Note: We’re excited to announce the second edition of our live workshop series, Mobile Growth Lab, to help marketers, leaders and execs prep for a post-IDFA world! 

In the first edition, we helped over 40 attendees: 

• See the map
• Prepare the groundwork
• Move forward
• Find acceleration 

We’ll cover all that (updated for the reality of the post-IDFA world) and many requested topics like web-based flows, creative strategy, conversion values, prep for iOS 15, and more. 

ABOUT ERAN: Linkedin  | Singular
ABOUT BEN: Linkedin  | AdColony
ABOUT PEDRO: Linkedin  | Liftoff
ABOUT JOHN: Linkedin  | Forbes



ABOUT ROCKETSHIP HQ: Website | LinkedIn  | Twitter | YouTube

KEY HIGHLIGHTS

🍎 ATT adoption trends : unmeasurable inventory, censorship metrics, extended probabilistic matching.

👎 How advertisers are coping with broken measurement – and marketers’ wishlists for Apple.

🧩 Probabilistic but non-tracking methodologies.

🔐 Picking right conversion models.

📲 The buzz around owned media channels and first party data.

🌟 The ascent of web-based flows

🌠 Consolidation in the ecosystem.

🤖   What’s happening in Android – and what’s coming

KEY QUOTES

Advertisers use probabilistic matching

Pedro Ponce de Leon: [Advertisers] may also be looking at SKAN, but SKAN is more secondary because they’re actually turning on P-matching (probabilistic matching) using this for 14.5 plus devices. And you can see there that it’s actually growing; 80% or so of our spend is utilizing P-matching.

Big social media publishers don’t support probabilistic matching

Eran Friedman: But it is worth keeping in mind that some of the biggest publishers out there — Facebook, Google, Snapchat, Twitter — don’t support probabilistic matching at all. You can’t really run any campaigns with them. So, from a macro point of view, possibly, a lot of the spend out there can’t really work with probabilistic matching. We’re working with some of the top brands out there and a lot of others just can’t afford to activate it because of legal or business concerns around this. In general, I would say that looking at the privacy trends out there, I wouldn’t count it as a long-term solution but definitely something to be aware of.

Blended CPAs have not changed, but Facebook’s reporting has changed dramatically

Shamanth Rao: Recently, I was looking at an account whose CPA on Facebook pre-SKAN/ pre-ATT was $40, while post-ATT, this was $2,500. That was the CPA increase reported by Facebook. And when we looked at the blended metrics, the blended CPA had not changed. It was exactly the same. That’s the challenge.

Ad publishers are hiding data from marketers

Eran Friedman: Basically, Apple is censoring some of the post-install events. So, some of your installs are generating post-install events but it’s hiding it from you. So, they basically see crazy CPAs, but that’s because 80% of their installs are censored. So, they can’t really see the true actions that their users are making. The problem gets harder in SANs, as Shamanth noted, because usually, a single Facebook campaign is running across multiple campaigns. So, the bar is even higher than just 30 installs per day. Facebook, for example, recommends a minimum of 108-128 daily per campaign.

Maximize signal density

Shamanth Rao: Something I would also recommend is that people try and maximize signals, just because this is one of the ways you can address the privacy threshold challenges. I’ve seen a number of SKAN reports where your bits 4, 5, 6, basically, there’s just no value because those just never get fired. You have a super upstream purchase event in there. And what’s the point if it’s never getting fired? So, you want to set up the conversion value schema so as to maximize signal density.

A wealth of data to a very small stream

Ben Holmes: One marketer went from 71 post-install events down to 2. It’s just a massive drop in how they do that: post-install conversion and the funnel they were trying to get down, but it’s the only way they could start to rebuild that dataset.

RESOURCES

Webinar deck

FULL TRANSCRIPT BELOW

John Koetsier: If you’re here for the State of iOS User Acquisition Webinar, you’re in the right spot; it is going to be good. We have some amazing panelists, we have some great data, and there’s some really cool things we can talk about. 

I, honestly, cannot recall a crazier year with bigger changes in mobile marketing. We’ve seen massive changes in platform spend, iOS, and Android. Of course, it’s all happening in a crazy global environment. The goal here is to provide some help, some insight, maybe a shoulder to cry on.

We’re going to talk about how to maximize insights from SKAdNetwork (Storekit Ad Network or SKAN), and why that is so critical right now; we’re also going to chat about some winning strategies. We’re going to talk about the latest UA trends that we’re seeing from the very best mobile marketers.

We have some amazing guests. We’ve got Shamanth Rao, who’s the CEO of RocketShip HQ. He’s also the host of the ‘(Mobile) User Acquisition Show’ podcast. He’s great, insightful, and you’ll enjoy him. We also have Ben Holmes who is the SVP, Performance & Exchange at AdColony. Great to have you, Ben! We also have Pedro Ponce de León who’s the Director of Sales Engineering at Liftoff. He has some data that I think will rock some worlds. And, of course, we have the Singular CTO and resident SKAdNetwork expert, Eran Friedman. Now, one note about the panel, it is an all-male panel, which is not usual for Singular. However, I have challenged Singular to have an all-female panel, next time. And of course, I’m the moderator. I am John Koetsier. I write for Forbes, consult with some tech companies, and do a bunch of other things. Eran, maybe, kick us off here. Can you introduce Singular for anyone who may not know the company?.

