fbpx

Our guest today is David Philippson, CEO of DataSeat, which offers in-house programmatic capabilities and a custom bidding algorithm to app advertisers. David earlier founded one of the first MMPs, Ad-X Tracking, which they sold to Criteo in 2013. 

We’re excited to host David because he’s seen the evolution of tracking on mobile from the very early days. He’s had a ringside view into the forces that have driven change over the years that have ultimately led us to the impending IDFA deprecation. Today we explore this history – and look at what lies ahead for mobile marketers.






ABOUT: LinkedIn  | DataSeat


ABOUT ROCKETSHIP HQ: Website | LinkedIn  | Twitter | YouTube


KEY HIGHLIGHTS

1️⃣When was the first time there was a need for having some accountability for mobile marketing dollars?

🤷🏻‍♀️How did marketers attribute performance prior to MMPs?

🤔What were the original user acquisition channels?

⚡️What was the dynamic that led to the rise of MMPs?

☝🏽What is fingerprinting?

👀What was the ‘Safari pop’ – and why did this happen?

🌏The need for a universal SDK and a common methodology to attribute and de-dupe.

🙋🏽‍♂️What was UDID?

🌪The period of No Man’s Land — there was neither UDID nor IDFA. 

💫How the No Man’s Land period actually created opportunities for in-house attribution.

🔄Very few people knew how to reset the IDFA.

📲What was the purpose and function of limit ad tracking? 

📍How LAT led to advertisers targeting those users even more aggressively.

⬇️What LAT users can become much less valuable post iOS 14.

❎How nearly all user acquisition and nearly all campaigns are a form of retargeting

⚖️Behavioral vs contextual bidding — what will prevail post IDFA?

🔒What does it take to achieve performance with contextual targeting?

💭Which advertisers will be affected by the changes the most?

💢How will hypercasual be impacted?

KEY QUOTES

Raison d’etre for MMPs

There were then these huge deduplication issues — ad networks counting their own homework or marking their own homework and charging double. That was what was happening in 2011: multiple SDKs. Then, the client and the advertiser were actually often paying twice —  two people for driving the same install. That is the dynamic that actually then led to the rise of the MMPs. 

What is UDID

The UDID was a Unique Device Identifier, so the acronym definition itself is a good description of what it was. It was a persistent ID per iPhone or iPad, per iOS device. Because it was persistent, it could be used for attribution.

There have been lapses in the identifier continuum

Then there was this bit of a No Man’s Land period because it wasn’t that UDID was gone and IDFA was introduced. There was a long period of I think 6 to 12 months where there wasn’t this persistent identifier replacement. 

Why opt-out IDFA didn’t have a huge impact

I think that remains the dynamic with IDFA is that few people knew how to reset it. That was also the same with limited ad tracking — few people obviously or intuitively knew how to opt out with limited ad tracking.

There was a workaround for LAT

Limited ad tracking actually backfired because those users were targeted even more and some were still tracked with fingerprinting. Some of these dynamics have led to Apple’s most recent policy change, so that “do not track” really should mean “do not track and try to target them even more, and then track them with fingerprinting.”

iOS 14 will be seminal

The effects of iOS 14 on app is enormous. It’s significant. It’s huge. It’s affecting nearly everyone. 

Why? Because nearly all user acquisition, nearly all campaigns are a form of retargeting. This is the difference between what was a more web based world and what app is. What our industry has grown up doing is taking device IDs from all advertisers. putting them into a device graph, and bidding UA on what you know about the device. So even UA prospecting in-app is a form of retargeting — you’re targeting someone because you know something about them.

Data in mobile marketing is unprecedented

When I speak to some desktop professionals, ad tech professionals that aren’t familiar with that, they look at me with shock and say, “What? Everyone’s sharing everyone’s data?” Well, yeah, some people put their head in the sand and pretend it hasn’t happened. Some people say, “Yeah, we know it happens. We’re addicted to the performance and the ROAS is great.” 

The future of app advertising

What you actually have to do to train a machine learning algorithm on contextual is you need to run budget over many publishers, 24/7. What you’re looking to do is to identify patterns — am I getting my installs from certain publishers, at certain times a day, at certain times of the day, at certain hours of hours a day, certain days of the week, are they on WiFi — all these many variables within the bidstream. When you start seeing conversions, you’ll then start being able to predict what variables lead to my conversions, and that is contextual targeting. That is the future of app advertising. 

