed to be a technical founder or you needed a technical co-founder who could build the software with you or for you. You needed, in many cases, non-trivial amounts of venture funding to get to a minimum viable product that people were going to actually use and to get it to market. And those days are gone.
Now, right? Like you can be a vibe coder and hack together an app in a weekend. And it may not ever be a venture-scale business, but like you can get real customers and real subscribers and real revenue very quickly. And that’s great on the one hand, because it means that it’s leveled the playing field and it’s democratized access to ⁓ building mobile applications. On the other hand, it means that
the app ecosystem, which was already extremely competitive, just became even more competitive. And there there are now I I I did another episode with with the VP of growth at ⁓ RevenueCat who talked about how in the last few years the number of new apps entering the app store every day, every week, every month has gone up. I think it was 7X over the last few
Jesse (38:53)
I saw this ’cause I think I caught them speaking at Business of Apps in New York last fall, I think, and or maybe it was early this year. Anyway, that number, I heard that. It’s insane. It’s amazing.
Phil (39:05)
Yeah. And so what that means is as a consumer, it’s great because you have more choices. It means there’s more downward pressure on pricing. So you’re getting better products at a lower cost. But as a developer, you have to be really thoughtful about how you’re going to create a product that isn’t just going to be instantly replicated by dozens of other developers. So I’d say that’s the biggest way that ⁓ value creation is changing. And that’s why delivering a great experience in the first 30 seconds and then hyper-personalizing the experience for each individual user is so important.
On the value delivery side, I would say that’s probably of the three, I would argue valid delivery is the one that has changed the least. But
We talked about how these AI tools like VO3, Eleven Labs, Suno, and there are many others, are just making it much easier to pump out significant amounts of ad creatives very quickly. And I would also say that another thing that’s changed, which is less directly related to AI, but more and more user acquisition is shifting from classic ad platforms like Facebook or even Instagram to a platform like TikTok, which is
less driven by the size of your follower base and and or the amount of money you’re willing to spend and more just driven by the algorithms and how good your creative is. And so what that means is even as a relatively small player, you can compete if you just get a creative that goes viral. And that that could be an ad or it could just be user generated content that talks about your product. And so I think
On on the value delivery side, I would just say you need to be using these AI tools to give yourself the best possible chance. And then you need to make sure that you’re using channels like TikTok that have uncapped upside and can really allow you to go exponential even as a small startup. And then value capture is the one where I think AI is causing the most profound changes because of the core insight that marginal cost of subscriber is now non-trivial. And so that’s where things like multiple subscription tiers, shorter trial durations, less generous freemium offerings.
Also, I I another thing I didn’t talk about as much is more I see more and more companies opting to use longer tail LLMs in certain cases when it’s good enough for their use case, right? Because using the latest ⁓ Anthropic or Open AI model is nice from a sheer quality standpoint, but it’s increasingly expensive. And so if you if you have a product use case where you can use something like LM Studio and identify a longer tail LLM that’s lower cost, and you can get just as good or close to a good performance.
then that’s something I think more and more companies are doing. And and Gamma famously did this early on when they were getting started. They they weren’t always using the most powerful LLMs. They were just using the ones that worked best for their use case. And that allowed them to both save costs but also speed up the the performance of of their products.
Jesse (41:43)
Yeah, that’s really interesting. ⁓ so if someone is listening to this conversation right now today, they run a subscription app. We talked through a lot of different angles here, but what do you think is one thing, if they could only remember one thing from this conversation, what should it be?
Phil (42:01)
I mean, I guess I’ll start with a bit of a cop out answer, but I think it’s the most important, which is you have to build a product that provides a unique and enduring core value promise. It’s at the center of the subscription value loop and it’s the foundation for everything else. And it’s more important than ever before because of the point I made that the app stores have just gotten so crowded. So I think that would be number one. Yeah. But number two, and maybe this will be a little bit more satisfying, is don’t get addicted to paid ads too quickly.
I think the mistake that a lot of early stage app founders make is it’s so easy to just pump a bunch of money into Meta. They do such a great job with their ad algorithm and finding new customers. But as you scale, inevitably your acquisition costs are going to go up because your marginal user is become become lower, lower in 10 over time. So your your acquisition costs tend to go up and your customer lifetime values tend to go down as you move past your early adopters and your ideal customer profile.
And so you need to be building the muscle early on to figure out how to scale through organic channels like word of mouth, like SEO or AEO, because ultimately that’s just gonna make it much easier to reach venture scale ⁓ hops. And that’s why companies like Duolingo and Strava and Tinder have been so successful. They it’s not that they don’t use paid ads, but they figured out how to grow virally or through SEO long before they started pumping tons of dollars into paid acquisition. And I think that really served them well.
Jesse (43:25)
Yeah, that is that’s that’s great. ⁓ if people are wanting to follow what you’re up to, kind of get more of these insights, ⁓ you know, or what Elemental Growth is up to, what’s best place for people to keep up with you?
Phil (43:37)
Yeah, so I’ll say a few things. One is I have my own podcast called Subversive. It’s on YouTube. It’s also on ⁓ Apple Podcasts at Spotify. ⁓ we do an episode every couple weeks with a top founder, CEO, or product marketing or growth leader at a consumer or prosumer subscription app. we’ve had guests on from almost every single one of the companies that I’ve talked about on the podcast today. And so ⁓ check that out if you’re interested. ⁓ beyond that, I teach a course on consumer subscription growth through both Maven and Reforge. And then if you ever want to talk to me about our growth advising and consumer consulting services. You can just reach me at Phil G Carter across all of the major social networks, LinkedIn, X, Substack. It’s just Phil G Carter and I’m happy to talk anytime.
Jesse (44:17)
Awesome. Well, thank you, man, for talking here. This has been really interesting to me. I think there’s gonna be a lot of insights for folks that take a listen. And then thank you to everybody who’s been joining us today on Growth Stage Podcast. Again, I’m your host, Jesse Paliotto. We’ve had Phil Carter with us today. I’m grateful to get to support digital product community through this sort of a thing and to get to hang out with smart and cool folks like Phil here today. So have a wonderful week. Big thank you to Phil and we’ll catch you guys all on the next one. Cheers.