Eran Friedman: Thanks, John. Hey, everyone! Singular helps brands grow faster by unifying all their siloed marketing data sets into a single source of truth. We provide the next generation attribution and analytics solution that pulls data from all your marketing sources, connects it with attribution, and provides the insights on your best performing campaigns, creatives, and publishers to help drive growth. We work with the top brands in the industry. 

John Koetsier: Thank you so much, Eran. Pedro, you have some super interesting stats and trends that Liftoff has seen. We would love to have you walk through those right now.

Pedro Ponce de León: Sure. I’ll get right to it. So, this is all from Liftoff’s purview on what we’re seeing in the marketplace as we run a DSP, and what we’re seeing in the supply, and also on the demand side. In the first chart, the blue on the right is showing the adoption curve of iOS 14.5+ devices. So, that includes 14.6. The surprising part was actually that 14.5 was not the OS version that actually saw that spike in adoption; it ended up being 14.6, where Apple really pushed it on users. So, 14.6 is really the OS version where ATT (App Tracking Transparency) actually kicked in. So, now it’s sitting close to 80-85 (percent adoption).

This slide here is showing the continual adoption on the supply side because there’s work on the supply side that publishers, exchanges have to do to actually support SKAN. So, this shows that adoption on the supply side is growing. You can see it’s hovering around 70% right now; and it will continually grow. And hopefully when Apple actually starts to enforce some of their things, the supply side will speed up on the adoption. That should get closer to 100% (adoption) as things progress. So the next slide here, at Liftoff, we look at IOS supply in three main buckets.

So this is before iOS 14.5, that’s that dark blue line. You can see that supply is diminishing. But then we have devices on 14.5+. These are devices that are subject to ATT, but we’ve got two lines that break those out. We’ve got the green line that’s growing, and then, the red line that should, over time, actually start to diminish. So, the green line basically represents 14.5+ devices that are SKAN compatible. So, that’s going to keep going up. And the red represents the 14.5+ devices that we see on supply that are not SKAN-compatible. 

That red line is significant because, from our point of view, and our reading of Apple’s terms of service, that the red line basically can’t be attributed by anyone not even SKAN or the MMPs. That’s because with the ATT prompt, if people opt out, most of that supply is not going to have IDFA in your tech; you’re not allowed to do p matching. So, in theory, this is unmeasurable inventory or non-measurable inventory.

We’re monitoring this to see that sort of disappear over time as supply comes into the fold and then becomes compatible. For this chart here, we’re looking at only 14.5+ devices, and using the presence of IDFA as a proxy for how users are responding to the ATT prompt. If it’s missing, then, they’re saying no to tracking; if it’s there, we’re seeing about 25% of supply has IDFA presence. We had predicted that 80% of people would turn down tracking.

In terms of Spend by Platform (pre-ATT), iOS had a bit of an advantage. They (iOS) were having closer to 60% of the total spend across both platforms. Post-ATT, there was a shift there; now, this figure stands closer to 50-50 (percent of ad spend across both platforms). This could be either because Android devices are growing faster but it also could be due to the changes that are occurring with iOS and SKAN.

Censorship is a topic that many advertisers are interested in, given that Apple is censoring the data in the post ATT world. There’s two censorship metrics – there’s the site ID that Apple censors, and that’s hovering pretty high (80% censorship) when you look at it at the campaign level, in terms of the post install events that we see, without site ID, conversion censorship rate is hovering pretty low  at 3-4%. 

If you look at the next slide, we have a pretty good understanding of what the censorship’s threshold is for SKAN. To explain this slide, the red line here is showing, at which point, if you look at all our campaigns, censorship goes down to zero. Essentially, it’s missing the vertical line at 30 installs per day. Where you start seeing that blue curve hit the red line, that is 30 installs per day. So if you’re able to get attributed 30 installs per day, you pretty much have zero censorship, and this is per SKAN campaign ID. So this is across all our campaigns; it’s a very clear threshold here of 30 installs attributed per day, and the censorship goes away.

Pedro Ponce de León: Next slide. So, this is  showing, from our point of view at Liftoff, how many advertisers are actually using what we call extended p-matching, but essentially, they’re using P-matching as their primary source of truth for metrics for iOS campaigns.

They may also be looking at SKAN, but SKAN is more secondary because they’re actually turning on P- matching (probabilistic matching) using this for 14.5 plus devices. And you can see there that it’s actually growing; 80% or so of our spend is utilizing P- matching. And again, this includes 14.5 plus devices. So, you can see that advertisers are essentially still using good, old P-matching.

John Koetsier: Thank you so much, Pedro. I’m sure there’s some marketers there that are using probabilistic (matching) and fingerprinting, and other things like that and they’re going, ‘Am I the only one?” Eran, any thoughts when you see that?

Eran Friedman: Yeah, for sure. We see that many of the long-tail kinds of advertisers in the industry are definitely still using probabilistic matching (P-matching). I can definitely see how, from Liftoff’s point of view, how this is trending.

But it is worth keeping in mind that some of the biggest publishers out there like Facebook, Google, Snapchat, Twitter, don’t support probabilistic matching at all. You can’t really run any campaigns with them. So, from a macro point of view, possibly, a lot of the spend out there can’t really work with probabilistic matching. We’re working with some of the top brands out there and a lot of others just can’t afford to activate it because of legal or business concerns around this. In general, I would say that looking at the privacy trends out there, I wouldn’t count it as a long-term solution but definitely something to be aware of.

