As you might already know, Apple is introducing the App Tracking Transparency (ATT) framework and starting from iOS 14, mobile publishers, namely advertisers won’t be able to target users unless they explicitly consent. iOS 14 is already there, but IDFA deprecation was delayed and should be enforced in early 2021.
We’ve held an AppGrowth Panel #1 where the leading app industry experts – Thomas Petit, Andrea Raggi, Johannes von Cramon and Ryan Goeden – shared their thoughts and valuable advice on the topic. Read this post to learn what some of the most talented, competent and knowledgeable guys in the app industry think of the IDFA deprecation and what strategies they recommend in order to adjust to a new app marketing reality.
This is a transcript of the app growth panel video. You may watch it or read the text. Use the table of contents to navigate this longread.
Johannes von Cramon, expert on mobile marketing, Apple Search Ads, ASO, mobile analytics and retention, App Marketer and Co-founder of Growfirst.
Thomas Petit, the author of Mobile.Ad.Ventures, a world-renowned Mobile Marketing Expert and independent App Growth Consultant.
Andrea Raggi, expert on mobile performance marketing and user acquisition, the author of the ASA Stack, Performance Marketing Lead at Phiture.
Ryan Goeden, expert on mobile attribution and ad tech, VP Sales at Kochava.
I think basically Apple will turn around the logic of the former limit ad tracking (LAT), which previously was a tracking opt-out for third-party apps, which reseted the IDFA for advertisers. And this will become opt-in, and in some kind of it already is because Apple already changed that you have to give this consent, but it’s not enforced inside iOS 14 apps right now, but you already have the ability to ask users if you want to. But currently the opt-in is given for most users.
I think that it’s a good way to explain where the problem originated. What I noticed is that this is an array of consequences and the main difficulty is splitting the big topic into the subtopics, because it’s sort of a number of consequences, affecting such issues, as: how to request the opt-on, what will happen to remarketing, how do I manage a self-attributed network? How do I manage open programmatic and others? I’m leaving it on the table now because we’ll dig deeper further.
But I think for clarity, the best when you talk about that broad topic is to (and what we will be doing today) isolate one thing after another to tackle it, which helps get a little bit of clarity on it, because a lot of the confusion is, “Oh, I’m lost because the consequences are so diverse that we’re talking about everything and we end up talking about nothing.” Johannes has summarized really well why it’s happening and now we’ll dig deeper into the subparts of this big topic. But I noticed the confusion came from that for so many people.
In addition to this ATT framework, Apple also published a big update and is now emphasizing the importance of their SKAdNetwork framework. So SKAdNetwork has actually been around for a couple of years now, I think since 2018. But it sort of flew under the radar and nobody really used it. And it was largely limited to just doing attribution of installs, fully contained within the iOS environment. But when they announced the ATT framework, they also did an update to SKAd, which is allowing for a what they call a conversion value method to be included in this attribution.
And what that meant then is that any advertiser that wants to use SKAdNetwork can actually now start tracking some type of engagement or post install event and attributing that back to the marketing campaigns as well.
I definitely think this is a huge change in the industry and many, I think, are still trying to get familiar. So I’m really excited to share what I know and discuss these hot topics with the guys. And just to talk about how marketers are still trying to get familiar with it.
I’ve seen a chart the other day where almost 50 percent of the respondents were still not at all familiar with what these new frameworks are.
And I think it is definitely the time to start getting familiar with them, and start testing them and talking about the App Tracking Transparency framework, which is, I think, one of the main subjects of iOS 14. Apple will require all different developers to come to consent for trying to track users and access their IDFAs – basically looking into their devices identifiers.
I think there’s quite a long list of things that need to be done. I’m not going to cover each and every little thing, but I’ll just cover at least some high-level things that I think app marketers should be thinking about and getting ready for.
So the first one actually following on your point, Andrea, is kind of recognizing the fact that this is still a big unknown and a lot of people don’t really know that much about it. And so first and foremost, it seems obvious. But just learn this, learn this stuff, spend some time diving into what the ATT means. What is SKAdNetwork all about? How does it work? And start thinking about linking your long-term marketing performance metrics back to this framework in the way that it works. So my first main point is: learn this stuff and then talk to your partners about it. Because everyone’s working out plans, it’s kind of a moving target. And the more conversations you’re having and the better informed you are, the better ability you’re going to have to ascertain what to do. And as we said, like a lot of people don’t know, so it’s just unfortunately on all of us – to learn what we can about it.
The second thing is maybe not everybody’s favorite topic, but I would start getting your legal team involved in these discussions. Because, you know, as we think about the ATT framework, this also sets against the backdrop of a bigger privacy push that we’re seeing across the world. And so things like privacy legislation come into effect as well as, you know, complying with the terms and conditions of Apple. And a lot of these things are moving targets. It’s constantly changing. And so just I would recommend starting to incorporate that kind of aspect to it, getting input from your legal team in these conversations earlier, because as you think about prompting for consent to share data, it’s also a good time to be thinking about just consent management overall from the legislation perspective too.
My third point is, kind of stating the obvious here, but I would recommend that people actually try to pilot some SKAdNetwork campaigns. The technology is already out there to utilize that. And so there’s no reason to wait until the ATT framework is in full force to then start doing it. You have a lot of tools at your disposal now to actually test it out and then do some pilot campaigns.
And then finally, thinking a bit more about the big picture here, this ability to track users of an individual level and share that data with other partners is largely going to go away. And so I think that that increases the importance for both advertisers and publishers to think about their first-party data and what can they learn about their audiences, both from selling advertising and buying advertising perspective, and trying to learn as much as you can so that you can ultimately try to find the right channels for you. So think about your use of first-party data to understand your users.