Who will feel the effects of iOS 14 most

A potential challenging sector will be hypercasual where they get the majority of their downloads from very specific targeted advertising then they monetize the majority of their inventory through behavioral demand. So if behavioral demand is how they monetize, that drops off the cliff, the eCPMs will drop and the ability to acquire users as well. 

FULL TRANSCRIPT BELOW

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

David: It’s a pleasure — thanks for inviting me.

Shamanth: We’re going to talk about how we ended up here in this time when we are confronting IDFA deprecation in a few months. The working title I have for this episode has been “A Brief History of Device Identification.” I think one of the things you said was, “there couldn’t be anybody better than me to talk about a brief history of device identification” because you’ve certainly seen how this has played out for the last decade. 

I’m excited to talk to you and dive into how things were close to a decade ago, how things have evolved since then, and how we really got here. To start off, in the early years of the iTunes App Store, how did marketers measure and attribute performance?

David: One of the downsides of this is I’m going to show how old I am because of how long I’ve been doing mobile advertising. You’ve really got to go back to when this started — the first iPhone shipped in 2009. It then took a while for app developers to start being able to see this as a chance. It was really 2011 that we then started first seeing the need — there were marketing dollars going behind apps, and since there’s marketing dollars, you need accountability, which is attribution. So, it was around 2011 that we first started seeing the need for this. At the time, before MMPs had even started, before we even started our MMP, the alternative to app marketers would be to work with multiple ad network SDKs. Again showing how long we’ve been doing this, back then it would have been AdMob, Quattro Wireless ended up selling to Apple, InMobi, and Millennial — those were the common names for user acquisition channels back then. 

The alternative would be that the app developers are going to implement multiple ad network SDKs. Now what happens is that there was significant overhead in implementing multiple SDKs and, of course, that means that all ad networks were counting their own homework in a similar way to SRNs today I might add. You certainly don’t want all ad networks counting their own homework. Of course,

there were then these huge deduplication issues — ad networks counting their own homework or marking their own homework and charging double. That was what was happening in 2011: multiple SDKs. Then, the client and the advertiser were actually often paying twice —  two people for driving the same install. That is the dynamic that actually then led to the rise of the MMPs. 

Arguably AD-X Tracking was one of the first. At a similar time, our now friends at Apppsflyer started, if you remember Mobile App Tracking, which became Tune, and, of course, Kochava; we were all addressing the same problem at the time, which was the need for a universal SDK and a common methodology to attribute and de-dupe. That headline still remains part of the value proposition to the MMPs today, but the world was very different then. An attribution is actually far less accurate than those. There wasn’t much deterministic attribution. 

We’ll talk about UDID in a minute. When we do, I’ll explain why it wasn’t used that widely. Early on, we started researching fingerprinting — it still is a standard today. It depends on your perspective. Probabilistic fingerprinting was matching the device type, operating system type, and IP address on a click. You then see it on a download. If it matches, you said that’s where it came from. Now largely, it works, but there’s lots of false negatives and false positives, so it’s probabilistic.

Some of you that have been in the game long enough may also remember when an iPhone app used to launch, a Safari browser may have opened. We used to call that the ‘Safari pop.’ That actually was a way that we invented a terrible user experience. I’m glad it’s not a standard today, but that was the first deterministic method to track because it would allow us to place a cookie on click. When the app launched, you launched the browser to check the cookie. That was the start of the first kind of deterministic tracking, there’s similar themes today, but the world has changed. 

Shamanth: You briefly mentioned UDID. What was that? What was the ecosystem like when the UDID was present?

David:

The UDID was a Unique Device Identifier, so the acronym definition itself is a good description of what it was. It was a persistent ID per iPhone or iPad, per iOS device. Because it was persistent, it could be used for attribution.

Now what wasn’t happening in the early days was ad networks didn’t always have access to it, so they weren’t always passing on the click. But that began to happen in 2012-2013. It became more common that ad networks that had their own ad server, that kind of affiliate type buys don’t always have access to the UDID back then. 

When ad serving became more common, ad networks were able to pass up the UDID on click. That was another part of the deterministic attribution methods. The problem with it was that it wasn’t resettable, so it was a persistent identifier that a consumer couldn’t say, “I don’t want you to do this” or “I don’t want you to use this to track me across different apps.” It was then removed by Apple.

Then there was this bit of a No Man’s Land period because it wasn’t that UDID was gone and IDFA was introduced. There was a long period of I think 6 to 12 months where there wasn’t this persistent identifier replacement.

