Our guest today is Ashley Black, founder at Candid Consulting. In this episode, we delve into the intricacies of Google UAC strategies on iOS post Apple’s introduction of App Tracking Transparency (ATT).
We love this episode as this is from an ex-Googler who gives us the inside track into the dynamics at play within Google that shaped a lot of the changes that happened post-ATT.
Ashley describes details about Google’s initial response to ATT and talks about modeled conversions. We also uncover the challenges and limitations Google faces with iOS ad inventory, particularly the inability to prompt for ATT in browsers and the significant enhancements brought about by SKAN4, including web-to-app tracking.
This episode has a ton of insight into how you can adapt to a post-ATT world on iOS with UAC – and I’m excited for this.
About Ashley: LinkedIn | CandidConsulting |
ABOUT ROCKETSHIP HQ: Website | LinkedIn | Twitter | YouTube
KEY HIGHLIGHTS
🗒 Google’s cautious approach to iOS changes stems from its complex relationship with Apple.
📈 Large and small advertisers perceive and use Google’s modeled conversions differently.
📍Tracking limitations on iOS, such as the inability to prompt for ATT in browsers, significantly impact Google’s ad inventory.
🔐 The introduction of SKAN4 marked a pivotal improvement in tracking web-to-app conversions and overall ad attribution.
✂️ Google’s strategy post-ATT involved a wait-and-see approach, which was often viewed externally as unhelpful.
🔍 Firebase enhances performance on Android but shows less impact on iOS due to limited access to device ID data.
✏️ Effective iOS campaigns with Google require a robust video strategy due to the specific inventory available on the platform.
FULL TRANSCRIPT BELOWSHAMANTH
I’m very excited to welcome Ashley Black to the Mobile User Acquisition Show. I’m excited to have you, Ashley, because you are going to talk about something that I think a lot of people just almost dismissively think of as a black box.
What’s so interesting to learn from you, and certainly what I did learn when we last spoke, is just understanding some of the nuts and bolts of what is very easy to dismiss as a black box: Google UAC.
You’ve certainly seen it from the inside and the outside. So, for all of those reasons, I’m excited to have you on the show. Welcome.
ASHLEY
Thank you. Excited to chat with you.
SHAMANTH
To kick us off, you’ve seen the UAC product from the inside and the outside. You worked at Google during and just after the introduction of ATT.
For those of us who were outside at the time, myself included, the impression I got was that Google didn’t provide much official guidance immediately after ATT was out. What, according to what you saw, were some of the factors driving this dynamic?
ASHLEY
You’re not alone. This was the feedback we got from everybody, and even internally, right? I led one of the sales teams at Google, working with app developers. I think they felt a lot of lack of support because there was very little communication on how Google was reacting to this and, therefore, what their guidance was to the advertisers when iOS 14 rolled out.
What I really comprehended about Google’s strategy here is that they have a very complex relationship with Apple, maybe more so than the other networks. Google is always a little bit hesitant to be the first mover on anything because it is very cautious about its approach.
They know that they have a target on their back, and they get a lot of scrutiny anytime something goes wrong. And so I think they were just very hesitant. Now, keeping in mind Apple’s instructions on the SKAD network and the rollout and everything, they were complex, and it took a lot of deciphering and understanding what was the best way to implement some of the recommendations or the changes that they were enforcing.
So, Google was just really hesitant to say definitively to advertisers, “This is what we’re going to do.” They were more in the mindset of, “Let’s just sit back and wait and observe.” But I think to outsiders, that was viewed as, you know, “You don’t know or you’re not helpful” because they were sitting quietly for a while.
SHAMANTH
Certainly. That makes a lot of sense. And for what it’s worth, I think a lot of marketers were just waiting and watching to gauge the impact of ATT to see if they should ramp down their budgets and how it impacted their marketing. So, the wait-and-watch makes sense, given that a lot of marketers are doing that, too.
Post ATT, one of Google’s first communications was about using more modeled conversions. I think there was a bar graph-filled post that I still recollect after this was publicly announced, saying, “Hey, here’s how we’re approaching conversions and conversion tracking.” You were working with some of the largest advertisers on Google at the time.
