Podcast: MDM Mailbag #8 (with Thomas Petit)

For the eighth installment of the MDM Mailbag series, I am joined by independent app growth consultant and advisor Thomas Petit. As a reminder: guests for MDM Mailbag episodes of the podcast are practitioners, and questions are sourced directly from the MDM community.

Thomas and I discuss the current state of measurement, the necessity of incrementality testing, and the challenges of scaling creative production in an automated environment. We also dive into the technical and ethical considerations of web-to-app funnels, as well as the ongoing legal battles surrounding app store commissions.

Among other topics, the questions Thomas fields pertain to:

• Whether incrementality testing will ever completely replace deterministic last-click attribution models for large-scale mobile advertisers and networks
• How the proliferation of media mix modeling vendors affects the quality of measurement insights available to smaller app developers
• Why the mobile industry continues to struggle with ad fraud detection despite the availability of sophisticated third-party verification services
• If the current trend toward massive creative volume is actually delivering diminishing returns for performance marketing campaigns on Meta
• What the long-term impact of AI-generated video creative will be on the traditional role of performance marketing creative agencies
• When the friction of web-to-app onboarding funnels becomes a secondary concern compared to the benefits of merchant flexibility
• Whether Apple’s proposed fifteen percent commission on external links will effectively stifle the growth of independent web-to-app payment ecosystems

Thanks to the sponsors of this week’s episode of the Mobile Dev Memo podcast:

• ⁠⁠INCRMNTAL⁠⁠⁠. True attribution measures incrementality, always on.
• ⁠Branch⁠. Branch is an AI-powered MMP, connecting every paid, owned, and organic touchpoint so growth teams can see exactly where to put their dollars to bring users in the door and keep them coming back

Interested in sponsoring the Mobile Dev Memo podcast? Contact Mobile Dev Memo advertising.

The Mobile Dev Memo podcast is available on:

• Spotify
• YouTube
• Apple Podcasts

Transcript

Eric Seufert: Welcome to the Mobile Dev Memo podcast. I am your host, Eric Seufert, and I am joined today for the eighth installment of the Mobile Dev Memo mailbag series by Thomas Petit, who is actually the first returning MDM mailbag guest. Thomas, welcome back.

Thomas Petit: Hi Eric. I am very honored to return and very curious about the questions from the audience.

ES: As a reminder to the audience, these questions are fielded from the Mobile Dev Memo community. I am in the middle of a move, so I did not get around to actually publishing the form until Sunday, and sometimes I forget that not everyone is permanently online like myself. We did get some good questions. Before we do that, why don’t you introduce yourself to the handful of people who don’t know you. You are a very prominent voice in the mobile space, but for anyone who is unaware of who you are, you can introduce yourself and then tell me what you’ve been up to.

TP: I am an independent, or people call that consultant or freelancer. It’s been about over a decade, twelve years or something. I specialize mostly in subscription apps, so I don’t do that much gaming or fintech, but all kinds of subscriptions. I like to work with indie developers, middle class, and very big subscription apps at every stage, largely on acquisition, monetization, and the data underlying it, so a lot of attribution, product analytics, and so on.

ES: It’s great to have you back. I have been remiss in scheduling these, and I think the audience really loves the mailbag series because it’s much more tactical and practical. Let’s start with some fan-submitted questions. I thought this was a good one because it can be interpreted in a number of different ways. We will kick off with: Is incrementality the way?

TP: Yes and no. For a very long time people were not looking at it enough. Regardless of what tool you’re looking at, the CPI or CPA or return that you see measured is highly dependent on the attribution method. It’s only since it’s got a bit fuzzy for a few years and also because networks are modeling a lot more conversion that people got interested into non-last-click and whether this is really incremental. Would it have happened if I didn’t pay for that? Which I think is great, but it’s not enough.

For most networks, that’s not going to be enough for the network to actually run the ads, even if Meta has some incremental conversion features. You want for yourself in terms of budget allocation the incrementality of it, but you still need the deterministic last-click user level because incrementality is not user level, and you still need those for the mechanics of ad auction to actually happen for the next impression to show up. It’s never been a replacement; it’s something if you don’t have it for sure you’re making mistakes, but you can’t have everything on incremental alone just because Meta, Google, Applovin, TikTok, whoever is not going to be able to properly work from it. It’s still a bit of both, and that’s what makes the game interesting, looking at different methods of last-click and incrementality and MMM. None is right, but you form a better opinion when you pile several of them up.

ES: It is interesting how the sentiment has changed around this. There was obviously this push into media mix modeling a couple of years ago in the immediate wake of ATT. Media mix models are great; you could have one and it could be very helpful to your business, or you could do some kind of real-time incremental measurement like our sponsors offer, Incrementality the company. But it’s not that alone is probably not going to be sufficient. I think there’s a bit of a recognition that maybe the probabilistic approaches were oversold in 2021.