John Koetsier: Yeah, absolutely. The bigger the brand, the greater the risk, right? And we know that there’s probably an end of life scenario there for that kind of P-matching perhaps in iOS 15 as Apple adds more and more privacy- focused features in there. Ben, any thoughts when you see those slides?

Ben Holmes: Hey! Yeah. Definitely! We’re seeing similar trends. One thing I wanted to point out was actually the SKAN adoption. It’s interesting. You know, we’re actually seeing a much higher SKAN adoption rate of 90%. And Pedro, I think your slides were saying 60-75% or so. And that’s by measuring by actual impressions, right? So, with the actual impression showing up, if you include raw app numbers and you look at just the sheer number of long-tail apps, it stands at 50%. But most of those apps are just never-opened, never-used, but in terms of actual impressions a paid media team can get access to 90% SKAN adoption.

Ben Holmes: AdColony is an SDK (software developer kit) provider, we’re a network. We have a DSP (Demand Side Platform), but our main product out there is our SDK. We help partners monetize because we get a little bit closer for that input of P-list management. So, a performance DSP – that’s one step up like a Liftoff or, you know, the other large players out there. If that P-list ID is not there, then it’s not going to look like it’s SKAN-compatible when it actually might be, you know, just depends on where that is. Everything else is pretty similar. We are closer to that 21% opt-in rate. So, we’re closer to that 80% opt-out; (79.7% when I checked yesterday). We are very heavy in the gaming space as a non-social player in the internet market. So, I am not surprised, it’s in line with what everyone expected.

Pedro Ponce de León: I would like to clarify that that chart was based on bid requests, so that’s the possible impressions we could buy; the actual impressions we end up buying is higher.

John Koetsier: Shamanth, any initial thoughts from you on some of those slides as well?

Shamanth Rao: Yes, definitely, we are seeing extended P-matching, it does happen. The one thing that is somewhat frustrating is that the folks who don’t adopt are losing out. So that’s a dynamic that is present. We just tell everybody that asks us, “Look, we don’t know how long this is going to be sustainable. It is risky territory.”

John Koetsier: Well, we’ll get deeper into it as we go. We’ll move on, however, there’s kind of a dichotomy that we’re seeing. Most users have adopted, they’ve updated. Some advertisers seem a little late and the competition on some of the iOS 14.6 campaigns is low. Why is that? And what do we see changing in the next few months? Shamanth, maybe kick that off.

Shamanth Rao: Yes, certainly. Just going forward in the next few months, we are seeing advertisers become increasingly comfortable with, what I like to call, broken tracking on iOS14.5 plus. Something I like to say is tracking is more broken than performance. That’s actually true across verticals that I’m seeing. What I basically mean is if you look at platform-reported CPAs, metrics generally look terrible. I know Pedro talked about the censorship rate at about 30 installs a day; that’s true for SKAN campaigns, unfortunately for SANs (Self-attributing Networks), you don’t know what the SKAN metrics are.

Shamanth Rao:

John Koetsier: I have to ask for a clarification.Did I hear you correctly that there was a CPA that you saw at $40; it went to $2,500? Did I hear that right?

Shamanth Rao: Yes.

Pedro Ponce de León: I can corroborate. I’ve seen the same.

Ben Holmes: We saw an account with a CPA of $100 that went up to $1400, so this is normal.

John Koetsier: Wow! And yet, Shamanth, when you looked at other data, that was just a figment of the data, it was not the real data?

Shamanth Rao: If you are comparing Campaign ‘A’ versus Campaign ‘B’ within a network, you can trust that number, but you don’t want to treat that as a source of truth.

John Koetsier: Yes. Eran, dip in here. What are you seeing here in terms of the dichotomy of this data and some of the sources of truth that Shamanth is talking about?

Eran Friedman: I think it’s definitely super interesting. We see things as welcome, especially around SANs. It’s an interesting evolution that the industry is going through because in the beginning, it was simply about getting the data, and looking at it; everyone got their SKAN reports available.So, I’m looking at the data now to see crazy numbers around CPAs and such. I think that the next phase in the evolution is heavily focused on the privacy threshold concepts.

Basically, Apple is censoring some of the post-install events. So, some of your installs are generating post-install events but it’s hiding it from you. So, they basically see crazy CPAs, but that’s because 80% of their installs are censored. So, they can’t really see the true actions that their users are making. The problem gets harder in SANs, as Shamanth noted, because usually, a single Facebook campaign is running across multiple campaigns. So, the bar is even higher than just 30 installs per day. Facebook, for example, recommends a minimum of 108-128 daily per campaign.

So, those kinds of practices and knowledge are critical to know how to actually optimize and measure your SKAN campaigns. The industry is kind of evolving to get that.

John Koetsier: You can do a little calculation there between Apple’s 30 per campaign ID and Facebook’s 180, and see how campaigns compare.Very interesting. Eran, we’re going to stick with you. We’re going to talk about this shift that we’re seeing, we’re going through from user- level data to aggregate reporting. How are marketers managing that shift? Because the data environment is so confusing right now. Eran, we’ll give you first crack at that..

Eran Friedman: Yeah, so, I think it’s progress, right? It’s like an evolution here. So, the first one was getting the infrastructure right.The companies have been shifting from trying to rely on device-level to aggregate because with Facebook and Google, for example, you just don’t have a choice. So, the first phase was building the reporting and BI from scratch, ready to rely on this, and stopping to rely on the IDFA (Identifier for Advertisers). The focus, now, is really on dealing with the data itself; figuring out why the CPAs (Cost per acquisition) are so high and how to optimize these. So, we are looking at all the privacy thresholds, monitoring them, optimizing them and getting better performance or better visibility to your performance with them. It’s kind of an interesting phase as marketers are now dealing with.