Ryan did a great job, yes – read, talk to your partners. Definitely talk to your legal team. The one I wanted to mention here is everybody’s jumping first on okay I’m going to need to require the consent. You actually don’t need to require the consent. So for me, the first question is: I am going to operate with the ATT, which means I’m going to ask people to consent or not, and then I’m going to have a universe of people who consent and a universe of people who don’t consent. And you have to deal with these two data sets that will be hard to reconcile. And you can do things with one, you can do things with the other that just don’t mix very well. But there is also a situation, and I believe most developers, they will end up requesting it and having this duality of datasets. But there are cases where you don’t necessarily need to request the consent. And if you think about your onboarding and you’re like: I’m already asking a lot of permissions – maybe location, maybe push notification, maybe opting to email, maybe the GDPR – that is lowering the cycle of the onboarding.
And adding another one is, of course, friction. You might want to do it because it’s really beneficial for you. But there are a few cases where you actually won’t necessarily have to request that consent.
And another example is kids apps. I think they’re not going to even be able to request it at all. So they have to live with only the non-consent population. But I’m also thinking, some developers who are not heavy on UA, don’t monetize through ads and actually don’t really use the IDFA internally at the moment – they don’t necessarily have to request it. And even beyond Facebook, because obviously they’ve got a massive trove of first-party data. They initially said that they’re not going to request it for the galaxy of Facebook Ads. I have the feeling that at some point they might go back on this decision. But I mean, it’s just to represent it not only because you’re an indie not doing UA and selling ads that you don’t have to request it. There are cases and I think this decision of do I request or not leads you to how you’re going to build your data structure, your acquisition, your monetization. And it’s a question that doesn’t have a necessary answer of yes, I need to consent. So at the very beginning, which comes after obviously learning and talking to your legal team, so it’s not the very first one. And I’ll jump straight up to one point, which is once you decide to request it, maybe you’re testing how you request it and you make those tests Ryan mentioned in regards to data here is how am I going to manage this data compared to today, is it linked to BI and so on?
I think one that is really big and it is important to prepare right now instead of waiting for January is which conversion I’m going to use. So Ryan mentioned in the first iteration, we only have installs, which for most marketers, is just not enough to measure performance. So Apple has just updated it with the conversion value here, just like two weeks ago, crazy late but anyway, there are limitations, which we’re going to go a little bit into, into that conversion value. What can I pass? Because it’s not going to be the same as the postback that we’re able to pass today. And here, if we look at the last maybe two or three years, a user acquisition has become a lot more important to say, OK, what events do I have that are actually indicative of good retention, good monetization in the long term? And a lot of people have started to work on it. I think right now is a perfect moment to go back at this analysis and do more regression analysis on what are really good early drivers – what I call the magic even sometimes – what are really good early drivers that are indicative of very good long-term retention and monetization? Because the amount of evidence we’re going to be able to pass is less than before. The timing in which we can pass them is going to be less than before. So I think here is a really good time to go back and see what do I actually have, like start thinking about what you will be able to pass and build this conversion value in this new way under the SKAdNetwork, which is very different from the event we were managing under the MMP. And even if you don’t implement them today, I think preparing by running this data analysis of what is a good indicator is something that it’s better to do right now in December than to wait for January. And because then you’re late.
There are other things, but there is also something that you can already do right now to save yourself time for January.
I think that those are great points and also to touch upon what Thomas has just said, I think you should really consider what to do with ATT. And it doesn’t necessarily mean that you have to request for the IDFA consent. Maybe your brand guidelines go against that, maybe you are very privacy-focused.
So I think you should be thinking about your company and what you should be doing with that. And then you could eventually test several UAC flows, see what makes sense for the users, maybe testing different ideas. You could perhaps try to prompt the ATT right after at the beginning or maybe after a happy moment. But make sure that you provide a reason why you’re doing that. Why do you need to use this data? And you can then analyze what’s working best. And then, you can optimize the onboarding and user experience according to that. And in the end, you optimize the user experience, but also minimize churn, which in the end is a win-win situation.
True. I’m really with Thomas and Andrea on this point because in all articles and panels and webinars about this topic, you always talk about gaming and e-commerce, all those very common verticals and big spenders in user acquisition. But I think this misses a lot of publishers who are in between that or even outside of these categories. And it really depends if the ATT makes sense or you can do this, actually. And also the transformation of the current event-tracking setup into conversion values in which you are in and what the business model looks like you’re aiming at.
So I think that’s it should really be the first question: is this at all affecting me? And I personally would start one step further and ask myself, how important is iOS traffic to my app in the first place? Because maybe it can be a good idea to shift some budgets to maybe Android, maybe web traffic, at least in the first couple of months, to do a little bit more environment-focused testing, which I believe will be a little bit different if all those changes will be enforced. Then everything we’re seeing right now, because 2020 is a crazy year with COVID-19 right now, we’re having this end of the year rally where everyone is spending a little bit more and a little bit different than the other months. So maybe it’s at least one thing to say, OK, maybe I can shift some budgets from iOS two to other platforms in order to test this life and then ask myself, can I do a pre-screen for ATT? Is this viable for my business goals?
To some extent, I think this question kind of harkens back to a lot of questions that I would get back in 2012, 2013, 2014, when advertisers would come to me. So at the time I was working at a mobile ad network, but I was looking at attribution. And one of the questions that we’d always hear advertisers say is, hey, this is attribution, it’s done just using the IDFA, it’s not so complicated. We’re not going to use your service – we’re just going to build our own attribution technology and bring it in-house.
And certainly there are some high profile advertisers who have done that with kind of varying degrees of success. But by and large, I think that question was answered with the two things. One is that don’t underestimate how much work it actually takes to build your own attribution tool.
And then second, what’s kind of the importance of MMPs is, as a neutral arbiter when it comes to establishing a single source of truth for the effectiveness of your campaigns.
And this is what played out in the early 2010s. And I see these points as being somewhat analogous and a bit useful now, when we think about what it’s going to be, what’s going to be the role of an MMP if this is the way the industry is going.
So the idea is that we are a bit of a neutral arbiter or a neutral party to establish what conversions are happening, and SKAdNetwork is going to work differently than the dynamics back in 2012.
Having an MMP that’s able to ingest your conversion data and report on it does create a nice sort of third-party aspect to establish what is true.
Because one of the things we see and this has been the case for a long time, is there are always going to be elements of data discrepancy between advertisers, publishers, ad platforms, et cetera.