I don’t know how these media mix model vendors survive. There are so many of them on LinkedIn. I don’t know how there are this many of them. Do you remember in the earlier days of the MMPs, there were a lot, and they would have one or two big flagship customers? It reminds me of Mad Men. All those little agencies had their flagship customer that kept the lights on. It was Lucky Strike for earlier on in the Mad Men series, and Lucky Strike left the agency and it collapsed. I think that was the MMP space for a while and then it consolidated or just eroded. Now there are just so many MMM vendors. They can’t all be chasing the handful of scaled advertisers that would not be in that sweet spot of being too small to build this myself but too big to be ignorant of the fact that I need it. That’s a tiny market space to be going after, and there’s a thousand of these vendors.

TP: It is even more surprising that both Google and Meta have released an open source model out of it. I was discussing with somebody in this field not long ago and they were telling me over eighty percent of the MMM companies that you think of are just a layer on top of those tools, of Meridian and Robyn specifically. They just repackage it very nicely. I don’t think that can last forever. I guess the reason there’s so many of them is maybe because none of these solutions is perfect. Clients are always going to be unhappy one way or another, so maybe some keep switching. I’ve stayed a bit away; I’ve worked with a couple of the bigger players and on the open source models, but there are many of them and I stopped trying to understand the market for that.

ES: I had Michael Kaminsky on the show a couple weeks ago, and he runs an MMM company. We were talking about open source MMM and the calibration, and all that integration stuff is the product. That is integrating with the client and their data and calibrating and aligning the output with the business goals. I also listened to a podcast with Olivia Cory on the Marketecture podcast a couple weeks ago, and she made this point that they are best friends with the MMPs. They don’t compete with the MMPs; they complement them. You need an MMP; all their app clients have to have an MMP. If they didn’t have that, they’d have a lot less data to be able to use for the reconciliation. I think we have settled in this configuration where app advertisers are accepting and acknowledging that you need both, and one or the other alone on its own is not going to be sufficient.

TP: I’m more surprised about how many of them still run only on the MMP, which is kind of crazy.

ES: That’s also just a broken way to operate. You heard about the controversy with Bill Gates’s daughter running that cookie stuffing fraud scheme with her startup?

TP: No, I didn’t read about that one. I must have been X-ing too much.

ES: Unfortunately, I’m X-ing permanently. She had a startup that was like Honey, where you did rewards and stuff.

TP: Oh, I know Honey. It’s very big.

ES: There was another one that got caught in a similar scheme years ago. They were cookie stuffing, so they were basically taking credit for conversions they didn’t deliver. This was a tiny startup; they had no revenues and weren’t that many customers. The only reason it was noteworthy was because Bill Gates’s daughter was a co-founder. I wanted to post that meme of the guy with the noose around his neck, and these people freaking out about having gotten defrauded by this firm, and if you’re a media buyer on that, you’ve got some grizzled veteran who’s like, first time? This is why you need incremental measurement.

People complain about fraud, and obviously fraud exists. I wrote this piece in AdExchanger years ago that got a lot of hostility to it. Fraud exists, fraud is regrettable, and fraud is a moral atrocity, sure. But if you’re doing your job, it’s unfortunate that it exists and it’s a drag on the digital economy, but if you’re doing your job well, you should probably be able to avoid the overwhelming majority of it.

TP: I also have this stand where it’s basically your fault as a marketer if you get a lot of it. It’s your fault. You can blame the system, or why does evil exist in the world, but we know that fraud exists in every aspect of life. When there’s a lot of money on the line, it’s going to attract more of it, and you have to be vigilant. If you’re doing your job, you should be able to avoid the vast majority of it. What bothers me is these voices in the space that tell you that you’re not sophisticated or smart enough to detect fraud. I just reject that out of hand. They’re always selling a service; only they are smart enough to build the tools to detect fraud, not you, media buyer.

I usually give the example where we built this tool or technique and it was really interesting. We got invited to a very large company to present to their marketing org. I went in there thinking it’s going to be some conference room with eight people and a pizza. It was an auditorium on their campus of three hundred people. I just asked how many of you have PhDs, and it was more than half. So some startup influence or ad fraud influencer will tell you I’m the only person that’s capable of telling you how much fraud you’re encountering, and then you go into a room like that and you say, no, I think I’m good. I think we’ve got it. Don’t worry about us. It’s not even that you need a PhD or a massive team, but I think there are obvious, well-founded techniques to root it out, and to say that you need to subscribe to some service or have some consultant come in and charge you five hundred thousand dollars a year to run these systems that they’ve developed as a team of one is just absurd on its face.

TP: What I see as a lost battle is that when you get there, you’re fighting on technicalities, on how it’s actually happening and detecting it. I guess the fraudsters are always going to be faster and most efficient in finding another way before it’s detected. I’m not interested in how it’s happening. I’m just interested if stuff I buy is largely incremental and there’s twenty percent of fraud mixed inside of it, well it’s still incremental as a whole. I consider that trying to fight it is going to cost more effort and money than actually knowing. I just want to be aware of the channels that are largely fraud, and I don’t need to know what they do, the technique, the exact thing, and those detection systems are pretty complex. I don’t need that. I just need to know that there’s a lot of it there and I can see it because the conversions are non-incremental, which I don’t see on last-click but I’m going to eventually see running any other method like incrementality, geo tests, whatever. It should be obvious enough that I don’t need a tool. I’m not arguing that some case might be helpful, but I’m trying to weed out a part that is so big that you don’t need this to see it. If you need it, then the amount of effort is so big that I’m not interested.

