📱How mobile measurement will work with iOS 14 in a post-IDFA world – with Gadi Eliashiv, co-founder and CEO at Singular🤓

Our guest today is Gadi Elisahiv, co-founder and CEO at Singular. Gadi is among the folks who has been deeply involved in conversations about IDFA deprecation – and what the world can look like post iOS 14. Today we talk about what might happen to MMPs – and how we might work with SKAdNetwork in a post-IDFA world.

KEY HIGHLIGHTS

👉🏽Fingerprinting, attribution hash, and customized popups – can these work under iOS 14?

👣The default path that Apple wants us all to use: SKAdNetwork.

⛔️The challenges and shortcomings of SKAdNetwork — conversion values, campaign IDs, postback, deduping, and more.

😬How the way the postback is currently architected can result in fraud.

👀What will be the role of MMPs post iOS 14?

🤦🏼‍Why aggregating data is difficult – and challenging for in-house advertisers to take on.

🌟How there will be a lot of innovation once the basics are established post IDFA deprecation.

📶Is conversion tracking becoming simpler or more complex with iOS 14?

🕰Will Apple release improvements to SKAdNetwork over time?

💸What might happen to ad spend and CPMs in October.

👩🏻‍💻How important it is for companies’ senior management to understand the IDFA issue and its ramifications.

👊🏽How advertisers are preparing for the upcoming changes.

🧐What sort of product changes apps are contemplating.

KEY QUOTES

How fingerprinting works

Fingerprinting is based on IP and user agent, it is more susceptible to fraud, it’s less accurate, the windows have to be a lot shorter. But that’s basically a pretty simple technology that everybody has. The challenge with that, and we have said this from day one, is that we don’t think fingerprinting is going to pass the Apple test.

Apple has always been privacy-conscious

We had one of our customers talk to Apple, and they said no. Basically the first time they said no, and they said something like, “If you could use fingerprinting to derive some stable ID, even if it lasts for two minutes, that’s off the table.”

Workarounds still need favors from Apple

There’s an idea called attribution hash that suggests, “Hey, if we can get access to the IDFA on the device and combine it with the IDFV and kind of make a hash, then only send that hash value to the server, that would basically mean we’re not leaking the identifier, and it would stay on the device.” The challenge with that is it means that Apple needs to do us a favor and give us the IDFA. That means, it’s unlikely because they just killed it very publicly. Why would they have an appetite to give anyone that access again?

There’s a tricky balance to be struck

The challenge is that there’s a tension between preserving privacy and giving the advertiser flexibility. The more flexible the mechanism is, the less privacy friendly it will be.

SKAdNetwork is not standarized

The first thing is that there is this notion of conversion value in SKAdNetwork that enables you to pass some form of a user value back to the ad network, so that they could optimize. That is completely non-standardized. For one advertiser value 17 could mean one thing. For another company, value 17 could mean something else — maybe something bad.

Setting up a framework

What we try to do is say, “Here are some standard models. If you want to get ROAS reporting, let’s use the conversion value, the 6 bits, to encode how much revenue, maybe ad revenue or IP revenue, your users generated.” Maybe we want to capture also, how many days has it been since the install? We’re trying to come up with all sorts of reporting because eventually advertisers would like to see some ROAS. I guess challenge number one is how do I govern conversion values in a standardized way?

Data can potentially be corrupted

There is a challenge right now where the way the SKAdNetwork has been proposed is that the postbacks go to the ad network. In some of the data, these postbacks have been signed by Apple, which is excellent. Some of the data has not been signed by Apple such as a conversion value. Also, you could say that the device that’s sending the postback, the geo of that device, is also not signed, it’s just sent on the network. So you can see what country the user came from. If it’s not signed, it means that the ad network could change it. I’m not saying they’re evil, but what’s interesting is, all it takes is one bad actor to disrupt the trust in SKAdNetwork. Think about what it means, it means that an ad network will report their spends, the number of installs, and their ROI — that’s insane. It’s like they can do anything, they can just report that it’s all whales.