But I think the more salient point now is more around the idea of not underestimating just how much work it will take to implement the SKAdNetwork framework, both in terms of development and infrastructure, but also thinking about the conversion value, an issue with SKAd.
So like I said, as you’re thinking about this, obviously, SKAdNetwork gives the ability for an advertiser to do conversion-based tracking on their ad campaigns an iOS, without using an MMP. That is there. But it’s a lot of work to implement and manage this. And an MMP, any MMP, not just us [Kochava], you know, we’ve already done a lot of the heavy lifting on this. We have the infrastructure to be able to ingest ad signals. We have SDKs that are ramping these conversion value methods and making it a lot easier for an app developer to implement SKAdNetwork than if they were to go on it and go it alone.
So, harkening back to the early 2010s, don’t underestimate how much work it actually takes. MMP really can be a way for you to take some of the heavy lifting off yourself.
There’s a third point on this, and that is really related to the conversion value method that everybody here has touched upon, and it’s basically to dive into the details a little bit with this conversion value method. As we said, this is what allows you to track in-app events and attribute those back to your marketing campaigns. But the way it works now is basically you as the advertiser, get to pass one numeric value in this conversion value postback. It’s a number between zero and sixty three. So you’ve got these different options to pass. And you could go and map that to individual in-app events that you’re tracking and just kind of have a one-to-one connection there, and do it all yourself and develop it. But if you do that then and you make any changes to it, you’re going to have to go and update that and re-release your app. And so that’s an area where, again, an MMP can really be useful because what we’re finding is certainly us and the other MMPs as well, they’re building sort of more clever frameworks on top of that conversion value method. That can then be utilized and changed in the server side so that it doesn’t require updating the app every time.
So, again, kind of thinking about what’s the role of an MMP here – I think there’s quite a lot of value that can be added by providing really useful frameworks to utilize this conversion value method, then doesn’t put all the onus on an app developer to have to figure it out all themselves, implement it all themselves, and then have to manage the ongoing updates of that.
I think there are going to be winners and losers.
We can start from the losers, and the losers are definitely going to be Facebook and their audience network lookalikes. And we can already see how their inventories are declining. And that’s due to the fact that the IDFA zeros are not being targeted in the ad networks. And also, I’ve seen one question and that was asking how many zero IDFAs do we see now? As of October, that was 34 percent. And that’s also going up. So we can expect that to be increasing over time. Also I’ve seen some other stats the other day, around 50 percent of iOS users are on iOS 14 in October. In November, those are already 70 percent. So all these developments are definitely impacting the ad network’s ecosystem. And Facebook shrinking inventories, CPMs go up, and more competition. And that’s not a very ideal situation for them.
Also, another thing that we have to remember is that Facebook has, as I said in a recent study, that approximately 50 percent of their revenue could drop without the personalized ads. So the deprecation of the IDFAs and the lack of deterministic data is certainly going to impact them.
On the other side, though, we can say that there are some channels that could get winners out of this, and I’m thinking of Apple Search Ads. Maybe even there’s going to be more budget shifted towards ASO, and I think those will become more important next year. But also in the short term, maybe even now, start testing what makes sense for you and start thinking about, I would say, Blue Ocean Strategy. Don’t stick to the Red Ocean strategy, which means, focusing on the same old solutions, on the duopoly, Facebook and Google. But think of different ways where there’s less competition, there could be more opportunities for you to take advantage.
I have a question that has just come to my mind on the topic of ad networks or mostly outside Facebook, Google and Apple, since fraud is a very big problem there, especially with the smaller ones. Do you guys think this will improve the situation on iOS? Because Apple itself will verify if a click or an impression will lead to an in-store or a conversion. This should be a good thing, right? So my idea is that maybe it’s a good thing for this type of fraud, but maybe also easier for fraudsters to mix in invalid traffic because you can track on the user-level basis. Maybe that’s a good question for Ryan. I don’t know. I’m not so familiar with the topic because we mainly focus on the big ones and there’s fraud, not that kind of a topic. And if you take Google display out of the equation, but I don’t know, I’m not sure if this is good or bad in the end.
Yeah, Johannes, I think that’s a great question, definitely something worth thinking about. I’m not going to pretend to be a savant and say that this is exactly how things are going to play out. But there are two areas with fraud that we see as potential ways that fraud could increase.
But I would balance that out with – I think you make a valid point there – that by Apple owning some of the attribution, that maybe certain types of fraud might decrease.
So one area is that with the ATT framework coming into effect, IDFA going away, we think that the importance of more probabilistic forms of attribution might become more important.
And with that comes a lot of potential to increase fraud. So, for example, you hear the idea of click flooding, which are, you know, different types of servers creating a bunch of bogus clicks in order to try increase the likelihood that some installs might be misattributed to those in more sort of probabilistic attribution environments. So that’s one area for which the overall increased importance of probabilistic attribution means that certain types of fraudulent click traffic might come about because they’re trying to win an attribution there in a fraudulent way.
So there is more of click flooding – this is something we know about, we have tools for that. So what I’d really say is that might become more prevalent potentially now a kind of a new emerging type of fraud that we’re anticipating – I’m not saying it’s going to happen, but just something we’re trying to be aware of – is bogus IDFAs being created in order to create some types of traffic that maybe aren’t actually real. Because what we see is that for the small subset of users, whatever that is, that do opt-in app tracking and allow for their IDFA to be shared, that’s going to be a premium type of traffic because you can do all the things that we’re used to doing now.
And that just seems like that’s too enticing of an offer for fraudsters to not try to find ways to create a bunch of sort of bogus IDFAs that give the appearance of users who have opted in.
It might be a little bit early to say this, but in my opinion, retargeting, as we know it right now, and mostly this is the e-commerce context where you see an ad with content you saw in the app like shoes or a dress or whatever you were browsing, and then retargeting ads saying, hey, this is 30 percent off. So this typically retargeting context is, in my opinion, may be disappearing almost entirely on iOS. Because the main reason for this is retargeting is very tied to the UX of an app. So it’s much more important to get the content right in the ads and then the routing, the deep linking to the app. And this is a thing which isn’t very easy to implement these days without all these IDFA changes. And it will probably be more or less impossible without user-level data. So my belief is retargeting will go away over time on iOS in this typical context.