ES: Exactly, it’s just a very basic cost-benefit analysis. How much is this going to cost me to root it out, and what does your service cost, and at what point am I actually losing money here by rooting out this fraud? I read a piece in The Current this morning, and The Current is owned by The Trade Desk, so you have to interpret everything that you read in that publication through that lens. They were saying that there’s this trend now or there is this impulse now for companies to actually reduce the amount of creative that they’re launching. That this impulse to always increase the volume of creative had actually become destructive and now people are launching too much creative and people were seeing success in pulling back on the amount of creative.

My sense is that may be true for specific companies, but I think in general people are not launching enough creative in terms of truly differentiated concepts. A lot of times you see people saturating concepts with too many variants, but I still think in general people are under-resourced on creative. What’s your take there? How do you see the benefits of increasing creatives if you take a client at random? Were they launching enough creative, and is that still a growth lever?

TP: I’m totally aligned with you that it’s mostly not enough. It went out of hand because people are launching close variants and saying, look, I’ve launched a thousand ads this month, and I look at them and say no, you’ve launched eighteen ads this month, and eighteen is just not enough because your thousand ads are all the same, which is a bit pointless. The sheer number because it comes from above saying we need to have this minimum, look, this company does twenty-five thousand ads a month, we need to do like them, and it pushes people towards just the number, which is meaningless per se. But obviously almost nobody launches enough differentiated concepts, enough attempts, enough messaging, enough sub-niche and specific audience, younger, older, gender, type of players.

I guess the only counter in that is that we’re both thinking about pure performance here. It’s true that if you run more brand ads or you have the conjunction of brand and performance, you may want to really choose which ads you give with the brand messaging. But on the performance side, it’s almost never enough. We’ve gone too far in close variants, but I don’t think we’re anywhere too far in the sheer number. The bigger companies I’ve seen have absolutely high numbers of total creative, but within it, it’s still a huge amount of different concepts non-stop. It’s an never-ending process. I don’t think we’ve gone too far. I guess it’s going to get even further. Costs of creative are so way down, so it’s more about the technicality now. It’s not so much into how you produce them, but how do I come with actually new creative that are not a copy or clone of the previous one that is meaningless because it’s exactly the same one. The other one is how do you deploy so many creative because when the volume goes up, there’s a bunch of false negatives. You haven’t tried enough, but you obviously don’t want to put spend behind every single one of them because it costs a lot of money. It’s still tricky to deploy, but I don’t think the sheer number of creative is supposed to go down. I guess it depends how you define the creative, but no, I would mostly disagree with that.

ES: Again, you have to try to work backwards and say what is The Trade Desk trying to message here. I think The Trade Desk is not in a position to do the same kind of creative measurement curation that Meta is, and The Trade Desk just serves a vastly different client base, so it’s a totally different advertiser category. Let’s say you onboard a new client, you’re coming in, you’re doing an audit, kicking the tires, what’s the first thing you look at?

TP: Usually the first thing I look at is how is the data flowing to the ad platform. Let’s assume I’m coming on the UA side, on the acquisition side, and people are expecting me to check the ads dashboard or their own dashboard. Usually the first thing I look at is the data flowing to the platform good enough, so typically for Meta in Events Manager or in Google in the conversion tools. There’s often a mess there, duplicates, bizarre names, sometimes people are not optimizing to what they think they’re optimizing towards, sometimes just a very dumb conversion. I just want to make sure what the network is receiving, and then only I would get into results internally or ideally not in the ads manager but in their BI.

There’s a bunch of definitions there, a couple of patterns there would be people who are over-simplifying the proxy results. It’s a bit like what we mentioned about last-click and incrementality before. What the platform needs to operate and what you’re trying to achieve are two very different things, and if you try to conflate the two into one, it will be sub-optimal. That’s why I start looking at what data is sent to the platform because I want to know how they’re going to interpret it. And then secondly I look, okay, now that we’ve got this proxy, is it what you’re looking after? Because that’s not going to be good enough. You’re going to be using averages and you’re not going to look at how the cohorting is going and so on.

Those would be first checks. I do look at money distribution, what kind of, I also try to, I mean especially if I’m coming, I’m like, okay, where’s going to be the low hanging fruit? It’s either going to be that they’ve made a blunder, and very often that’s in integration and conversion signal and so on. The second one is okay, where is the most money going? Because that’s what I’m going to focus on. This small channel that is spending two percent of the spend, I’m not going to look at it first month because I’ve got bigger fish to fry first. I’m going to look for the big wins. Typically I’m going to focus on the one or two bigger channels and see. I’m looking for blunders very often. They often are. As an external it’s also good that you find these blunders in the first month and it gives you a lot of credibility because there was an instant win, and then only you can get into the high-effort wins that are going to take three months to build and stuff. But if you start straight up with the hard stuff, people tend to lose trust. They haven’t seen an impact within three months and they’re like, what is this guy? So I always look for a blunder basically. Honestly ninety-nine of the time there is a blunder.