Non-standardized processes can be problematic

I’ve seen people in the past try to build some of these things in house, and it always eventually fails. It sucks. One person wrote it, he leaves the company, they’re stuck. It’s not standard, it falls behind, etc. So I don’t think this is any different. There’s been examples of companies who built their own MMP stack internally, and it would have a lot of holes, it wasn’t really smart, and it had fraud, etc. I think it’s gonna be the same thing, but just different because you will still need to figure out: “How do I receive all this SKAdnetwork data?”

MMP still have a part to play

In a way, I think advertisers will still kind of hire an MMP. They’ll say, “Hey, I need to figure out all this mess. Please be my technology provider,” which is how you buy SAAS software, and help me sort out all this mess.

IDFA is an organization-level change

We talked to some folks whose upper management doesn’t even know it’s coming. It’s like, “Oh, yeah, my technical team will solve this IDFA issue,” but they need to understand it up to the CEO level.

First mover’s advantage applies

There’s these stories whenever a new channel comes out, and those people are early adopters, and they enjoyed really low CPIs and just grew their business really well. I’m sure it’s going to be the same. Some will adapt quickly and will take the advantage, some others won’t.

Expect a learning curve

They’re figuring out what changes I need to do to my stack? What is the new reporting gonna look like? Let’s say that I use the conversion value to encode revenue for the first day or two, can I even rely on my buying patterns? So they audit every one of their games, and they’re trying to figure out — can I live with that kind of reporting? They’re trying to figure out how should I encode my conversion value? They’re talking to us and other people to figure it out because they’ll need to have a way to transmit that back to the ad networks. That’s the preparation they’re doing. They’re auditing scenario by scenario.

FULL TRANSCRIPT BELOW

Shamanth: I’m very excited to welcome Gadi Eliashiv to the Mobile User Acquisition Show. Gadi, welcome to the show.

Gadi: Thank you. It’s great being here.

Shamanth: I’m definitely thrilled to have you because you’re certainly been very engaged in dialogue around IDFA deprecation, and certainly you and your team have been among the folks who have proposed solutions to what may be perhaps the best ways to navigate it. So definitely thrilled to have you today and definitely thrilled to talk about the huge changes that are overt that we are encountering.

To start off, and again, in very many communities, there are many possible solutions or ways to navigate iOS 14 that have been discussed. Out of all the things that have been proposed and discussed as possibilities, what do you see as realistically possible? What do you see are not realistic options?

Gadi: I’d say it’s a great question. I guess it’s great because I’m asking myself the same thing all the time as we’re navigating this. I think there’s a few phases that we’ll be going through. The first one is when Apple just released it, there’s the initial shock. We’ve been preparing for it. I think everybody’s been speculating that it’s gonna happen for a long time. We’ve been preparing for it for a long time. We started this mobile attribution privacy group about a year ago. It used to be that we just invited everybody to our office — it was 15 companies — and now it’s over 900 companies, so I’ve learned a lot from that group.

I could try and summarize my thoughts. One of the things we try to do as much as humanly possible is be pragmatic about our approach to this because, yes, there are some advantages to our platform over traditional MMPs, but we didn’t want to offer solutions necessarily to highlight these. We wanted to do what’s best for our advertisers, and it’s hard to do because I need to run a business. I’m the CEO of an MMP company at the end of the day. So how do you do that? There’s a range of different solutions that I see right now. Some may provide certain advantages to individuals, some may not.

Let’s start with the first one: fingerprinting. That by definition has been the default fallback for every attribution flow. If you don’t have IDFA, you have LAT on or maybe it’s a mobile web campaign — the MMP will default to fingerprinting. It has a lot of issues.

By the way, just to clarify, every time I’m saying something about a method that is unlikely, it doesn’t mean I don’t like that method. I wish it would work, but I just don’t think it would pass the Apple sniff test. For example, with fingerprinting, everything they’ve done in the last few years suggests that it’s unlikely that it will work.