There were discussions with the industry as a topic, which was open UDID

Shamanth: Did the IDFA get introduced as a result of pressure by developers? What happened?

David: I believe so. Despite personal relationships you might have with past employees, getting official lines or answers from Apple isn’t very common. So I can’t say definitively it was because of that. Then again going back to the definition, it is an ID for advertisers, so it makes sense that Apple was respecting the need for advertisers to create valuable content for iPhone users. Advertisers would have been articulating the need for this persistent privacy-friendly identifier. It certainly appears that that was the result of empathizing with advertisers.

Shamanth: You did speak about the 6 month period that was a No Man’s Land that there was no identifier. What did advertisers do during that time to measure and attribute?

David: It actually created opportunities for MMPs because in the early days of MMPs, The competition was less between AD-X and AppsFlyer in Kochava. The competition was actually advertisers doing it themselves. That was actually a very common theme — in-house attribution. With UDID, you could understand why a gaming developer or an advertiser could actually try and do some of the matching themselves. So actually that No Man’s Land with no UDID, it then needed a partner that had some expertise around probabilistic matching, around the whole cookie tracking –like I said it wasn’t great. But that period of uncertainty actually created more demand for the MMPs back then because when there’s uncertainty, you’d go to subject matter experts.

Shamanth: From my understanding, you did say UDID was a permanent identifier and IDFA was resettable. I remember from my memories in 2012 -2013, I don’t think a lot of consumers would go in and reset it. So would it be fair to say that IDFA was resettable, but nobody reset it, and therefore it was more or less like a UDID? Or is that incorrect as a characterization?

David:  I think you’re right, but your sample set of the two people that you’re asking is a very narrow sample set. If I go to the average million iPhone users, my mother is one of them, what percentage of them know how to reset or put the limit ad tracking in place — it was very few.

I think that remains the dynamic with IDFA is that few people knew how to reset it. That was also the same with limited ad tracking — few people obviously or intuitively knew how to opt out with limited ad tracking.

Shamanth: Which brings me to my next next question, which is about limit ad tracking. Even though it’s been much more in the news lately in the last couple of years, it’s very old. In my understanding, it’s as old as 2012. Can you tell me about what limit ad tracking looked like in its relatively early days? How did it function back then?

David: You’re testing my granular memory of iOS 6 back in 2012. I’ll speak to more of a high level on it. I think it’s completely reasonable for Apple to have introduced limit app tracking, whether it was 6 years ago or 3 years ago. The intent was to give the consumer the right and the option to say I don’t want to be tracked for the purposes of advertising. So I completely agree with the principle. 

The interesting thing that I know it is now widespread within industry, but it kind of remained a bit hush hush at the time, is that advertisers or smart advertisers in the beginning as limited ad tracking percentage increased – we see it around 16% different in different countries, but 16 to 18% is common in programmatic. That’s a very good sample set. You see billions of ad requests. You see what percentage of limited ad tracking — it’s about 16 to 18%. Then, what we saw advertisers beginning to do was target limited ad tracking users aggressively and specifically. 

Now that was when I went back to the sample set of the two people. But folks who have limit ad tracking as a social demographic group are tech savvy, high net worth individuals. When you consider that, so actually the mechanic of how to set limited ad tracking on, meant that someone has to be familiar enough with their iPhone settings, and actually have the persistence to go look and do it. By definition of your million iPhone users, they’re probably tech savvy and tech savvy are probably high net worth individuals. If you’re an advertiser, that is a good demographic to target ironically, and almost amusingly, perhaps unfortunately, because it may have led to some of the deprecation we’re facing today. 

Limited ad tracking actually backfired because those users were targeted even more and some were still tracked with fingerprinting. Some of these dynamics have led to Apple’s most recent policy change, so that “do not track” really should mean “do not track and try to target them even more, and then track them with fingerprinting.”

Shamanth: Were they targeted even more aggressively just because the limit ad tracking was a FYI feature — like advertisers: FYI this person has turned it off, but we will take no action if you target them. Is that why this dynamic exists?

David: The reason it grew was because people experimented with a hypothesis. The hypothesis was, someone said that it probably earns more money, so they will probably buy more of my coins or buy more of my ecommerce transactions. The hypothesis was tested. They were targeted. They were still able to use fingerprinting to calculate downloads and to analyze performance — and the performance was great. As soon as a marketer sees good performance, that’s where the budgets go. Obviously, word soon spread. Even today in programmatic, many advertisers will have specific line items or campaigns, which are limit ad tracking campaigns, but that will change.