ES: That’s such a veteran consultant; you just gave out millions of dollars in consulting fees to aspiring app growth consultants. That’s such a veteran insight, just how to optimize the engagement. But it’s true, and it’s not like it’s extractive or manipulative. You come up with a quick early win and it’s going to be a more productive engagement for everybody because you’re not going to be fighting skepticism. Everyone wins.

TP: There’s a trust factor. Somebody external comes in, you need to see that they’re really in, and the wins help everybody. Sometimes when you don’t find it, it gets tricky because then you get onto stuff that takes longer and you don’t have the buy-in to actually make it happen. Everything goes slower and is more frustrating. I don’t think that’s bad to do that. And on the contrary, I’m going to try to take what takes the lowest amount of effort first, let’s get a win, and then we move on to the further. Sounds like a logical way to proceed to me.

ES: You’re right, it’s win-win, it’s just managing expectations and establishing credibility, giving yourself a clear pathway to operate. We’ve seen this surge in apps submitted to the App Store. Vibe coding has given anyone the tools to make an app and I think that’s fantastic. I think that’s great. I think people have a very clear tractable pathway to bringing their visions to life now, and it’s just going to result in more people having more creative output and more choice and more personal fulfillment with their spare time. App store’s crowded, there’s a lot of junk, but now distribution’s become the critical skill. So let’s say that you’re talking to one of these people that’s launched their first app, some solo dev, they did all the coding in Claude, they’re very proud of it. Let’s say it’s a really great product and they want to scale it, maybe they want to turn it into a business. What’s the playbook? How do you go from zero to five K in ad spend?

TP: The zero to five K is kind of the worst. I do it for a couple of friends because I like them, but as a business it’s a terrible business. Nobody wants to be in that segment, basically because it’s the biggest effort for the least impact. You’re not going to be able to get paid properly, and it’s going to be harder. There’s no data, you don’t know where the creative win is, it’s really the hardest. So the first thing I’d say is have you tried anything that wasn’t paid that because maybe I’m going to see a signal that there is something, could be a messaging, could be a very particular audience that you may find outside of the normal path. And maybe cleaning the store page and whatever. But usually it’s going to be down to start posting videos on TikTok, and it’s TikTok not because TikTok is better just because there’s still organic distribution in there. You can have a fresh account and publish fifty crap videos and there’s a good one and TikTok is going to show it, which that used to happen on Instagram as well, doesn’t so much. So it’s much harder. And then when I find this angle, I’d usually put my money first on Meta than on TikTok. So use TikTok as a what is the angle, what is the messaging that gets people interested, then put the money behind it. I would most likely do it on Meta just because it’s not necessarily the easiest, but it’s kind of the one that had the algorithm that picks really fast into who’s going to be interested into that product. They’re just really, really good at it. Even though competing networks can be very good at it, and typically Google or Applovin are very good at it, but they require more volume to do that, so I’d rather do it later down the road. But yeah, my first reply is just don’t anymore because it’s very tricky. And the second reply is find a message before you throw the money out. It’s funny how many conversation I see among these zero to one people indie about how bad the LLM advice is on how to integrate the Meta SDK, how to set up your conversion, how to set up the first campaign. It’s really, really bad. And I guess that’s the reason most platforms are making it so easy to create your first campaign. Next, next, everything is set up by default, which is both good and bad, but I guess it opens up the possibility for the very long tail of advertiser to get started. LLMs are good at many, many things; doing the zero to one in UA is not necessarily one where they’re very good at.

ES: I agree that’s a purgatory for a business. The problem though is getting the scaling up the creative production piece. It’s the whole orchestration and the pipeline. It just takes a lot of unknown unknowns. How do I make video creative? How do I do that? And I’ve got to set up some form of attribution, how do I do that? Now I think if you’re going into the five K just start with Meta and stick with Meta up to ten K a month, and then you agree with the TikTok trying to get organic distribution is where you start.

TP: But even for Meta I would try a couple TikTok to see what’s speaking so that you don’t burn your five or ten K in the first creative testing to realize they were all shit. At least you’ve done the effort a little bit to see this is catching attention so at least you otherwise I’ve seen a lot of people burn their five or ten K because they haven’t done that, and they come to Facebook with their two videos, and their two videos are complete crap, and the first five K are burned already. That’s why even for paid I would start on TikTok organic just to say that one is a bit more promising, I’m going to put my base of my five K into that one rather than burning it out entirely.