You’ve seen ITP, for example, Intelligent Tracking Prevention, on the web, and what they’ve done in WebKit. They’ve been very aggressive to attack cross domain tracking, to standardize how you pass parameters between ad clicks from one page to another, and they prevent certain API’s from going into Safari because they’re afraid it’s going to open up fingerprinting. Also, in the WWDC video, clearly everybody was playing forensics with those videos — watch the same 30 seconds again and again and asking, “What did they mean by that?” So there’s a screenshot where they say that tracking includes a fingerprinted ID, and we’re like, “Okay, what is a fingerprinted ID?” It’s not very clear, so we were very skeptical.

Then recently, probably the most important news, is

we had one of our customers talk to Apple, and they said no. Basically the first time they said no, and they said something like, “If you could use fingerprinting to derive some stable ID, even if it lasts for two minutes, that’s off the table.”

If you think about it, the MMPs bifurcated into all these different solutions, so you now have two vendors that are talking about device graphs, and it’s very probabilistic, but it’s also very accurate. They talk about fingerprinting or maybe they don’t use the word fingerprinting, but I just don’t understand how it’s gonna work and how it would work with Apple, so I am not as hopeful about this method. If it does, Singular would have happily adopted it etc, but I wish someone from these companies would call me and say, “Gadi, here’s where you’re wrong. This is why device graphs are going to be okay with Apple.” So that’s method number one.

Another approach was how do we bring back the attribution we used to have and love. There were some cool ideas that came out such as attribution hash that basically said you have a few assumptions. First one is, you need to get consent on the publisher side. So we have that flow where you see an ad on a publisher app, you install it, you install the advertiser app. So if you get consent on the publisher app, maybe it’s okay that the advertiser app would know where you came from because you gave consent in the publisher app — that was the logic, which I actually like.

The implementation relied on these clauses that Apple included in their policy. There’s a clause that they added that talks about on-device tracking, and it says, “If you’re able to do any sort of tracking that remains on the device, that’s okay with us as long as you don’t send it to the server, and you can’t identify the device, etc.”

So there’s an idea called attribution hash that suggests, “Hey, if we can get access to the IDFA on the device and combine it with the IDFV and kind of make a hash, then only send that hash value to the server, that would basically mean we’re not leaking the identifier, and it would stay on the device.” The challenge with that is it means that Apple needs to do us a favor and give us the IDFA. That means, it’s unlikely because they just killed it very publicly. Why would they have an appetite to give anyone that access again?

Maybe they could do something else that could give you a hash value, that attribution hash as well, but it also has some privacy issues unfortunately. We kind of explained that in the Slack group in MAP, but essentially once you have the IDFA of a user once, you’ll be able to track him everywhere. There is a way to do that computationally, but it has some holes, and I think with Apple’s tolerance to privacy is so low or high or however you call it, they just wouldn’t accept anything right now. So that’s another approach.

There’s a lot of discussion about maybe we can get consent, maybe everybody will be really good at getting consent — the publishers, advertisers, etc — and we’ll have IDFA. And we can just use IDFA. Recently, I want to tell you of another angle, which is really interesting. There’s the GDPR consent and CCPA and all sorts of pop ups now. In Apple, also, I saw that they changed the word consent to user permission because they didn’t want to say their pop up equals GDPR consent, it’s a different thing. So what if companies could come up with their own pop ups, get approval, and then use that to to justify fingerprinting. Meaning: “Hey Apple were not going to use your identifiers, but we’ll get the consent, and we could use fingerprinting.” The challenge is that Apple said, “If you do any form of tracking and fingerprinting is part of it, you have to use our pop up. That’s our policies for our ecosystem or platform. If you don’t follow our platform policies, we can kick you out.” So I think that’s the range of solutions.

Shamanth: I think that covers all of the options I have seen out there. And you’re right, it will come down to how tolerant Apple is to anything that could identify a user individually. I think that’s really the crux of it. Of course, the default solution that Apple would rather have everybody use is SKAdNetwork. Just to jump to that, SKAdNetwork has been proposed by Apple, the details are out there, the documentation is out there. If an advertiser is looking to adopt SKAdNetwork, what could be some of the challenges there in the way it is currently architected?