I mean, you could always re-target users just on a higher level and say, OK, we’re targeting users who installed the app and did something with it and say, OK, come back to the app, here are some features that are really good and people might not know about – something like this. Or you have new contents within the app, but not this targeted as it used to be.
I have I thought about something else, but I think I’ll bounce on what Johannes has just said so that we transition smoothly here. And let’s say that retargeting capabilities at user level are entirely gone for remarketing. Let’s mostly agree with that. But let’s say this is the situation. One strategy you need to work on is what alternative do I have for re-engagement? And I think here, at least the way I see it, remarketing can be a very powerful tool for a lot of apps and I’ve actually seen it going way up this year. I don’t know if people are anticipating it’s going on or whatever, but the way I see it, you remarket people when you don’t manage to re-engage them on your first-party property. If the user is still opening the app, you can do a lot of things like in-app messaging and so on. User has the app and doesn’t open it? You still can send push notifications. Even if a user has removed the app, you can still email.
All of those, they are very complementary to remarketing; they’re not better or worse, just complementary. The thing is that if remarketing goes away, then one strategy you should think about is doubling down on all these alternatives, on re-engagement. And where are you on these tactics of life-cycle if you’re doing very little there and a lot of remarketing, it’s time to think, OK, I need to act. Which, I mean, it’s not the same. It’s not going to bring the exact same results. But you will avoid the loss before it happens, and these probably gains for you to have.
So if remarketing has been very important for you, and it is for a lot of gaming and e-commerce due to the repeat nature of of valuable user, I think it’s the time to double down on CRM, on push, on LiveOps and move your focus from re-engaging user outside to re-engaging user on your property, maybe within other apps, maybe on your website – there are lot of options.
So that’s the first point. The other one for me, which is kind of big, it’s not really a strategy, is also part of the preparation – it’s strategic thinking, which is the way you will measure your UA efforts on which channel it is, where it’s a losing channel or winning channel, like Andrea said.
I think the perspective is going to be very different because the nature of the data is going to be very different. And I think here people are not very prepared on what’s ahead of us in terms of how the data will look like. So you can preview what SKAdNetwork data will look like if you run this test, that Ryan mentioned, and most MMPs already provide information: hey look, this is how it’s going to look like. So you can go to whichever MMP you have and they already have: this is what it’s going to look like.
So you start interpreting what am I going to do with this and how I’m going to look at my campaign. And for me, the way I see it, acquisition is a bit of a two approaches: one is how do I optimize my campaign? I’m the UA manager. I want to get the best of it. I want to get profitability out of my campaign. This part is going to be rendered really hard because we don’t have impressions anymore. The data is aggregated and the conversions are earlier than before. The optimization is going to be a lot tougher than it used to be. But on a very strategic level, let’s say you’re the C level, you’re the CMO, you’re the one making the decision. And I’m putting 30 here to Facebook and 20 here and 10 here, to Apple Search Ads or whatever. I think here it’s a really good time to get prepared: the reality I’ve lived in about the profitability of each is going away. I need to understand better, which is working and not. And I think here it can only be done by the combination of: what I’m looking at the campaign so mostly through SKAdNetwork, partly through possible extrapolation from the IDFA cohort. And a topic that has become really big before this enforcement has come – is starting to measure much more incrementally, which is basically: are these really extra users or they are the users I would have had regardless which has grown. And if I look back over the years forgetting about the iOS 14 for a second, you would go back six, seven years ago and people were: everything is on CPI, I’m getting an install – I mean, it was so cheap, it didn’t matter because it worked.
And then marketers got smarter. They started moving to events and measuring CPA and then we moved down to ROAS. I need profits on this. I need return [on investment]. And for a lot of people ROAS has been sort of the end of the line. I’ve moved from install to even to ROAS, I’m good.
And I think it’s just not the end road here. And sort of after ROAS, people usually start looking at the payback period. It’s great to make a return on it but if it’s going to take four years, there is a price to pay on that, and usually after ROAS people start looking at the payback period. And I think incrementality is the one that came before, and the smart marketers already moved there by nature this year. But I think this push about changing attribution is going to wake up a lot of people. It’s going to create a lot of difficulty on optimization, but it’s going to open the eyes of a lot of people on what is the actual effect of their paid campaigns. And I think it’s bad generating a good thing that we’re going to have to rethink about that and be a lot more strategic about where we actually place our bets and how we make this. And some channels that have been partly cannibalizing a lot of other activities, and they’ll look good when you look at it deterministically, typically search, because it’s at the very end of the road here and it’s been affected by other things. But I don’t mean that search is bad – it’s just you have to look at it through the lens of multichannel. Today, if you’re operating at a certain scale, the same users down TikTok, down Instagram, they look in the store, they’re also playing other games and they see how all of this is working together. And I think here, the fact that our data is going to be less precise is going to force us to think a lot more strategically, which, I admit, tactically is going to be a nightmare. But strategically, it might be a good thing.
I’d just probably add one thing, because of the lack of deterministic attribution, I think that if your business model relies on identifying whales who spend a lot on IAP (in-app purchases), that’s going to be a lot more difficult.
And the knock-on effect of that is almost like a strategic rethink of business models as well, because it’s not going to be so easy to just go and kind of carve out the exact segment of inventory that you want and go after it.
SKAdNetwork, as Ryan said right at the beginning, was launched in 2018. And I don’t think almost anybody has to use that. But it has gotten some interest lately, right? And it is this anonymized attribution solution that enables you to keep tracking installs coming from different networks, but without compromising anonymity. So that’s the keyword. Well we’ve been talking about deterministic and from that deterministic and individual data targeting, we’re going to move into having aggregated data, so no user-level data, no capacity to measure the users, to attach them to a specific campaign or network. We’re also moving from what we were used to, from deterministic to probabilistic. With the SKAdNetwork, we’re going to get limited signals for campaign optimizations. It is really impossible to attach real revenue figures to the advertising efforts. We’re also going to have only 63 conversion events.