ES: What tools are you seeing people use to generate creative? People talk about Higgsville as a creative production tool. It’s not. I don’t think anyone’s using that for creative production; it’s too expensive. People are using that for fake influencer videos. That’s not being used for ad creative. And that’s I think the problem that companies are going to run into at scale. That’s the scalable product; it’s ad creative that I don’t think people are going to go to tools like that for. I think they’re using it for influencer video generation, which you’re going to get a lot of people running small amounts of spend there and then not having any success and then stopping. I think the way that they’re projecting out revenues my sense is they’re making a lot of assumptions about the durability of scaling those tiny influencers into large influencers and there’s just by definition there aren’t a lot of large influencers. But what tools are you seeing people use to generate creative?

TP: I almost don’t want to give too many names because my experience has been it’s so fragmented and diversified. Obviously a bunch of people use the main models directly because anyway everything is based on the main models anyway. I see a lot of fragmentation in there because a lot of tools are good at just one thing. One is good to make variants, one is good to make the voiceover, so I see people using eleven labs a lot for voice only but then not for the video, and then they would use this to make variants and that. There’s one that a lot of people have used around me that is called Arcads that has grown quite a bit. I haven’t had the best result myself but I know a lot of my clients have used it at some point. And I see them high and I guess it’s because we’re still in there. To me it’s only about a year that we’ve got AI video creative. Everything that was before summer 2025 was pretty shit because the video models were not good enough. And that year has been a lot of experimentation of jumping from one tool to another, and I’m not seeing yet forming like and everybody has their own process and so on. I see people who work in this field, they’re trying to glue different steps into one process that they can package and sell to their clients because they know they’re not going to replicate a model and because they know it’s painful to say, okay so I’m going to have to generate the model, let’s say it’s a AI UGC, I’m going to have to generate the persona, and then there’s the script, and then I’m going to need to edit it, and then I’m going to need to make the voiceover and then the text over. And actually it’s the whole process of sticking this one tool for each that is quite painful. I’ve got a couple of friends that are working on that, but they’re very small tools and if I look at my portfolio, I don’t see a very clear pattern that is established like that one is great to do AI UGC. And then I see also a bunch of agencies that had two tools for themselves and are slowly moving from, okay nobody wants to pay for my service anymore but they want to pay for like but I had to build something to stitch these different tools together. But they’re usually small. I haven’t seen any one of them in particular really become standard in my portfolio. That’s why I don’t want to give a couple names because I think there’s myriad of names. And they change every three months because oh look, the new shiny toy, and then I see everybody jumping on it and then three months after nobody talks about it anymore, which is very hard when you build a SaaS in this environment right now because even if you manage to get the momentum and a lot of influx of leads or whatever, the whole castle can tumble within a few weeks because there’s a new shiny toy. It’s pretty tough. There’s also the fact not only because there’s a new shiny toy but because there’s a price war now, and a lot of people would switch just because the cost of credits cheaper on the other one and don’t really give care which output is exactly better. It’s just like, oh look, free credits over there, let’s swap. And so I keep seeing people swapping quite a lot. I think it’s not going to last like this forever, but because it was the first year for video, it’s still pretty fuzzy how it’s going to concentrate. I don’t have it very clear yet.

ES: That’s a very tough space to be launching into right now. It’s a race to the bottom on cost. You’ve got really performance open-weight models. If I was at a big company, big meaning upwards of ten mil a month mobile, I would distill Kimmy, K3, into a model that is fine-tuned on my own creative. But then you say that because we’re in August 2026 and it’s possible that in November you’ve got another name to throw. But that would buy you six months, I think. You have to do it every but if you’re a big enough company spending ten million a month, that’s worth you’re going to spend less doing that than you are running two thousand creatives a month through some outsourcer who’s just or through company basically that’s just wrapped around probably one of the frontier models that’s the enterprise plan. So you’ll spend less doing that. And maybe you get a year, I don’t know, at some point it just gets good enough. I talk about this a lot, but once you’ve done these integrations, often times you’ve validated this against a need and it meets the need and why would you ever upgrade? I do all of my transcription of my podcast with Gemini 3.1 Flash preview. Why would I ever change? It’s so cheap and sure I could go to Flash 3.5; it’s more expensive. Why? The transcripts are great. I have no problem with the transcripts. I’m never going to change that until Google deprecates it on Vertex AI and then I’ll just update to whatever the next most advanced model is, not the cutting edge. I have zero problem with my transcripts. I think they’re perfect. I don’t think you could improve upon them. So why would I ever change that unless I’m forced to because it’s costing Google money to process those queries, those prompts? And the same even goes for the video production, the clipping, all that stuff. I think it’s great. I have no problem and you could tell me well it might be eight percent better if you up okay so what, I don’t it’s not worth anything to me. It’s worth nothing to me to get fifty percent better because it works fine. It meets the need. There’s certain use cases where yeah, I would generate more revenue or whatever and then we could have that discussion, but for these things that are just satisfying the checking the box or satisfying the need, they’re serving the purpose and it’s binary in that way. I’m never going to upgrade, I’m never going to upgrade. And you might get there with creative. You might get there with creative production.