Gadi: That’s a great question. When I mentioned all the methods, I forgot to mention that one because I wanted to mention the alternatives to the default path. But yes, Apple released SKAdNetwork, and they said, “Here’s our suggestion.” It almost feels like they were proud of themselves that they innovated, and it’s a pretty cool idea that they innovated to create something that’s privacy preserving, and it gives you attribution that doesn’t leak user identity.

The challenge is that there’s a tension between preserving privacy and giving the advertiser flexibility. The more flexible the mechanism is, the less privacy friendly it will be.

I’ll give you an example. There’s been a discussion about, “What if we could just get multiple conversion values out of SKAdNetwork or postbacks?” Very easily, you could prove that this could leak the user’s identity, which is why Apple created such a limited mechanism because they really want to be strict. It depends on how much risk they want to take of someone exploiting that. But what they built right now is really hard to exploit, meaning that it will be very privacy friendly.

Now, what are the challenges for advertisers? I think it’s not just advertisers, it’s everybody because an advertiser lives in an ecosystem. He has ad networks that need to work with him, and he has some solutions in the middle, MMPs, etc. We’ll talk about that later. But the single advertiser can work with a bunch of ad networks, and they want to know that when they onboard the new network that they want to test, they speak a common language. Vice versa, the ad network needs to work with a lot of advertisers, hundreds of advertisers, so there needs to be a common language.

SKAdNetwork, as Apple proposes, in its current form is very bare bones. It’s okay, they gave us an API and said, “Go figure it out.” One of the things that we’ve done very actively is we said that we have to be pragmatic. If this is going to be the new reality, we have to accept that people will be stuck with this solution, and we’re going to be stuck with this decision for a long time. How do we make it better? Here’s some of the challenges that I think people will face and what we’re proposing to fix.

What we wanted to do is say, “Hey, let’s define methods to standardize what kind of models we want to use this value for.” Now, it’s very limited. It’s only 6 bits, which is 64 values. There’s all sorts of quirkiness with how the mechanism works. You have these timers, and they’re trying to anonymize the message.

What we propose, and we wrote a document called SKAN, which is short for SKAdNetwork, and it’s our proposal for how to implement that. We built it in a way that’s not Singular specific, it was actually benefiting every advertiser, ad network, and MMP.

What I just described is the most simple implementation. There’s a lot of sophistication you could do there all the way to you do LTV prediction, and based on that, you assign a conversion value on the predicted LTV for that user. There can be a lot of fancy methods as well, but someone needs to standardize that. So that’s challenge number one, conversion value.

Challenge number two is campaign ID. SKAdnetwork gives you 100 campaign IDs to work with — super limited. You could change what each campaign ID means. So, if I’m an ad network, let’s say I’m IronSource, and I want to serve ads. Their machine learning is probably going to assign a different meaning for campaign ad 5. Maybe one day it will be to US users with certain publishers and another day to Germany users with a different device model. They’re first going to control that conversion value. The challenge for advertisers is, what the hell does that mean? How do I translate a number between 0 and 99 or 1 to 100 to something readable? So that’s challenge number two.

Challenge number three is one of the worst things in my opinion. I wasn’t very vocal about it in our article, but this is a good chance to talk about that.

So that’s a big gap in my mind, and I think maybe Apple will fix it. I’m hoping they will. They probably want to fix the country issue, so that’s still going to be an issue.

What if I wanted to buy users in the US, and I’m getting users from another country, I can’t verify that. We propose an alternative in SKAN that the postbacks will be sent directly to the advertiser and whatever stuff they have to receive these postbacks. I think Apple should agree to that because the advertiser is probably the most trustworthy player out of everybody. Also, the postbacks have no PIIs, so what are you going to do with that? It’s not like they can abuse it. I’m really hoping Apple would approve that tiny change to send it to the advertisers. Then the advertisers can ensure it’s not being changed. I think if you’re an advertiser, and you’re listening to this video, please ask your Apple rep to push for that. I know our customers are going to push for that. I’m sure other MMP customers are going to push for that. I think these are some of the initial challenges.