Shamanth: Just to jump on to something you just said, in programmatic there are line items of limited ad tracking. Do you think they could still continue to be targeted in a non targeted way? 

David: With iOS 14, the changes I am expecting will be that a far larger percentage of people will be opted out and will be limit ad tracking users. Going back to the million users, one of which is my mother, which will probably opt out, so that group is far larger. It can no longer be considered tech savvy, high net worth that’s worth targeting because it’s very easy to opt out. If it’s very easy to opt out, far more people opt out; therefore, that social demographic targeting is no longer valid. But in the bidstream, you see a device ID or you can see a hashed out device ID, which is a limited ad tracking device ID. So if you saw a device ID that was 000 that is a LAT device and certain campaign types, certain advertisers would target those specifically for that reason. But if everyone opts out, and everyone’s in that group, it no longer becomes effective.

Shamanth: That’s a good observation. So the LAT category is just going to become less valuable. 

David: It includes everyone, including my non tech savvy relatives. 

Shamanth: As we’ve talked about just now, there’s been a gradual progression from UDID to IDFA to LAT actually being enforced. Of course now with the upcoming IDFA deprecation, this has been a gradual progression over very many years. How does this compare with how privacy policy has changed and evolved on the web?

David: It’s a good question. I have drawn comparisons, and I had specific experience although my kind of expertise is all around app and attribution and programmatic for apps. We sold AD-X Tracking to a business called Criteo that were well known for desktop retargeting at the time. Now, when you look at the impacts of ITP, which is Intelligent Tracking Prevention from Safari, then it became ITP2, and then you saw Chrome following soon. Where that had the most impact, It wasn’t on prospecting, it had impact on behavioral targeting. Prospecting is more just trying to find new users — you don’t know anything about them — it’s more contextual. So ITP in terms of tracking prevention and what happened in Chrome and just the general cookie, kind of controls what has been put in place over recent years. 

The most negative effects on ad tech players and advertisers was around behavioral targeting and retargeting because the impact wasn’t actually huge. Of course, it was a significant change, but the impact wasn’t huge. It was significant to any retargeting web player because they were restricted considerably. The changes that are now happening in iOS with iOS 14, they’re similar to the changes from ITP — the objective is similar. However, my observation is that while the impact on web wasn’t huge, it did affect behavioral targeting and retargeting.

The effects of iOS 14 on app is enormous. It’s significant. It’s huge. It’s affecting nearly everyone. 

Why? Because nearly all user acquisition, nearly all campaigns are a form of retargeting. This is the difference between what was a more web based world and what app is. What our industry has grown up doing is taking device IDs from all advertisers. putting them into a device graph, and bidding UA on what you know about the device. So even UA prospecting in-app is a form of retargeting — you’re targeting someone because you know something about them.

Whether it’s cookies, or iOS IDFAs, the negative effect is on behavioral advertising. Now, the difference with apps is nearly all marketing dollars are going to a form of behavioral targeting whilst that wasn’t the case in desktop. I’m expecting and predicting that the impacts of iOS 14 are far greater on ad tech players in app than ITP was in desktop,

Shamanth: I like that characterization that all of UA is in the sense retargeting. 

David: It’s one huge retargeting campaign.

When I speak to some desktop professionals, ad tech professionals that aren’t familiar with that, they look at me with shock and say, “What? Everyone’s sharing everyone’s data?” Well, yeah, some people put their head in the sand and pretend it hasn’t happened. Some people say, “Yeah, we know it happens. We’re addicted to the performance and the ROAs is great.”

Other people will say, this isn’t sustainable. It’s got a huge cannibalization effect. Apple will probably want to stop it, and it’s probably non-compliant. That’s the route it has gone. 

Shamanth: With that context, it’s understandable why Apple is doing what it is doing. Certainly, it also sounds to me like that as precedent on the web for operating without that kind of deterministic targeting, there are very many web-based businesses that operate at fairly large scale without requiring that sort of exact precise knowledge of the individual users’ behavior. Would you say that that is the way the world will look potentially in a few months post IDFA deprecation?

David: I think that is what the new world will look like. I think the change will be a bit like the S&P or the FTSE with the pandemic – it will drop off the cliff. Because of iOS 14, the opt out is going to happen. I think the transition will be gradual. I think there’s gonna be lots of chaos. It’s gonna be hard to measure ROAS and retention. Everyone’s scrambling to get SKAdNetwork kind of solutions in place. But I think the new normal, if you consider – there’s two types of bidding – behavioral and contextual.