TP: We’re not there yet, and especially because the video generation costs a lot of money, but your analogy is true for certain niche. I see people who have their process for voice, and especially around translation. I’ve got this creative that works, I need it to have it in twelve languages now. I don’t really care if it’s five percent better left and right because they found something that kind of works, the process is there, and the time gain into just applying this and at some point they might switch, but we’re not there yet for video generation. But for scripting is also an easy one where it doesn’t really matter that it’s ten X better.

ES: Web-to-app. I feel like this feels like old hat; this is not a new field anymore. This is what everyone’s doing. Talk to me about optimizing the web-to-app onboarding. How do you think about that in terms of length or friction? How do you think about that from a measurement and unit economic standpoint? What’s the impact on the unit economics? Is there a universal optimum? I don’t think there is; it’s going to be very context-dependent, very client-dependent. But talk to me about the process you go through to optimize that web-to-app onboarding and then how do you think about that in terms of conversion but also the unit economics of those people, do they tend to be better? Can you overcome some of the frictions going straight app-to-app with just a more optimized onboarding?

TP: If I was in an ideal world, I would want my acquisition to be half-half, half on the web, half direct to the store. Less dependencies when one thing breaks, which always happens, I’ve got the other one. The reality is that it’s never like this. It’s always almost a hundred percent zero or zero a hundred percent. I’ve seen many clients for whom the web-to-app economic they tried really hard just didn’t work. I’ve burned myself over a million dollar on one that I was sure was going to work and it didn’t and that’s on me. But it always end up that one of the two is just a better fit for that particular use case. It’s especially true the more the time to experience the aha moment in the app is short, the harder it is to move to the web. So that’s true for gaming, that’s true for a lot of photo video editors that can show the wow moment instantly in the app. That’s true for a couple of utility apps where you don’t need those personalization quiz and so on. And then the opposite way in health and fitness obviously, a lot of education, and a lot of the look-maxing stuff as well, and dating, fintech, anytime there’s more complexity into the app, it kind of makes sense to do that. And I think the wrong conversation all along has been oh yeah, you’re going to pay five percent fee or three percent fee to Stripe instead of thirty percent of whatever, and this is such a bad reason to do web-to-app because conversion rates are different and it hides the whole problem around it. But it’s more I think one of the issues is actually the App Store itself is such a bottleneck. Even if you can change your screenshot and stuff, the experience is terrible. There’s not a lot you can say, and if it’s from the ad if you need something to warm up the user or an explainer or to show something, the App Store is a terrible place to do that. And so especially if you’ve got low install rate, it kind of makes sense. If you’ve got long onboarding, it kind of makes sense.

Sometimes I’m trying to watch what we’re doing, are we doing it for the hype of Webflow or does it actually make sense for the type of user we’re trying to reach? Older users are a bit easier to reach on the web, for instance. Does it make sense to have a long flow? Yeah, or once we’re there, I’m going to mainly check at some I’ve got a couple of numbers in my mind that should be acceptable. The drop on the very first page, so basically people who bounce, they click the ads and they would bounce on whatever you’re showing them first. A reasonable number there is fifty percent. Fifty percent bounce is okay; misclick, they were not interested, whatever. It’s actually hard to move it higher, and whenever you move it higher, you lose people later, so those are battles that but if it’s very low, there’s probably something to do. I’ve been working with a company recently whose bounce rate on that page is around eighty, eighty-five percent. There’s definitely something we have to do on that page to improve it. And then actually the next I look is not even how many people reach the paywall, so after this whole questions and stuff, but that one is not really one that I try to maximize anymore because what you’re doing through those flow is trying to raise intent. In which state of mind is the user going to be when they’re going to hit that paywall, which defines the conversion rate to pay. And actually there are some people you can actually easily trick them into getting to the paywall but they’re never going to pay. It’s a little bit useless to try to maximize from this fifty percent to how many see the paywall, but rather okay, how is this change impacting the payment conversion rate without touching the payment? Because that means I’m going to the right direction in selling my product better, in people believe the solution I’m giving is actually going to solve their need. So that would be the second metric. And some of these change are on the paywall itself, whether you use a weekly and what’s the price point and localization price point and the UX of the paywall and blah blah. But very often they’re on this page and it’s never like oh, that question is dropping ten percent user. No, that question is dropping nobody, but people arrive in a state of mind that they drain when they arrive at this paywall and they don’t convert. So very often I look at this. There are some tricks about I love this the Psych framework that was from Dari, it was posted on Andrew Chen’s blog like twelve years ago or whatever, that every action gives energy or removes energy, and I think it’s a very good way of framing it of you know am I psyching up people into believing what I’m believing what I’m selling is actually going to solve their need. And so changing some question into reinforcement stuff and curves that go up and social proof and whatnot. The next metric for me is really doesn’t bounce to pay conversion, and I’m not looking so much at the intermediary steps because they’re quite misleading. Unless there’s a huge bounce on one page, unless the paywall is really horrible, I’d look at the whole thing, and that’s up to the paywall and that’s the second thing I would like maximize on. Lately I’ve had this battle with a lot of people which is when you make the paywall harder you’ve got a few more people who are going to get in because they really want to see what’s left, but obviously you got more refund which become more and more of a problem when you’ve got a lot of chargeback. So one of the tactics you can try to fight the chargebacks or you can try to have just less chargebacks and be a little bit more freemium. The playbook here was like make the paywall complete harder and whoever is not paying, forget it I don’t care, and that’s how you increase conversion, which is true in the short term, but eventually having a button maybe later, having a second offer, letting eventually people who’ve gone through that whole flow download the app and all, I believe in the long term maybe beneficial, not for everybody but for some apps that one you get late conversions, two you don’t need all this CRM to do cart abandonment and so on, and three you’re going to get less chargeback. Chargeback are a problem. The same way Apple is going to go at you if you’ve got twenty percent refund on IAP, Stripe is going to go at you, other merchant of records are going to go at you. So you need to balance this a little bit as well. So that would be the third one. Bounce, full conversion rate from non-bounce to payment, and then how do I manage the ninety-five, ninety-seven, ninety-nine percent of people who actually don’t pay.