Then there’s other challenges such as how do you dedupe when you have consent? Some users do give you an IDFA and some users don’t. How do you combine your anonymous data that has a lot of deep cohort information with your very limited SKAdNetwork data that only gives you like two days of activity? There’s been some cool ideas in the industry about that. Lastly, what else is challenging is testing. How the f are you going to test this now? No one is ready. There’s barely any documentation. No one’s barely speaking about SKAdNetwork. Ad networks are starting to adopt it, but the big ones are being cagey about it, so you can’t even test it, and it’s coming out on September 15th — holy shit. So yeah, that’s a lot of challenges, unfortunately, but there’s no choice. We have to do it.

Shamanth: Certainly that’s the default option unless Apple says yes to fingerprinting or quasi fingerprinting, however you call it. You’re right. It’s certainly a huge shift from the older paradigm that we have gotten so accustomed to just getting precise ROAS numbers. What I would agree with is that I think the challenge is that a from lot of the ad networks, including Facebook and Google, we’ve heard sort of blanket party line statements from Facebook and Google, but nothing super concrete yet. So it remains to be seen how much of an impact that was going to result in.

In all of this in the earlier paradigm, I certainly think that MMPs played a very critical role in adjudicating where the installs were coming from. That centrality of that MMPs role is no longer going to be present come September. So what do you see the role of the MMPs being? They may not be MMPs, they may become something else. But what do you see that role being post iOS 14?

Gadi: It’s another very interesting question, and I could give my POV. I think certain individuals like to talk about, “Oh, this is the death of x or y or z.” I think it’s overly dramatic. The way I see it is as follows. The world just became more complex, the chances that customers that are going to build all this stack themselves are very, very tiny and, by the way, there have been examples.

Singular sells two products — we sell an MMP product, and we also sell a data aggregation product for all your marketing data, etc.

Let’s say that Apple agrees, and you now get it as postbacks. So are you going to build your own real time server to get these postbacks, and then forward to other ad networks? You could, but it’s insane, you can do that. Then you need to make sure that they are not deduping. They’re not sending duplicates. So you need to start setting up your own fraud mechanisms and then also do the BD with ad networks to talk to them and say, “Hey, you’re sending me duplicate postbacks, you’re trying to cheat me.” You could, but it’s intense. You need to build the layer that manages the conversion value.

When you send events, you need to translate that to conversion value, and then also tell the ad network what it means. So can you do that? Yes. But again, it’s intense. You need to aggregate all the data, all the spend from every single channel you’re using on the campaign. You need to map the campaign ID from SKAdNetwork to something readable, you need to build the reporting. Again, all these capabilities, the world doesn’t change as much to me because of the necessity I think MMPs always had is they solved the technical problem that was pretty hard to solve. And yes, they were in a central position. But a lot of the power of the product came from just like how you buy any other SAAS product.

My company could build Zendesk, I can build it, but it would suck. That’s why I’m buying Zendesk. I could build Salesforce — I mean, it’s a table with attributes, but my sellers wouldn’t know how to use it, and it wouldn’t be standardized. When I hire a new seller, he wouldn’t know how to use it, so that’s why I go for the standards. I think that the role of the MMPs will simply evolve, and you still need to provide unbiased measurement.

And yes, maybe the way that clicks were being forwarded to the MMP before, some of it may change. Actually, that’s not even going to change because what if you get partial consent, then are you not going to look at the IDFA? You probably still want to have it, so you’re going to have the MMP just for that portion. It’s just gonna be such a complex universe, which is why I think you’ll still need it. You’ll still need to have someone help you with all that stack.

That’s that’s where my confidence comes from. Plus I mean it’s happening in a few months. It’s insane. Who can even get ready for that? You need to build a massive organization to fight that, to build that. So that’s why I’m pretty confident. You can challenge me. I’m an engineer by background, so I know it’s not just some bullshit salesy stuff. I know what it takes to implement this. And it’s massive. So I’m like, “Who’s gonna do that?” By the way, there’s Android and there’s the web. There’s deep links. So there’s all these other workflows that you still need to figure out. Probably a million other things I’m not thinking about right now, but this is my honest on the fly answer.