So you have to be super strategic about how to choose conversion events in an appropriate way, making sure that possibly you select early funnel events, because we’re also going to move from real time optimization to delayed data.
With the SKAdNetwork, that could be a delay of up to 24 or 48 hours depending on the timer. So I’ll just go back on how this SKAdNetwork works. And basically when a user decides to click on an ad, then the specific ad network generates a signature, then the signature is stored as the user installed the app. In the meantime, of course, you have to register for SKAdNetwork, you have to make sure that it’s available for your ad networks.
And after that random timer is set, the ad networks receive a confirmation by a postback of that install and then the install can be verified. So it pretty much works like normal attribution, but, you know, it is aggregated data, there’s no cohorted data. And yes, it is very important that you also start working on that right now, because, as we said before, those inventories, perhaps on Facebook and self-attributing networks, the inventories are shrinking. So there’s more competition.
Also, one of the limitations of this is that fraud, which was touched upon a little bit before, the event conversion values will be unsigned, which means that the app developers will not be able to verify whether the event took place in the way they’re used to. And this could open the door to fraud. So re-engagement is not factored. Thomas mentioned that there’s also not going to be any impression attribution. The SKAdNetwork does not support deep linking. These are just a couple of limitations now I can recall of but it’s going to be a very different ballgame. And, you know, we need to become more strategic and start thinking out of the box. And if you get there before your competitors, you can certainly take advantage of major return on ad spend and even better performance.
Apple has made an exception for Apple Search Ads in terms of these IDFA changes. So this might be the only way to reach users without consent because they won’t see any app install ads on other platforms, like it’s already the case with Facebook, for example. But I think in the current setup, this won’t change much for publishers because we have these really search-focused ad campaigns on Apple Search Ads. But if they move a little bit more to display-like environments like with the tests and the news and the stocks right now, then this lack of scalability they have right now with Apple SearchAds, which is very profitable for most publishers, but not that scalable for many of them. And this can be a solution to this. If they move into display or maybe other channels and have always 100 percent of users they can target, with the exception that users with LAT on and off are not behaving the same in most scenarios. So they should be treated like different user groups because the ones with LAT on are a little bit more privacy-focused and a little bit more aware of the things which are going on in this tech environment as a whole. But you can also talk to those users within Apple Search Ads, and that’s in terms of what Ryan said, when you’re targeting whales or very specific target groups which monetize better on iOS – this is very valuable, I think. But it’s hard to say right now how this will look like, because we are in the really early days and Beta stages of those display texts, and search will always be limited in the App Store. So there’s no scalability, I think.
Even though I agree with the observation that the limited scalability of pure search and if Apple is going to move away from that. They are two bits, like there’s one thing I want to specify, which is today on Facebook and Google, because they rely heavily on the IDFA not showing the app install ads to the LAT inventory. You can reach them through Apple Search Ads, through ASO, you can also reach them outside of the walled gardens, because open programmatic would still target them, the SDK ad network will target them. I think here are two things to consider.
One is, do they see the ads or not? So on Facebook and Google today, the answer is no – on Apple Search Ads and SDK network it is yes. And how do I track them? So typically today, Apple Search Ads show the ads to the LAT people, but they don’t report them to you at all.
Facebook and Google don’t show the app install ads to these people. On SDK ad network they show the ads and they track it through fingerprinting. So there are always two things to consider: 1) do users see the ads – yes or no? 2) how do I track them?
I think the situation is very much of today, and a lot of people have not been aware that this inventory was missing for a very long time. They get a word this quarter because this percentage is rising a little bit. But I think this is going to be an entirely different situation in the sense that Google and Facebook can not go…I mean, I don’t know how they could escape by showing only the ads to 70 percent of the inventory – they’re showing other ads to these people, like web ads, the e-commerce, direct and so on. But next year, they’re not going to do it for everybody they don’t know if they can consent or not. So at the moment they will start tracking through SKAdNetwork, the reach on this network is going to go back to 100 percent.
That’s due to the fact that the trackability is going to change. If they track it through IDFA, they decide not to show the ads, if they track it through SKAdNetwork, they will bring it back. So the situation is a little bit about to change.
Going back to Apple Search Ads, I think here the reason Johannes was speculative about it and I would completely be in line – is that we don’t really know what Apple intends to do with this display network that they have been testing for eight months and didn’t tell anybody about it. It’s just been spotted in the wild, but they didn’t communicate anything about it. My personal bet is that they’re going to realize that news and stock has the same problem, as search – that it’s very limited. Even though there are millions of people who are looking at Apple News, it’s not going to have the scale of another ad network. So I’m betting this is the first step towards something much bigger. But that will take a little bit of time. And for me, there is a big unknown on the Apple side is they have not communicated how we’re going to track Apple Search Ads next year. The users are going to see these ads, whether they’re LAT on or not, whether they consent or not. But as far as everybody knows, Apple Search Ads is not going to be tracked through SKAdNetwork and Apple has not told us how it’s going to be tracked yet. My personal feeling is they have something for themselves. The whole industry is going to go through SKAsNework, and they will go through another system that they have not communicated yet about.
And if Apple is listening to that, hurry up, because we’re a month there and we all need to prepare. Please at least tell us a little bit about it because we need to prepare. I don’t know why they haven’t communicated on it, but they are walking on a very thin line because at the moment they start making an exception for themselves by selling ads. And I think from a legal and competitive point of view, this is at the very border of what can be targeted by the regulator. And that might be the reason why we don’t know yet. And I can’t tell you how it’s going to be, but I can tell you, I’m super curious how Apple will get away with “you’re not tracking the same way as everybody else but this is legit to the eyes of the regulator”. And I’m looking at this topic with a high interest also because I’m an operator on Apple Search Ads. But I’m a little bit concerned that in general, we still don’t have the answer. So let’s see how it goes – I’m curious about that one.