ES: It’s important though because I think there’s no free lunch here. You make the tactics more deceptive or you make it a little bit too frictionless to subscribe and your chargebacks go through the roof. That could be disastrous. I don’t think people realize that you have to pay a chargeback fee. So a chargeback you refund the money and you pay a fee. And it’s really problematic and you could try to fight it but it’s almost impossible. I’ve had very few chargebacks, I’ve had like five chargebacks, but it’s a very different use case. Everyone wants to pay on your stuff and it’s not you’re running aggressive UA into it. I’ve tried fighting it and I’ve successfully fought one. These are egregious. People are clearly lying. They obviously forgot to unsubscribe. I’ll email the people when I get the chargeback notice, I am happy to refund your money if you’re not happy with the product, I will give you your money back. Just please cancel the chargeback, and I think oftentimes people are just too embarrassed to do it and respond. So you try to fight it and it’s like I had this person once, they did a chargeback on the annual sub. I emailed them and asked why you did that, I would have just refunded it. You didn’t need to do a chargeback, can you cancel the chargeback please. And this person writes, well the sub renewed on Saturday, and so I saw the charge is pending on my credit card. And so I thought if I canceled it because the credit card payment wouldn’t resolve until Monday, that I effectively was canceling before the renewal date, and therefore you charged me after I canceled. And I looked this person up and they had an MBA. I was going to write back to them like if I publish this your alma mater would go bankrupt because no one would ever get an MBA from there again. But I didn’t do that, of course, I just let it go. I know that you know that’s not how credit card charges work. But going back to the subscription use case, yeah you make it too easy to subscribe, you’re just going to get a surge in chargebacks and it’s going to cost you money. It’s not just that you refund the cash; it’s also that you have to pay a fee on the chargeback.

TP: There’s layers and layers of consideration on those chargebacks. One is the cost of those chargebacks, but there’s actually a threshold until you get kicked out and you need to change the provider and so on. I think in the recent case, there was this very prominent case from the FTC against Genesis that has a lot of subscription recently. One of their complaint was that it’s hard to cancel. One of their complaint was that yeah the paywall is not very explicit and the full price is not mentioned. They had a lot of small points here and there which I believe actually don’t matter, even if they’re true and it’s okay. The real problem the FTC had is that they constructed a whole network of sub-payment companies behind with the goal to actually dilute the chargebacks around accounts because if they were operating through one the chargeback rate was so high that the payment provider would have kicked them out. I think they were the main problem for the FTC of you’ve designed a system to actually hide the fact that your chargeback rate was very high and to escape the consequences of it. I don’t know if it’s true, but it’s my interpretation of it. Let’s say that you’re not as smart as Genesis that has built, has executed in a way that I believe was actually really well done and they’re very smart people, but you’re not them. And it’s going to bite back. And if you’re smart as them it still bites back because then you reach that level where the FTC comes back. There are much bigger problem than people realize. They’re also a problem on IAP, it’s just Apple deals with that problem and usually the consequences are pretty harsh and you’re, oh I’ve been kicked out of the store for no reason. Yeah, show me a little bit, I don’t think that was no reason. But you never read the story about oh Stripe kicked me out for no reason, but there’s a lot of people who get kicked out of Stripe or even other payment provider as well just because and then you go to more shady and more shady payment provider that would accept let’s say higher refund rates.

ES: I just have a visceral cringe reaction when people say that they prefer web-to-app because it’s harder to unsubscribe, that you can side-step Apple’s very easy to use unsub tool. That’s just no way to run a business. First of all, it’s just unethical; I think you shouldn’t be making a product that people don’t want to use, that you know they don’t want to use and that makes money just as a function of being complex to unsubscribe from. But there’s just no free lunch, like even if you took that approach, you’re going to pay the piper eventually, you’re going to get chargebacks. That’s no way to run a business ethically or just ruthlessly; it’s just no way to run a business.