Shamanth: The one counterpoint I might offer to that would be if the primary function of the MMP is aggregating disparate sources of data, they get 10 ad networks data, is that something that requires less of a technological expertise than the current function of MMPs? If all it takes is getting 10 spreadsheets in a certain format, and just imbibing it into that internal data system. How would you think about that?

Gadi: It’s a good point. I think it’s just not gonna be as simple. I think that you will need to get all these postbacks, so it’s going to be a ton of data. It’s not gonna be 10 spreadsheets unfortunately. Ideally, we get the postbacks directly sent to the advertiser or to the MMP server and not to the network. So you still need the real time processing capabilities of MMPs. You still need to build on the analytics framework on top of it. So kind of translate what it means for each individual network.

Maybe the reason I’m so confident is that we’ve been in this business for a long time aggregating the data. That’s even part of the solution I’m talking about. It’s difficult. People think it’s easy — it’s not. That’s why we’ve been successful as a company. It’s difficult; otherwise, I wouldn’t be here. I would not be able to speak to anything. I have six years of running the business to see as experience to people that thought it would be easy. So that’s one thing. The second thing is that’s the basics of where MMPs would begin.

There’s still other workflows, for example, whenever you have a deep link that opens the app directly, that flow works as usual, so people would expect all these functionalities working with an MMP. You have the deep link, you need to, in real time, check where the user came from, run some tests, maybe parse the parameters, you need to track the user and all their cohorts, etc. So some flows are going to remain the same, and I don’t think advertisers are going to throw them away. Then from there, that’s where the sophistication will start building. You’ll see certain MMP are starting to build in different directions.

For example, what is the next level of insights that we can give you? For example, SIngular has been working on the ability to look across a lot of customer data and give active recommendations and benchmarks, etc. There’s a lot of discussion about probabilistic attribution. The problem is every time someone says that it kind of sounds like fingerprinting, but that is different. It’s like how do we take all the aggregate post SKAdNetwork postback data? Maybe we can figure out what is the probability that someone came from a certain campaign or maybe earlier I mentioned predicted LTV, and I said people figure out what the LTV is.

This opens up a new chapter for innovation in the attribution systems, and I think you’ll see two phases. Phase one is we’re helping people figure out what the hell’s going on because we ourselves have to adjust. Everybody has to adjust to this new process and figure out what are we going to do? How do I still run reports? How do I see ROAS? That’s the question people ask me all the time. Unfortunately, as sexy as we may all want it to be, there’s going to be a lot of foundational work to make the basics work. Once the basics are there, you’ll see a range of innovation and sophistication coming out of vendors.

By the way, anybody can innovate, I just think MMPs will be in a prime position because they’re still very central in a lot of different platforms. They’re still needed to support a lot of different user cases. We’re so deeply embedded in all this stack, we’re directly connected to the network, so you’ll see a wave of innovation coming from there. Everybody can try. It’s a free market.

Shamanth: It’s a good point. In fact, we had an episode of the show out today with David Philippson, who founded Ad-X. I was asking him about what the early days of MMPs was like, and he said something very similar — a lot of companies tried to build their own attribution solution in 2012-2013, and found out it wasn’t nearly as easy. So I can certainly see where you’re coming from on that.

The other perspective also that I would offer to what you said is certainly I think there’s someplace for the MMPs in the ecosystem. One dimension of that is because SKAdNetwork network has one conversion value whereas in the past, part of what MMPs did was make sure there were 30 different conversion values, level 20, level 50, 20 IAPs. So that complexity was something that the MMP was handling. Now that that complexity goes away, do you find that that simplification adversely impacts what an MMP is actually going to be set up to do?

Gadi: It’s interesting because I actually see it the opposite way. In the past, life was easy. We just forwarded events. There was an event, we got it, we attached the attribution parameters to that based on the device ID, and we sent it to the ad network. That was simple. Now, life is more difficult. You have to live with that limited value. How do you encode that value? If anything, by forcing you to a more limited space of information, I think that algorithmic complexity is going to increase.