John Koetsier: Ben, your thoughts on that?

Ben Holmes: Everyone’s seeing the exact same drop in performance. Everyone wants the ability to just know which creative is working best. We have machine learning deployed to be like, ‘Alright, well, this engages with this most likely, it was the one that drove it!’ But it’s not for real. And that’s one thing that we don’t get from SKAdNetwork. We don’t have the user level; simple things like blacklisting or exclusions, they just stopped working. There’s all new methodologies that everyone’s testing out. When do I stop showing an ad? Maybe, they already converted and maybe they didn’t. I have no way to know. And that alone is just inefficient media spend and driving up costs across the board. According to our paid media reporting, everything is awful. Nothing works, everything’s broken. But when I look at my raw data, we’re continuing to grow at the exact same pace we were (before). The transactions and everything is exactly the same. So, one’s wrong, I think, we all know which one it is. And that’s why people aren’t relying on it, yet.

Ben Holmes: I only know of one example where an advertiser just straight up said, I’m not doing SKAN. Everyone, for the most part, is still running SKAN because they have to figure this out. We’ve already all said it, a P-matching, probabilistic fingerprint, whatever you want to call it – it’s temporary. We all know Apple’s trying to do something so that they won’t allow it to happen because it’s impossible to monitor and to play whack-a-mole. So, they’re all testing it out, but unfortunately, nothing’s really shown to be like, yes, this is the silver bullet, or this is what’s working. Everyone’s slowly getting better, but no one wants to make that hard shift just yet.

John Koetsier: Pedro, I think we’ll turn to you for a second and talk about what needs to improve for marketers to be able to make better use of SKAN, what needs to happen on all the sides. Pedro, you first.

Pedro Ponce de León: Yeah, sure. Apple is making some progress; they’re committed to this. You can see that with the updates. They’re providing advertisers with SKAN postbacks directly, instead of relying on media partners to share those postbacks; they’ve added VTA (View Through) attribution – that was something that was missing. You can tell they are actively working, but there’s a lot of things that are missing. One would be just more clarity around how attribution actually works. There’s a lot of advertisers that suspect that Apple is essentially trumping attribution for SKAN, even though they don’t use SKAN themselves for those services, so just clarity would be helpful. But, in essence, going for SKAN, we lost a lot that was standard and we need to see some of those things return, even if they return with some compromise. So, I can call them out but being able to set VT/ CT (View Through/ Click Through) attribution windows, cohorted performance data, true ROAS reporting versus ROAS cutoff at a certain time period, ability to run in some privacy-friendly way, like incrementality tests – we lost that. So, you know, we’ve got a lot of things that are missing that will make this usable. And I really do hope that Apple works on a bunch of these.

John Koetsier: Shamanth, same question. What needs to improve for marketers to be able to make better use of SKAdNetwork?

Shamanth Rao: I share a lot of elements of Pedro’s wishlist. I just don’t know if I’m optimistic that Apple is going to provide a lot of things like cohorted reporting or incrementality tests. The one thing that I would definitely like to see is improvements on the privacy thresholds. But just to echo Pedro, even clarity around the privacy threshold would be extremely helpful. We all talked about how we have seen thousands of dollars in CPAs; what would just be helpful is make the SKAdNetwork actually usable. That’s really the primary one, just so we can use whatever Apple is providing the way it was designed to without just second guessing ourselves.

John Koetsier: Yeah, I think you’re making some assumptions there when you say use it the way it was designed to.

Pedro Ponce de León: The challenge is like how much can they provide and still keep their privacy priority.

John Koetsier: Pedro, I want to stick with you. You showed it on your last slide there, it’s probabilistic matching, and let’s just be honest, it’s fingerprinting. There’s a significant amount of that going on. As Eran said, it doesn’t happen on SANs, which is huge. It doesn’t happen on all networks and it’s not all advertisers, some of the top advertisers, the biggest ones are not doing it, but many of them are. What’s going on there? What’s happening? And do you see a future there or not? Pedro, you get a first crack at that.

Pedro Ponce de León: I think we’re all in agreement that Apple’s intent is that probabilistic matching is not allowed. If the user says no to tracking via the ATT prompt, you shouldn’t be able to do P-matching. Most advertisers’ reading of their intention is the same. This, to Liftoff, effectively means that for the most part, the only Apple-allowed effective way to attribute performance to in-app paid media campaigns is through SKAN for 14.5 plus devices. So, the future is SKAN. It’s just a matter of when that’s going to be the case. Apple doesn’t want probabilistic matching to be happening if the user opts out. The reality, however, is largely due to Apple’s inaction on policing P matching, now, it’s a difficult thing. Given a choice, most customers are electing to continue to use probabilistic matching. That’s what we see at Liftoff and that’s what that chart shows. It’s mostly intuitive given that, like right now, there’s actually not much downside to opting to use P-matching, and there’s actually lots of upside. If Apple were to enforce ATT, customers would likely only risk a temporary rejection from the app store and you can get around it, correct it and resubmit your app. So there’s minimal risk and a lot of upside. So, currently ATT enforcement is delayed as far as we can tell. We don’t know when Apple will take action, they will. There’s a strong chance that they’re going to do it and they’ve been steadily investing in SKAN, so, you know, they’re committed to it. What we did at Liftoff, and, maybe, as an industry, is we mistakenly thought that the release of iOS 14.5 was the prompt for using SKAN. Now, it’s been moved to when ATT is going to be enforced, when Apple figures out a way to stop probabilistic matching from happening. There may be a technological way to make it not possible to do probabilistic matching and maybe that’s their approach. So, the new date when everyone’s going to have to truly figure this out is when they enforce it either through technological means or through some sort of policing.