TP: I counter that sadly people who forget are a reason that many app developers and I’m talking on IAP alone here are actually surviving. Without the people who unsubscribe a lot of app business wouldn’t operate, so it’s a bit of a sad reality, but then it doesn’t mean you need to maximize this cohort. That’s just bad. But it’s not the only reason people move to the web. I’ve always been an advocate on audience and channel expansion, that the auction of Meta is not the same when you run web traffic and app traffic. Google auction is completely different when you run app campaign and non-app campaign. So for me this was one other reason. And even on payment themselves, even if it’s not to make it hard to unsubscribe, but the flexibility that you have on payment is just so much better on the web and in particular to do upsell. Many apps that I know that run those web-to-app funnels, once you get somebody to subscribe, they just said yes, really easy moment, are you sure you don’t want this PDF on top, and this method, and this body scan, and this supplement, and this, and it’s a very good way to actually increase LTV that is really hard to do as an IAP or at least the friction is higher. That’s a very good one. But also the management of those payments, you can do it in an ethical way that is better for you and the users. I have something for example that I remember once where we were making it easy for people to cancel, but when they ask, we had this like would you be interested if I refund you half the amount and you keep the whole subscription, and we had a giant number of user say oh yeah that’s awesome I want that. Apple is slowly moving into that direction with this new retention messaging API that they’re offering. But it’s just so much more flexible doing it on the web, there’s so much more. And I wish developers see the expansion to web not as a fee stuff but as a feature expansion and audience expansion. I hope Apple and Google sees it and say okay we need to do better, and I think Apple has started to get better on this but they still have a lot to do like on refunds, on upsells, on the win-backs, in this case it’s not even win-backs it’s like you win back before you lose them. Hopefully there’s the fee itself and I hope the IAP are going to standardize at fifteen percent one day or another and I’ve always advocated for this, but I also hope it we get closer to feature parity with the liberty of how you can handle your customers on the web that honestly is just better at this point. It’s better for the user to be on IAP because it’s centralized, it’s everywhere, sometimes for very small developer because all the tax, all the refunds, everything is in one place, Apple does a lot for you. If you’re a large developer, I hope we’re going to get a lot more features around payment on IAP that incentivize developers to also stay on IAPs.

ES: I’ve always said that you should not move from app to web to save the commission because you’re going to be encountering a lot more friction that you’re probably not accounting for. You should not move from app to web because it’s just inherently harder to unsubscribe; that just makes you a revolting person. You should move there because it gives you so much more freedom and latitude to do merchandising, to actually provide a personalized product that’s impossible to do in the App Store, and therefore to provide more value to the user. So you can counter the friction with just providing more value and making the value proposition stronger and more compelling, and then you’ll overcome the friction. Quickly, I know we’ve got two minutes left, link out. We don’t know where we’re going to land there yet, but you can do it for free now on iOS in the U.S. because of the injunction. But Apple made the new proposal, fifteen percent on link out. Tim Sweeney’s objecting to it. I have a feeling, and this is just an intuition, that Judge YGR is going to accept it. What do you think if fifteen percent’s the link out commission, is that viable?

TP: For some it’s viable, those who retain, but the enforcement of this is going to be a mess like on actually proving that this oh this came before the store, this came after the store is going to be so messy. Just the finance through it of justifying what should be getting fifteen percent and what not, it’s such a mess. I think it’s been a huge win for the developers while it was at zero percent. Some developer will still gain with ten or fifteen percent on it, but a lot of developer just for the hassle of justifying I say fuck it, I just do IAP and pre-App Store, but it’s just too much hassle to do it. So only the bigger developer with most retained users and so on and that are able to develop like this extra value because personalization and so on, it’s not effortless. I think it’s only going to be good for the bigger developer. And the whole battle of Epic from the very beginning was allegedly on helping the small developers, which I think for the small developers is going to be like that’s it, it’s too complicated.

ES: Don’t get me started. I think Tim Sweeney has just dropped the mask completely on this. He clearly just wants the App Store commission to be low. I understand his battle, but his battle is not the small developer battle. That was the banner that they were flying originally, but this is clearly just for Epic to make money on an Epic Games Store that charges slightly less than what Apple charges on commission.

TP: We all defend our little piece of meat, you know, we all do it.

ES: Yeah, but we’re not so sanctimonious about it. He’s just lost all credibility here in terms of the actual reasoning for this crusade. I actually wonder how it’s viewed internally because my sense is it’s seen as a disaster. This is seen as tilting at windmills internally. They’ve spent allegedly a billion dollars on legal fees, they just laid off a large proportion of their workforce. I don’t think this has been seen as a successful effort internally. Thomas, this was great. Thank you. How can people find you? How can they interact with you? How can they hire you?

TP: They can’t hire me. They can interact with me in the MDM Slack. I’m on Twitter/X and LinkedIn. My LinkedIn inbox is pretty packed, so better you talk to me on MDM.

ES: All right, there you go. Well, what a seal of approval there. Thank you, Thomas. I appreciate your time, as always.

TP: Thanks.