If anything, in the past, we got away with doing pretty simple things. Now, the sophistication needs to increase. So no, actually, and I could see it. I’m talking to my team. I have extremely talented engineers with the range of ideas and innovation that’s coming up because of this. By the way, caveat, I would love to go back to how they did in Avengers — go back in time, and stop Venice from flicking his fingers with all the Infinity Gems. I would love to go back and still have IDFA and wake up from this dream. But it’s true, it’s gonna cause a lot of innovation. So if anything, I don’t think it’s going to be simpler. Hell, it’s gonna be way more complicated because unless advertisers say “Yeah, I’ll stop effective UA,” but no one’s gonna say that, there’s going to be a race to what’s the better way to predict the value to encode the conversion value, to provide certain reporting like this.

That’s going to be a new generation of innovation because of what Apple did. I want to say, though, that I have a belief that there’s a balance between overly complicated solutions that are not very clear to practical things that work at scale, but I think that what we’re imagining with SKAdNetwork right now is not going to be the standard, it’s going to improve it. It’s not going to be insane, but it’s going to improve to a point where there’s more sophistication. it cannot just be that people report one value for level one and one value for level two. it’s gonna be better than that. It’s just complexity.

Shamanth: If I understand you correctly, you’re saying, “Look, that’s just one value, but you still need to figure out what’s the best way to represent that one value.”

Gadi: I’m sure Apple would create improvements for SKAdNetwork over time. They did version two. Unless SKAdNetwork is going to die and not get adopted, which I think is unlikely, then they’ll probably work on version three. They even added a field right now that says, “What version is this?” I don’t think they did that because they thought of the future. They even said to some of our customers that are going to work on iterations in the next few months, I’m hoping they’ll make it more useful. I’m not sure we’re gonna get a lot more flexibility.

Shamanth: We can be hopeful, but my impression was that the version is there because SKAdNetwork was already that — there was version 1.0, and this is 2.0. I would like to think there will be improvements. I don’t know how much we should bank on Apple to do things.

Although I will say, again, going back to the episode I was just talking about about the early days of device identification – when Apple phased out the UDID, there was a No Man’s Land period where there was no device identifier. And then Apple said, “Oh, we need to put something in, let’s bring in the IDFA.” So there is some precedent — that’s 7 years old. I don’t know if we should rely on Apple to solve our problems or listen to us.

Gadi: What people are hoping is that Apple would say, “Oh shit, I destroyed a vital part of my ecosystem by doing these changes. Maybe I need to help them a bit.” I’m hoping that part of it will happen. I don’t know if it would be publicly like they wouldn’t bring back an identifier, but maybe they’ll improve SKAdNetwork, etc. Maybe they’ll do something else.

Shamanth: Obviously one aspect of this entire shakeout is going to be what’s going to happen to ad spend and CPMs in October 2020. How do you see that playing out?

Gadi: That’s the million or maybe billion dollar question. I think that it depends on each advertiser’s readiness and the network’s readiness, which is why there’s so much urgency right now to provide people answers. I’m working with some other great partners, advertisers on improving SKAN. The reason we are doing that in a rushed timeline is because of what you just mentioned. If there was a way that was tested and verified to run some UA effectively, that would give people a lot of comfort because they can follow a path. People need something that’s been tested, even if it’s not the best, even if it’s not something very futuristic, but if it works, and it gives you some basic capability, this will give a lot of people confidence to run their campaigns. Doesn’t mean necessarily it’s going to be the same capacity that they do today. I think there’s going to be a learning period.

What I can tell you, which is interesting, a lot of our top advertisers see it as their advantage over the market now because there’s going to be a period of time where maybe the people are unprepared who don’t ask questions. We don’t know what they’re going to do.

This is a dramatic shift to your business. The companies are smart about it, and our top customers are scarily smart about this. They will have some advantage over others by using methods they built in kind of adapting our stack faster than others. I’m sure they’ll use that advantage.