John Koetsier: Yes. And we should be clear that there’s multiple forms of probabilistic, right? Fingerprinting is one where you’re tracking advice based on what it looks like, but there’s also incrementality and other things like that that are probabilistic, but are not tracking- related. Eran, maybe ping in on this fingerprinting piece right here. Do you think that Apple’s going to make it harder and harder? Maybe cloak devices and make them look the same? Maybe obfuscate data that they throw off or limit the amount of data that they throw off in iOS 15?

Eran Friedman: Yes, for sure. I completely agree with Pedro here. You see how Tim Cook speaks in the videos about how privacy is important to them. It’s really hard to imagine that this would be kind of the future. I mean, already in iOS 15, we see the signs. Private relay, basically, blocks all their IP addresses, essentially, kind of hiding them from the server side. So, it would appear as if they are the same. They’re also blocking IP addresses in Safari by default for all HTTP requests. So, it seems like they’re trying to go through the tech route and kind of make more blockers to make it extremely difficult, just inaccurate, to fingerprint users. And I expect that this is probably going to continue, maybe, throughout even iOS15 and the next version.

John Koetsier: Which means that those that are learning SKAdNetwork right now, using SKAN right now, and getting good at SKAN right now, will have a better position. Want to turn to Shamanth here and talk about some of those probabilistic but non- tracking methodologies: media mix modeling, incrementality testing? Should marketers layer different attribution methods?

Shamanth Rao: Yes and no. So, I think, what methods they use really depends on the scale they are at and what’s right for them. For the smaller folks we’ve worked with, definitely, even just doing pre/ post tests or just looking at geo-wise blended numbers, that’s always a good start. I also try to caution folks that processes like media mix modeling or incrementality testing are not for folks that are not at a significant scale just because you need a threshold amount of data for a lot of these techniques to be usable; like you want at least four to five channels, low to mid six figures in spend for incrementality testing, media mix modeling. You just need to be seven figures in spend, you need your internal data infrastructure, you need reporting. It’s a huge rabbit hole, so people need to be ready for it and it’s certainly a conversation I’ve had with a number of folks. But definitely, I think, it’s worth doing basic pre/ post tests, basic blended analysis if folks are at a smaller scale.

John Koetsier: I think that’s intelligent. I think that’s smart. It makes sense to look at a lot of different pieces of data. You are getting some deterministic data with SKAN. I’ve seen marketers using what they’ve learned on Android and applying to iOS because guess what? These are big pools of people should be fairly relevant in terms of certain things, maybe creative, as well. Not perfect, also. Want to move on to conversion models and we’ll kick off with the Eran there. One of the challenges is defining the right conversion model for your app. What’s the best way to approach that, Eran?

Eran Friedman: Yeah, so, obviously it’s all about getting kind of the signals you need to optimize. I recommend the advertisers only start with SKAN, start simple, start with the most basic model that you can test like just initial KPIs (key performance indicators) you had from the first 24 hours, just to see how the data looks like. I saw all kinds of advertisers who started with really complex and theoretical models that today, that just weren’t there. It was completely messy. So you have to get the baseline right and be confident that the framework is working for you even with all the privacy and all of that. From there, you can iterate and make it more advanced. I am familiar with some of the more sophisticated markets out there; already got to the stage where they are encoding their predicted LTV (Lifetime values) through SKAN, for example. You can get there, but it usually requires a lot more work; make sure that you really trust the data first.

John Koetsier: Absolutely, Shamanth, anything to add there?

Shamanth Rao: Well, there are a couple of folks I know that have tried doing predictive LTV. They have not made it work. So, I echo what Eran said to start simple.

Something I would also recommend is that people try and maximize signals, just because this is one of the ways you can address the privacy threshold challenges. I’ve seen a number of SKAN reports where your bits 4, 5, 6, basically, there’s just no value because those just never get fired. You have a super upstream purchase event in there. And what’s the point if it’s, it’s never getting fired? So, you want to, really, set up the conversion value schema so as to maximize signal density.

That’s, really, my key recommendation.

John Koetsier: Absolutely. Shamanth, I’m going to stick with you. And we’ll talk a little bit about tactics and strategies. There’s been a lot of buzz around the increased importance of first party data. There’s probably been a lot of acquisitions around that as well. There’s been a lot of buzz around owned media channels. What’s going on there? 

Shamanth Rao: Yeah, I think there’s definitely an increased focus on owned media, cross promotion, a number of studios I know, they have ramped up their efforts with CRM cross promotional stuff. Of course, as John, you pointed out, there’s an increasing amount of M&A activity out there. So, that’s definitely important. I would also caution that it’s not a silver bullet because I have run cross promo campaigns in the past. It’s hard to make those work. I’ve done a cross promo from a farming game to another farming game and it still isn’t the best just because you’re not always getting the best users, so it’s hard to make work, but that has been an area of focus that I see in a number of apps that I’ve had some familiarity with.