Behavioral bidding will stop because what you know about the device is going to be stopped — your primary key, the device ID, is gone. So you can’t build a record of what you know about the device. What are you left with? You’re left with contextual. In fact, that is what prospecting UA in the desktop world always has been. There’s never been this huge cross pollination of data. So yes, that will be the new normal.

It’s also important that I define what I mean by contextual. What I mean by contextual is that one advertiser’s data, or performance data, will only ever be used for that advertiser to optimize performance, and then you’re going to go wrong. 

But how do you optimize performance? You have to go through a phase of training machine learning algorithms to predict outcome. That is the objective of any machine learning algorithm — it’s to predict an outcome. What the outcome will be is, is there going to be an install?

What you actually have to do to train a machine learning algorithm on contextual is you need to run budget over many publishers, 24/7. What you’re looking to do is to identify patterns — am I getting my installs from certain publishers, at certain times a day, at certain times of the day, at certain hours of hours a day, certain days of the week, are they on WiFi — all these many variables within the bidstream. When you start seeing conversions, you’ll then start being able to predict what variables lead to my conversions, and that is contextual targeting. That is the future of app advertising.

The whole behavioral model, what you know about the device is limited. Of course there are exceptions and outliers to all the statements I made. There could be targeting on Facebook Canvas, for instance. I completely agree there are exceptions, I’m more talking about general principles.

Shamanth: Do you think this is going to impact the smaller advertisers more than the larger ones because they have fewer dollars to sink into exploring and learning and whatnot? 

David: I think the ones that are the most affected are the ones that are most reliant on paid installs. If you look at all advertisers and actually understand the ratio of their installed base in their whole monetization mechanic — how many come from paid versus organic? If you had 90% of traffic from paid installs, the short term transition period where there’s lots of uncertainty, lots of stuff is gonna get broken, and Apple has no urgency to fix it all for us in the short term – is going to be challenging. I think those that will feel the impact the most are the ones that the majority of their users come from paid channels. Then a multiplier effect on those that get affected the most are ones that monetize the most through advertising.

A potential challenging sector will be hypercasual where they get the majority of their downloads from very specific targeted advertising then they monetize the majority of their inventory through behavioral demand. So if behavioral demand is how they monetize, that drops off the cliff, the eCPMs will drop and the ability to acquire users as well.

Again, there are exceptions to everything I say, I completely acknowledge that, but there may be a variety of effects. I have one customer that’s 90% organic. They’re one of the known gaming companies, and they’re a bit more relaxed about this. I have another customer that’s completely the opposite. And they’re like, “Ah, David, what should we be doing?” But I think that the spectrum is paid versus organic and monetization for advertising.

Shamanth: Yeah, we are indeed headed into what could be a perfect storm. David, I think all of this helps us contextualize why we’re getting into what we’re getting into because this has been more than a decade in the making. Thank you for sharing all of that perspective. I think it also helps inform us about what the future could look like and how we could adapt. This has been incredibly instructive — perhaps this is a good place for us to wrap. Before we do that, could you tell folks how they can find out more about you and everything you do? 

David: You gave a kind introduction earlier, but I’m David, I’m the co-founder and CEO of DataSeat. I founded DataSeat, which is an in-house programmatic DSP for app developers. I actually see these changes as an opportunity for in-house. Because if you think about in-house,  if an advertiser was to in-house their programmatic UA, that would never be behavioral retargeting, that’s never based on a device graph, which is why in-house UA always struggled to compete against the status quo of behavioral bidding. Actually a world that’s moving towards contextual, I believe is in favor of in-housing, which is one of the reasons we thought this was gonna happen, which is why we decided to focus on that opportunity. 

But if anyone would like to speak with me, please find me on LinkedIn. Our website is Dataseat.com. We’d love to speak to people that are interested in the future of in-housing and particularly contextual bidding algorithms. 

Shamanth: We’re going to link to your LinkedIn and Dataseat.com in the show notes. Of course, we’ll have transcripts and notes for everybody to reference. For now, David, it’s been a pleasure having you on the show. Thank you so much for being on the Mobile User Acquisition Show.

David: Thanks for inviting me. I’ve enjoyed it.

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!


WANT TO SCALE PROFITABLY IN A POST IDENTIFIER WORLD?

Get our free newsletter. The Mobile User Acquisition Show is a show by practitioners, for practitioners, featuring insights from the bleeding-edge of growth. Our guests are some of the smartest folks we know that are on the hardest problems in growth.