I’m sure some companies will say, “Oh shit, I’m not prepared. I’m gonna pass on my spending on iOS and be very careful.” That I think is inevitable. Do I think Apple will care or will notice? That’s what I’m wondering. I don’t know.

Shamanth: Apple will notice if there’s a big drop in the install volume on the App Store, which is a possibility I’m prepared for, I think it is within the realm of possibilities. Right now, I know I can pay $40 CPI for high value cohorts or users. Once I know I can’t, I will cut my spend. I cut my spend, my install volume drops. If 2000 people do that, no matter how smart or savvy they are, then Apple’s download velocity across the App Store drops. I think the question is, how much does Apple care and how big does that impact have to be for Apple to care?

Gadi: Others trying to punish Apple by taking their money out of Apple Search ads, and I wonder if they even care because I feel like Apple always treated their ads business like a stepchild. It’s a different building in a sad corner. I’m wondering if they even care about the ad revenue or not. I just don’t know.

Shamanth: Out of curiosity, you did mention some of your customers are preparing and are some really smart people. I’m curious, what are some of the things they’re doing especially considering that there’s so many question marks around all of this.

Gadi: Here’s the thing. Going back to your maybe your first question in our interview, which was what are the different options? I think there’s people who are being very pragmatic and preparing for the worst. When we released SKAN, we said the same thing, which is we’re not in love with SKAdNetwork, but we may be stuck with it for a very long time. So the responsible thing would be to get ready. That’s what they’re doing. They’re doing a responsible thing.

They also have a lot of other things they have to prepare for, which is outside the realm of just SKAdnetwork. They’re figuring out, do I even need to show the pop up? What kind of product changes do I need to do if I want to show the pop up? What is Facebook going to do? Because maybe if Facebook doesn’t show the pop up, no one shows the pop up because it’s not cool. But maybe Facebook does it, everybody does it. So there’s a lot of product implications related to this as well. Can we incentivize users to show the pop up?

Here’s another idea. This isn’t even my idea. It’s a really smart guy that shared an idea, which is what if we only show the popup to the people we care about? The whales. You’re eager, you’re in the game, you just made a purchase and are like, “Okay, yeah, I made a purchase. I don’t mind being tracked. I trust this brand. I just gave them money.” So, if you only get consent for whales, that’s maybe okay. The payers are the only people you care about really in the game — maybe you don’t care about other people. So that’s a really cool idea. I think that that’s some of the ways people are preparing for this. This is probably what you should do.

Jayne from EA is releasing a blog post soon. It should get listed on our blog. She describes from her angle, which is really interesting, way more interesting than what everybody else has to say that’s not an advertiser. She actually needs to make all these decisions for EA. So check out her blog, it’s gonna be really interesting.

Shamanth: I will look out for the blog post. And Gadi, this is perhaps a good place for us to wrap. There are certainly very many questions, but I think we are beginning to make sense of what the world could look like. Thank you again for sharing here, but also offline, because I’ve certainly learned a lot from everything that you and Eran on your team has shared. We will of course link to Singular and your LinkedIn in the show notes, and we’ll have a full transcript and notes as with every episode. Any closing thoughts as we wrap?

Gadi: The one thing I would recommend to everybody listening is we have this Slack group that is formed with probably 800-900 companies by now. It’s called Mobile Attribution Privacy. There’s extremely fascinating conversations with all the people from the companies you’d care about. It’s about IDFA, SKAdNetwork, and limitations. There’s a lot of advertisers, and other MMPs, and different vendors in there, and it’s open for everybody. So I would recommend you guys join. I found it to be an amazing resource. There’s a website called joinmap.org. You go there, there’s a link. It’s a pretty ugly website, I didn’t update it in a while. You can also share with your colleagues. If you’re not there yet, I would highly recommend it because it’s a good resource. You could also search it and see what people answered. There’s a lot of interesting discussions.

Shamanth: Excellent, Gadi. We will link to all of that. Of course, we’ll have show notes and the full transcript as soon as we go live. Thank you so much for being on the show today.

Gadi: Thanks so much.

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!