John Koetsier: It’s really interesting when you say something like that, because we’ve seen some of the big studios buy smaller, hyper- casual studios and publishers, and they’re getting access to more idea feeds and stuff like that. It’s interesting because we’ve run cross-promotion campaigns pretty simply in the past. ‘Hey, you’re playing this game. I have this other game. Why don’t you try it? What if we tried.’ You got a little sophisticated and said, ‘You know what? You’re a great player in this game. We’re going to set you up at level 10 in this new game if you try that game.’ What would happen then? Could be interesting. Eran, any thoughts on this whole buzz around the first party, data and owned media?

Eran Friedman: Yeah, I think, we’re probably seeing these acquisitions because the main trigger is Apple is okay with tracking users even without consent across your own apps and websites. So, you have IDFVs between the apps and the same developer. And even if you have like a log in their email, basically, a first party identifier, you have a user coming from your website or to your app, it’s okay to track it. You don’t even need to show ATT to do that, essentially. That’s why it’s a huge advantage for the big enough companies who can afford it and I agree with you, John. I think more companies are trying to think about how they can leverage their first party identifiers to scale their user base.

John Koetsier: Excellent. Ben, we’re going to ping in back with you. And as we look at tactics and strategies, we see a lot of web to app flows. We see that, hey, I can do things on the web I can’t do in an app. I can learn things on the web. I can connect identities. Maybe, I can do other things. I have limitations on that and there are regulations on that as well, but there’s some other things I can do. What do you see in web-to-app?

Ben Holmes: Yeah, that’s a great question. You know, in our space, we saw a big surge of it, and then, when P- matching started, all of a sudden, it dropped off because the web app was basically a loophole. I’m going to take you to my mobile website that I’ve set up in a day, and then, I’m going to take it to the app store and because it’s owned and operated, now that I can track it across the board, that’s a glaring loophole. That’s clearly not what they were trying to do. And then P- matching picked up and it completely fell off. So, is it happening? Yes, it is and we have a few partners that are doing it. There’s a development site, we’ve got to make them a website, if this is going to be brand new for you, especially if you’re an in-app company only. If you have a whole website and you have an app as an extension play, then, it’s a completely different scenario. But, you know, the majority of gaming clients, it’s not like that. They’re setting up these overnight websites. It’s a loophole, sure, it works, you know. I kind of put in the bucket of: is it long-term? Is this something that Apple will just clamp down later? What’s happening with cookies out there in general, is that, eventually, not going to work? If you want to do it now, sure, it’s totally viable. But you know, once again, you got to figure out SKAN, that’s just what has to happen.

John Koetsier: I’m kind of waiting for SKAdNetwork to be natively supported in mobile Safari and maybe a couple of other places like that. And probably, as a plugin or something like that in Chrome. Shamanth, what are you seeing in terms of web to app?

Shamanth Rao: Yeah, we are seeing a lot of adoption of web to app flows, seeing a lot of success. I think there’s definitely a skew towards lifestyle subscription apps, because again, those are users that are more accustomed to web- based flows. If you have a meditation app or health and fitness app, those users are more accustomed to going to a web- based flow. In fact, a couple of apps we work with in the lifestyle subscription space are growing month-on-month on iOS, just based on this. I think this is just because these are the kinds of apps that can lean on longer copy to really sell the product in a way that the app store cannot. What we are also seeing is these also tend to be somewhat mass- market products. The relatively niche ones have struggled with web flows just because, again, you’re still not addressing your targeting challenges as well. But definitely, this is something we’re seeing a lot of success with for a number of advertisers, certainly, something I expect to see continue.

John Koetsier: I love what you said there because you have full control on the web. You can do what you want. And if you direct ads on the web to your website, then you can present your solution, what you’re doing in your app. And guess what? It’s not in the context of competitor apps. If somebody searches, maybe also a competitive search on the app store, where there might be some ads from your competitor swooping in to steal what you’ve got. Now, if you’re doing that, however, you need to measure them, right? Eran, how do you measure them?

Eran Friedman: The idea is if you want to promote your mobile app, you can buy users on the web, which has UTM parameters. It doesn’t have IDFA tracking, buying ads in Google or something to your websites, and then directing users from your website to your app. It can be with a direct link, then it would probably require probabilistic matching like Ben mentioned. Or you could ask for a registration, like enter your email or phone number and you’ll get them into the mobile app later on. And then you can use the first party identifier for the tracking. So, basically that’s the answer to your question, John, there are multiple ways here, but I think kind of the classic extended way is to get users to your service on web, register the users, then you have the first party identifier, then suggest either in email or on the website itself to download the app. And you can basically cross reference these cross-platform and know that this is the same user, which is, again, in a sense, allowing the app to track this user across your own websites and apps.

John Koetsier: Wonderful. I’m going to turn to Pedro now and ask what we’re seeing in the mobile  marketing ecosystem, changing how you view brand and performance marketing. Is brand and performance evolving as we have these changes? Pedro, you are on.

 Pedro Ponce de León: At Liftoff, we’re definitely seeing more traction from our enterprise brand initiatives. I do think there’s more interest in spending, brand budgets from app advertisers. So, I definitely see something there due to Apple’s changes.

Pedro Ponce de León: Ben, anything to add there?