One of our guys noticed that Apple literally just released an API that allows them to claim attribution on users who don’t opt in the ATT and associate it with keywords. And I don’t want to get too speculative on the long term implications of that but it seemed like that might kind of suggest that they were going to go to the SKAdNetwork framework.
Thomas, I definitely share your concern. I definitely share your thoughts that if they were to build their own outside system that kind of favors them, that seems like that would be quite anti-competitive.
Yeah, at the moment, they haven’t registered themselves as a network on the SKAdNetwork, but they might move there because the other solution is too dangerous. I think this API will be sort of an iteration of the self-attributed API that they’re already offering.
I will continue with the idea of rethinking business models, and we sort of think that, you know, like I said, whale targeting on IAP is a huge business opportunity for a lot of app developers. That’s going to be more difficult. And we wonder if there might be a bit of a shift towards almost higher value services that users have, like subscriptions assigned to them. And there’s a lot of buzz about subscription-based businesses now, they tend to perform really well when they have a good business model. And so we wonder if a lot of these changes are ultimately going to orient app publishers towards providing services that have a subscription tied to them.
I think when Apple announced this change, some people thought, oh, it’s because they’re privacy-focused, they implemented this for the benefit of the user. I mean, it’s partly true. You can buy this narrative if you want. Some other people started saying, oh, they want to sell more ads. So that’s why they’re doing it. They want to favor themselves. I think this is still pretty unknown, like just Ryan and myself commented. But too few people are commenting on this shift of business model. And something that is clear is that when you monetize through ads, Apple is not seeing a cent of that. When you monetize through subscription or IAP, Apple is getting 30 percent, in some cases 15 percent – I don’t want to open that Pandora’s box today – but they’re getting a cut out of it. And for me, this is a major incentive for Apple to have made that move, is that if they provide a strong incentive to developers to monetize more through the place they get cuts versus the other one, it’s beneficial for them from a business perspective. So this was not a pure user-driven move. This one is a strong reason for Apple to keep developing the services business by incentivizing developers to move to the subscription or IAP doesn’t have to be a subscription. I think this is a strong, strong reason we’re seeing this change.
One very last thing about business strategy in Apple’s business strategy – what I think they’re trying to do is trying to become the leader in tech. And I think that the SKAdNetwork, the deprecation of IDFA is certainly going to hurt Facebook and its business model. And they’re the main competitor. And with this move, I think they’re also going to influence the industry with their view and possibly, you know, kind of forcing and not forcing Google to follow the same stuff.
Yeah, what I think the steps towards ASO will be – more companies will be going towards that direction, not only diverting from the duopoly, Facebook & Google, and budgets will be distributed more evenly. And ASO could actually be one beneficiary if this really happens. However, I do not think that ASO is actually the silver bullet that could mitigate the consequences of IDFA depreciation. I do think that, you know, there’s going to be more budgets and more and more attention to that, so there’s probably going to be more activities on ASO.
Maybe, as those budgets start to shift towards ASO, we might see an increase in testing new assets in the App Store. I don’t really think or foresee new approaches to testing methods per say. But, you know, if indeed there will be less emphasis on UA, then it will be more important to stand out from the crowd that creatives will become more and more prominent, will be trying to test different value propositions, different tests, A/B testing, of course, will be adopted.
And I think, you know, in the coming months, this will become increasingly important, and not only for ASO. So creatives will start to play a much more important role, for paid UA as well.
I think creatives will become a key element in converting users.
Not that they were not before, of course, but it will be another weapon. Let’s say that we should leverage more in the coming months.
For me it’s also hard to say what exactly will come into place next year. But preparing for this panel, I thought maybe iOS is a little bit going away from the very strong, like Thomas mentioned, ROAS focus we have today, and CPA & CPI that were before, going away from this more to awareness and a branding-oriented campaign structure.
Going away from this more to awareness and a branding-oriented campaign structure.
Because of the lack of tracking and the high costs we have within this, if we will act targeting users specifically on the user level, maybe it’s a little bit more beneficial to say, OK, we’re going to the upper funnel and it’s targeting users on maybe reach, impression rate, something like this, to talk to these users on this level and with with other business goals. That’s the thing I could believe is happening. And in general, I also believe that for many verticals where the transition into the ATT and conversion values, as we discussed before, is not that easy, marketers will shift budgets away, simply as that. And maybe two other on iOS, but more and more also on Android, if they want to go a similar path in the future, which I don’t believe, but maybe we’ll come to this on a later stage and also web, of course, with landing pages, with websites also depending on in which business model you are in.
Question on chat: How much more expensive Facebook traffic will become on average?
This question is related to the point how performance will evolve, even if it’s a narrow scope of it.
My answer is it’s not going to be more expensive on the inventory side, might be more expensive for you because it’s less converting and less profitable. So you might consider it more expensive in that matter that you made less of a return from it. But I think it’s going to be cheaper. And the reason for that is one you look at today networks that have the LAT traffic versus IDFA traffic, and the LAT traffic is cheaper, somewhere between 60 and 30 percent cheaper. Does that mean Facebook is going to be 30 percent cheaper? No, but it means it is likely to be a little bit cheaper. Even though in life, I don’t really believe in the hidden hand in the market, I think if the targeting capabilities of Facebook and Google are going to be less because there’s no lookalike, there’s no user level data or the machine learning they built on identifying these highly converting users as well and so on – is going to be less than today. And I’m not going to say it is going to be zero. They have very strong first-party data to make extrapolation – but let’s say it’s a little bit less powerful. It’s likely that marketers are going to adjust and say, oh, I’m getting less return, so I’m bidding less, I’m shifting budget on the web and on Android and on whatever, just like Johannes said. And all this is potentially going to put less pressure on the inventory. Also will have one hundred percent of the user on an app installed compared to 60 to 70 percent today. So I think there are several forces that align to make that inventory. And I’m talking here: purely CPMs are going to go down as a reflection that how do I convert an impression into a dollar on my side is likely to go down because of the optimization. So the answer and I think it’s just a narrow point of your bigger question, but I think it’s a good answer to the question above – how much more expensive is going to be Facebook? Facebook is going to be cheaper on the inventory side. Your role is to make as much of it, as possible, from a profitability point of view and from the limited data that you have. Is this lower price point still something that is worse for you investing in? This second answer for some developers is going to be yes, for some other developers it is going to be no, and this is actually a very interesting question: how you transform this inventory into retained users and into money. But I think the inventory is going to be a bit cheaper. And I agree that this move from Apple is hurting directly, its competitor and its rival in leading tech. Because ninety something – around 98 percent of Facebook revenue is coming from the feed CPM, and if the CPM is going to be going down, this is really bad for them.