Ben Holmes: Yeah. We brands are soaking it up, man. Let me tell you why. When I say brands, I don’t mean a brand new user acquisition, but brand awareness, that upper funnel. I need to go back to March 2020, April, 2020, brand new for the pandemic, brand spend just went away and UA campaigns soaked up all the inventory left behind and we kind of saw the opposite happen, now. On the iOS side, UA spend, you know, Pedro shared those slides. It went down from user acquisition tactics. Brand awareness started picking it up. At AdColony, we definitely saw this happen because the in-app space has always been kind of difficult for those higher upper-funnel metrics because of the massive data lakes we’re trying to do, just matching with other big data lakes to get those device IDs in there. And we all know the match rates were not that great. If you had a really high match rate, it was more like, I just don’t trust that data; it’s not first party data. Where do I even get that? So contextual line was already a big part, right? Hey, if you’re a CPG and you’re trying to get that 30 to 44 year old female, they’re playing match three titles. We know that from Nielsen data, all that kind of good stuff. And we’ve seen brands just lean even more heavily into it because now it’s even more efficient because the CPMs have gone down. The metrics are all really high whether they want for those brand awareness campaigns. And we’re seeing them just buy it even more now. So, it’s been very interesting to see that shift happen.

John Koetsier: Eran, I want to throw one last question your way. We’ve talked in and around this issue for this whole webinar we’re going from this very simple data model that we used to have, where there was one number, one identifier that was the one ring to connect them all. And it’s not happening anymore. We could connect installs, everything with that. Now, we’ve got SKAN, we still have IDFA, we’ve got Android ad ID, Google ad ID on Android, maybe, we’ve got incrementality if we’re over six figures, maybe we have media mix modeling if we’re over seven figures in spend, as Shamanth was talking about, and we have all these different data sets, as well as your campaign data, that’s going into and what the platforms are telling you that you’re doing. Your attribution data set, your performance data set is more fragmented than ever. How do you consolidate all this stuff to get a single truth out of this and understand what’s performing?

Eran Friedman: Measuring the performance just became much more complex than before; we had the easy days before. You had IDFA, just worked very straightforwardly. Now, it’s much more difficult. You’re going to need to live with multiple datasets that are fragmented, exactly like you said. You have to prepare your reporting and your infrastructure to be able to ingest all that and look at all these together. We’re hoping for a SKAN to become kind of the source of truth. It’s probably not there yet, it’s progressing there but it’s not going to be the only thing. You’re still going to have some sample side for IDFA, which can be very valuable for sampling data and models. You’re going to have the organic and IDFV, like Shamanth mentioned, the blended metrics that you can look at to see details, you can add incrementality to it, and it can provide you valuable insights. The performance marketer is going to have to have multiple tools that they need to kind of look into and understand what are the trends from a strategic standpoint, while the more granular details that they can understand in their campaigns and creatives. They need to work with a large set of tools and understand how it all fits together in the same picture. Interesting times, I think.

John Koetsier: We do have to get to some of these questions here. And I think I’m going to throw the first question to Shamanth. And this question is from Andrew Nelligan and he’s saying, ‘How do you expect the iOS changes will impact consolidation trends, particularly in mobile gaming?’

Shamanth Rao: You said consolidation trends, I’m assuming he’s talking about channel consolidation and I think that’s already under way for folks that are relatively small. We just don’t go outside of SANs. In the past, pre-ATT, we would just say, ‘Hey, let’s test some, this SDK networks.’ Especially in gaming, I would say we would test some of the SDK networks or test DSPs; I think at the smaller budgets. That’s just harder to do because of the elevations we talked about.

John Koetsier: Interesting but the data that I’ve seen suggests that there’s actually a different trend happening is that people are going away from SANs and going to more channels. We’re seeing that actually happening. Pedro, how about you? Consolidation in mobile gaming? Is it going to be more and more common or we can get uber studios with massive multi-billions of players?

Pedro Ponce de León: Yes, that’s a possible outcome, definitely consolidation on the supply side with more apps; as we mentioned before, IDFV is important here. So yeah, I see that trend continuing.

Eran Friedman: Just to add to that, there’s another advantage in scaling. That’s the budget. We are already talking about how you can barely get any data with tiny budgets and when you consolidate those studios, you can spend more and you can get more data. Now, like everything goes kind of aggregated, so it becomes important. So, yeah, I agree that you are probably going to continue seeing this trend.

John Koetsier: Interesting. And you can also learn if some of your titles are good ways to get people into your house of brands, house of apps, and then you can actually move them out to others as well. It’s an interesting world that we’re living in and moving into. Ben, we’ll stick with you. And this is a question from Chris O Croix, and he’s talking about lag and data being passed back to platforms. How’s that impacting those platforms? What data do you get? And you’re getting into 24 hours, 48 hours, seven days? How’s that impacting your ability to optimize a campaign?

Ben Holmes: It’s impacting it. It’s true. He also brought up algorithms, right? Is it hurting algorithms and the answer’s yes, it is. And we’ve talked to DSP partners as well. And it’s very similar trends. You know, this goes back to the conversion value mapping, right? Because the timer resets because of privacy thresholds. I was a really big fan of what Eran was saying about things like, ‘Hey, start simple and then build upon it.’ We actually have the exact same message. But it’s from optimizations and targeting. You gotta start big, start at the country level, everything you thought you knew, it’s gone. Whatever it might be, you have to start over and the more willing you are to do that, almost like a completely new learning phase, the better, and you need exploration models to go longer now. You know, if it was like, ‘Hey, we thought we knew that the first three days; now, it’s going to be the first 7 to 14 days. And it’s because of that time delay and it’s affecting all the algorithms. There’s one marketer I was talking to and

they went from 71 post-install events down to 2. It’s just a massive drop in how they do that: post-install conversion and the funnel they were trying to get down, but it’s the only way they could start to rebuild that dataset.

John Koetsier: Wow. 71 to 2, a wealth of data to a very small stream. I’m going to throw the next question to Pedro. And this one is from an anonymous attendee: a major mobile game publisher cut their bookings forecast by a hundred million, due in part, to ad pricing, higher CPIs, lower ROIs. If you’ve been following the news recently about public companies in mobile gaming, you may have seen this already. Question to you, Pedro, is: are you seeing this broadly or is this likely more specific to this one particular game company?

Pedro Ponce de León: So the higher CPIs, it kind of ties to the other questions. Under SKAN, there needs to be an adjustment to your metrics because it’s fundamentally a different way to attribute. So, you’re looking at measurement and metrics through a different lens now, and not through the way that MMPs were doing it. So, it’s higher CPIs, the metrics all look worse, but it’s also because the measurement is literally different, right? The way the measurement is done, the way the attribution is done it’s different. So, yes, the algorithms are impacted. We’re all going to get better at improving the algorithms, maybe use more upper funnel signals. Hopefully, people will spend more, maybe the thresholds will be adjusted, so we’ll be able to do more with the algorithms, but ultimately the way attribution works is very different. So you have to adjust that. So like, when, when these questions get asked, it’s hard to answer. What lens are you looking at?

Pedro Ponce de León: Are you looking under MMP attribution today or under SKAN attribution? If you’re looking at SKAN attribution, you’d have to adjust your expectations. I think that’s part of it, but I think we’ve already said here that a lot of advertisers are seeing bad metrics, compared to what they used to see. So is that happening? Is that likely? Yes. Is it gonna stay the same? I think so. Because ATT and how SKAN works is fundamentally different.

Ben Holmes: I think there’s a bit more like the ad monetization, right. You know, if it’s a hundred million cut in bookings and that tells me that their ad revenue that’s taken a hit. Not knowing anything about them, that kind of sounds like that might be more of a hyper – casual where ad-mon and, you know, retention were their main source of revenue. They’re not doing any in-app purchases or anything like that. And that’s definitely happening. Especially, the gaming pub that was overly relying on iOS and didn’t have a nice, healthy Android mix as well. They’re seeing it, lower CPIs and the inability to efficiently target. That’s how this works, right? Because if we are on the vendor DSP network side, if we’re transacting in a CPI and we’re not getting those installs, then we cannot turn around and give out those really high CPMs. And then the publisher, they don’t like it. So, that’s why you’re seeing the switch also to CPM pricing, to try to get that revenue back over to the publisher. So it’s definitely a hit. Now, is everyone going to be hitting, taking a hundred million dollar loss? How many companies even have a hundred million dollars to start with?

John Koetsier: Ben, that was a great segue actually, because you know, it is an iOS webinar, but hey, we got to hit Android. It’s there, it exists. And it’s a massive part of the world. It’s a bigger part of the world. Here’s a question from Ori Ger and I’m going to shoot it over to Shamanth. With the Google announcement, the hiding GA IDs, when users are opting into limit ad tracking or opting out of ads, personalization as is on Android, what do you think the future of Android attribution will look like? Are they going to follow the privacy changes in your opinion of iOS?

Shamanth Rao: Well, I would point out this is the equivalent of LAT, which was on iOS many years ago. So, one way of looking at this is that ATT is going to be at least two to four years away. So, that’s certainly one way of looking at it, but definitely, knowing Google, they have been followers of everything apple have done privacy- wise. So, definitely something to look towards in the future. I don’t know if we are necessarily worrying about it just now, because I feel like dealing with SKAN will give enough of a playbook for us to work with Google. Google also has a lot of privacy initiatives. Those should give enough of a leading indicator of how to prepare for this. So yes, this is going to happen. I don’t think anyone’s losing sleep over just now.

John Koetsier: This is Google, it’s an ad network. So they’re going to be a little different from Apple, which is a hardware company with software and services. So, it is a different world, but I think all of us are aware that we’re going to see more privacy initiatives from Google as well. And we see some of the initial announcements there and there’s just going to be more because that’s essentially the trend of the world.It is the one hour mark and you have all been amazing. I have to thank every panelist here. It has been great. We also have a ton of questions we didn’t get to. We will answer them on the Singular blog. But I have to thank the panelists here. I mean, Shamanth, great insight, Ben, really appreciate what you brought, Pedro, thank you for the data and the insight, and Eran, as always, much appreciate the wisdom and perspective. It has been wonderful for all attendees. Thank you so much. And this is John Koetsier signing off.

A REQUEST BEFORE YOU GO

I have a very important favor to ask, which as those of you who know me know I don’t do often. If you get any pleasure or inspiration from this episode, could you PLEASE leave a review on your favorite podcasting platform – be it iTunes, Overcast, Spotify or wherever you get your podcast fix. This podcast is very much a labor of love – and each episode takes many many hours to put together. When you write a review, it will not only be a great deal of encouragement to us, but it will also support getting the word out about the Mobile User Acquisition Show.

Constructive criticism and suggestions for improvement are welcome, whether on podcasting platforms – or by email to shamanth at rocketshiphq.com. We read all reviews & I want to make this podcast better.

Thank you – and I look forward to seeing you with the next episode!