I’m not saying you should sell your Facebook stock, but that might be a decent idea actually. And don’t like my investing advice that is just a disclaimer [all smiling].
One of the things that we’re thinking about a lot is really the heightened importance of first-party data, both for publishers and for advertisers. For the publishers, the better ability that they have to segment their audiences and then allow for precise targeting just within their own systems and not relying on IDFA, but their own internal identifiers and their ability to serve ads due to their targeted audience. We think that that’s going to have a big impact on their ability to sell at higher CPMs and maintain their margins there. But it really comes down to having really good first-party data that you can use to segment your audience and target.
On the flip side of that, from the advertiser perspective, if we were thinking I mean, I don’t particularly like the word hack because I don’t think this is a hack per say. I think actually this is just doing your due diligence and really putting a lot of effort in to do a good job with this. But your ability to really understand exactly who your users are, so that you can then go out to the publisher world and try to find similar audiences, this is going to have a big impact on your ability to get high quality users and ultimately turn a profit on that.
But it starts with you having a good sense of who your users are. And to do that, you either need to be collecting your own first-party data. So kind of thinking about your app and how you might collect that or, absent that, potential sort of audience enrichment type stuff as well, or maybe, looking at data acquisition. I don’t know if I’d call it a hack, but just just thinking about how much can I really pin down who my users are, then go out into the publisher world and try to find similar audiences – we think that’s going to be a renewed importance in direct deals, which we often see on the brand side of things. So that’s one thing that we think might play out.
We had a little chat with Johannes previous to this, considering what hack we are going to recommend. And I think we both had pretty much the same answer. And here what we see through what Ryan is saying is: this is not a hack that in one day you start exploiting, this is a strategy you build over time. This is something that you have to think strategically for a very long time ahead and deploy progressively, striving to enrich your data. And I think that is the answer.
If you’re looking for small hacks that produce a huge uplift tomorrow. Well, good for you if you find one. But many of those are just temporary. And I personally hate the term growth hacking because it makes people think that any small change might bring you 50 percent uplift in performance.
And growth has to be seen through a sustainable point of view. So which hacks were efficient and are not going to be efficient? In my opinion, thinking about hacks is already the wrong way to think. It is with a sustainable strategy you’re going to put in place and doubling down on your first-party data is definitely one, but it’s not sort of a hack here. This is definitely something about thinking big.
And there’s an anecdotal point that I didn’t mention, not a hack, but some strategies people have been developing say, oh, there’s no more IDFA. And we’ve seen some strong consolidation, especially in the hypercasual space and gaming this year, big acquisition. It’s an open question, not an affirmative. But I was thinking maybe these developers, they’re preparing to have a bigger portfolio so they can exploit a bigger first-party data across all these apps and actually prepare for the fact that the IDFA in the outside world is going to be less. So maybe your internal portfolio, IDFV (the identifier for vendors) is more important. And it led me to think something very interesting. Of course, this doesn’t apply to the big developers. But Apple just announced that they would lower the cut on developers making less than a million a year.
And what this means is you remain independent. Let’s say you have a portfolio of 10 apps, each making one million a year. Then you have a strong incentive to keep them separated, to have a lower fee on it, which means that you’re not going to be able to use the strategy of bundling them together to use the IDFV. And I don’t think this is the reason why Apple did this change. And they have a much bigger reason to do it. But I think it was funny that they were making a move that also blocks part of this possible strategy that you might want to deploy, if it’s relevant to your case, might be in some areas. It’s just my very narrow thought on the big question.
I think it almost and you want us to have something to it, or Thomas has already presented your point of view here.
I’m not a big fan of the term growth hacking either. Maybe a small point, Andrea, I guess, mentioned early on: if you’re searching for this hack (or what you want to call it) – so this may be a little bit short-term boost to your marketing strategy, which is (I’m totally in line with the others) – not the best way to think about it.
But if you could think about it like this,
Invest early on in new channels.
And this might be iOS14 with the campaign structures we have there, this might be something like TikTok, this might be a new ad network, this might be Apple Search Ads if new channels approach there. So taking these sometimes cheaper and and good placements, if they come up and don’t just wait months before it’s common knowledge and everyone does it – might be a little bit kind of a hack if you want to. So this is always a thing to consider in your strategy along with this whole big approach, like Ryan and Thomas said.
We have iOS 14 now, and with iOS 13 and before, the LAT on users were actually split between LAT on and LAT off. That is not going to be any more the case. So you will be finding iOS 14 users only in total downloads. So Apple is not going to make any distinction anymore for between LAT on and LAT off in Apple Search Ads.
For me, there are several fundamental problems in SKAd, but there is one that any conversion that is coming after day 1 is uncertain that you will get it because of how the timer is built, which is kind of a topic I don’t want to dig into too much. But the summary of it is everything that happened after the first day has a high level of certainty. And given that, I believe there is a wrong incentive here for developers to fit into the first day something that is valuable to them. So typically, I work a lot in the subscription business and you see some subscription. They’re very smooth and open about their paywalls. So you can experience the app before you reach the moment you have to pay, like, for example, in Duolingo. And some developers are a bit more what I call aggressive, putting the paywall very upfront. And I think this change in SKAd is going to push developers to adopt more of the aggressive monetization strategy, which eventually is not great for the user, is not necessarily great for Apple either to have this particular user experience. And I think here is a big problem of providing an incentive to developers to be aggressive. So, on my wishlist for SKAd is: please extend a little bit the window of time in which we can get the postback and this conversion value because it’s in the interest of us to measure better the action we do, but it’s also in the interest of users so that we’re not incentivized to urge users to pay. I think this is a bad consequence of how SKAd is built, and I seriously hope that Apple is going to make an iteration on this timer thing that is actually quite complicated to embrace.
This is a big unknown, to be honest. And it’s actually not yet clear whether or not what we’re starting to call probabilistic attribution is actually going to fall within the ATT framework or not. So to some extent, it’s still kind of TBD on where that goes. But I suppose just at a personal level and again, you know, sort of thinking this is a bit speculative, but I struggle to see how probabilistic attribution is going to go away entirely because there are some big channels even on iOS that rely on that. And the big one is mobile web, where (I know that’s been outside of the performance marketing world), but there are still lots of brand advertisers doing mobile web campaigns and they might want some type of conversion tracking. And so, you know, it’s just kind of thinking about the implications of actually doing away with probabilistic attribution. I just think that it would totally destroy some big segments of the advertising landscape that would actually impact Apple as well. So I don’t want to say it’s not going to happen. It’s still something that’s changing as we speak. But there would be some huge implications if they get away with that entirely.
I’m with Thomas, I also believe that traffic might be cheaper on Facebook, depending on if they are able to sell it elsewhere. And this will be hard if we are going to see non-IDFA users 70 or 80 percent. So I think then they are not able to sell it on the web or e-commerce or that kind of stuff. So, traffic might be lower because people are spending less due to lower conversion rates and lower ROI and ROAS on it. Facebook and other networks have to react on this because this inventory will not be sold if they are going to lower prices. That’s my opinion but it’s really broad because like with most things today, we don’t really know, to be honest.
I wouldn’t suggest that you get rid of LAT on users because you would get rid of valuable piece of inventories. So I would suggest that you keep targeting them. Some people argue that those are even more valuable than LAT off users. And in regard to the second question, how iOS installs will look like – well, in Apple Search Ads right now, those installs are undifferentiated and those are being attributed in your total downloads column. Actually if you subtract the LAT off and on downloads from the total downloads column, you can even calculate how many iOS 14 users you’re actually acquiring right now.
IDFA is not completely gone. It’s gone as the standard for attribution.
And the new reality is going to be that there is a world, a cohort and universe with the IDFA, which is the one you will need to request the consent from now on.
Before you were going, users were coming in, you’re just pulling the IDFA, a few people didn’t have one. That’s life. In the future, by default, you don’t have it, but you can request it. This is what the ATT framework is doing. Like Johannes explained at the very beginning and we mentioned, you might want to ask for the consent or you may not. If you don’t request the consent, the IDFA is gone for you entirely. You may want to request the consent in some cases. And here you have to read on the line of Google’s documentation.
What Google is trying to do is push onto you the request and force a lot of developers to request the consent so that they will keep feeding the machine learning, which means the optimization will still happen partly, because they will extrapolate from the part of the inventory that has given the consent, and they will be able to maintain the current products that they have.
So if developers don’t request IDFA and IDFA is gone for them, it gives Google less data-point to actually push where we’re likely to exploit this data the same way they were before. So is the IDFA completely gone? No, unless you decide to not request the consent because you don’t need it. If you’re spending a lot of money on Google UAC, my understanding of that is Google is going to extrapolate from this cohort. So you have a big interest in requesting the IDFA, providing this data for Google so that the optimization keeps happening. This is speculative because 1) I’m reading between the lines and 2) the new Google product that is going through SKAd is not entirely released at the moment. So it might go through changes at the last minute. I can’t say exactly how it’s going to be, because I haven’t seen it with my own eyes. But what this text is strongly suggesting is they want to keep it extrapolating from the IDFA cohort and then they push you to request, this is in their interest. It might be also in your own interest to align with Google interest and make this work. I hope this brings a little bit of clarity on why the IDFA is gone and why Google is asking us to request permissions. Anybody disagree with my interpretation, which is, I admit, speculative?
No, because they already do this to some extent with the search traffic, right? They don’t say that like in the spotlight, but they are already doing this. So might be the easiest thing for them to do. And maybe they’re just pushing this a little bit further into 2021, because iOS is not their main source, it’s Android. So they’re not that dependent like Facebook on this. And on Android, they have first-party data all over the place. So they’re in a pretty good position to just sit and wait. What other their competitors like Facebook and also to some extent Apple will do and then react on short notice. So I’ve nothing to add on this. I think that’s a good way to think about it.
This depends a lot on how you’re mapping your conversion value in this case. Let’s say you have a main event that you’re gearing to maybe starting a trial or maybe completing the first level (the tutorial) or maybe doing the first purchase. You are able to pass this conversion value through SKAdNetwork and then you are somehow capable of measuring your CPA. There are a few new nuances here. One is Apple mechanics to attribute is different to the one you had before, so the CPA is not the equivalent. We mentioned that there is this one day limitation. So it’s only the first day, and you might want to extrapolate what is the first day conversion compared to the all time conversion you had before. There is a limitation that before we were provided with the view-through conversion – impresion view attribution – and here we’re not.
So you do have a CPA, but you need to make a bunch of extrapolation to be able to compare it as before. And I think here is just we’re going to have a new reality. There’s going to be a new number and a lot of the work is just going to figure out what is my new target. I used to have a target CPA of $20. What it is going to be in January, I can’t tell you. You need to map out how much impresion view is, how much everything after the day one is. Is this event exactly the same and so on and so on. And that’s why I pushed at the beginning to run new regression analysis to determine if the event you’re using today is actually necessarily the one you’re going to use tomorrow. And will you keep using CPA and stuff? Yes, but not in the same manner as you were before, because the whole mechanics is different.
Editor’s note: Big thanks to the brightest minds of the mobile app industry: Thomas Petit, Johannes von Cramon, Andrea Raggi and Ryan Goeden – for sharing their thoughts, concerns, insightful recommendations and actionable tips. Sharing is caring.