What response did you see to the modeled conversions, and how did that differ based on the size of the advertisers you may have been talking to.
ASHLEY
To put it simply, anybody who was big and spending a lot with Google and also spending a lot with other networks wasn’t going to be using modeled conversions for their primary attribution.
The smaller advertisers, who, let’s just say, were slightly less savvy, therefore, didn’t really understand how to leverage the SKAD network and maybe weren’t running across as many networks, is that you know, I mostly worked with mobile game companies. And so they tend to, they could be running across 8, 10, 12 different networks, right?
Some of the smaller guys might only be running with Google or maybe just running Google and Meta, and in that case, I think that they were fine taking the modeled conversion sort of at face value. Also, keep in mind, like with the SKAD network, most people at the time, and even to this day, had to use an MMP to get all of that data.
And those can be costly. So you did have this big disparity between the bigger advertisers, the ones that generated the most revenue for Google, really not relying on modeled conversions, and then the smaller guys, the long tail advertisers, who were totally fine with accepting it.
The modeled conversions were tough because even internally, it was hard to say definitively, “Hey, this is how they were modeled,” right? Or “These are the factors that went into it.” And then just to see the level of disparity because of some of the SKAN network shortcomings and how they measure things.
We’ll talk about that later. Specifically, regarding Google’s network, the discrepancies were massive. Also, it’s worth calling out in both directions. Everybody assumed Google’s number would be way more inflated than what SKADnetwork.
I saw that a lot, but roughly 10 percent of advertisers saw more conversions on the SKAdnetwork than they did on Google. And so, for us as salespeople, I think it made it really challenging to go to an advertiser and say, “This is the source you should be using,” when we saw these extreme differences.
SHAMANTH
Why do you think there was a delta where SKAN was overreporting versus Google?
ASHLEY
Google, I will say this: they hold themselves to a very high bar in terms of how they modeled out those conversions. And so they have to have a really high level of confidence in order to say yes, this happened.
So, what we saw a lot of times is, you know, if an app was new, right? Like brand new, hadn’t spent with us, and Google didn’t have any historical data on it, a lot of times we would see SKAN overreporting versus on Google because it didn’t have historical data to model off of, also depending on where your traffic was coming from, like let’s say YouTube, where they didn’t have any opportunity to capture IDFAs because YouTube doesn’t prompt for ATT.
Again, you couldn’t use that information to inform the model then. So, for people who were spending 90 percent on YouTube and maybe a new app, Google just didn’t have the level of confidence to say yes, and this conversion happened, whereas with SKAN, it didn’t matter.
SHAMANTH
If I understand correctly, modeled conversions were based on some form of deterministic data, whatever was somewhat legally clear for Google to collect. This could have been IDFAs, but it could also have been first-party data or past historical data – so they could say now we have a reasonable amount of confidence that this conversion did happen.
ASHLEY
Exactly.
SHAMANTH
And, of course, if a lot of the traffic is on YouTube, all of that deterministic data just goes away. Did I understand that correctly?
ASHLEY
Absolutely, yes.
SHAMANTH
Can you say that YouTube is one piece of it? What about the rest, other parts of Google’s inventory? Is there a lot of deterministic data available on those? How would you describe that?
ASHLEY
I always like to explain that there are three primary sources of traffic for Google on iOS.
It’s Search, AdMob, and YouTube. That’s it. In comparison to Android, where you have the Play Store, which is a huge percentage of the traffic, you really only have those three. And the way that I break it down is which of those platforms has the ability to prompt for ATT?
When I was on the inside, I was kind of frustrated that there weren’t more communications from the MMPs or thought leaders or anything because they seemed to overlook the fact that Google’s inventory is just different from so many of the other networks.
I know we had talked about this before. If you take Meta, for example, I don’t know what percentage of their traffic comes from apps, but I have to imagine it’s very high. TikTok is just an app. If you look at Snap, it’s just an app. And then all of the other networks are mostly, you know, their inventories on apps.
Google’s different. If you, you know what’s Google known for? Search, right? Where do most people search? In a browser. You cannot prompt ATT in a browser. And so initially, prior to SKAN4, which we can talk about separately, there was really no way to capture the ATT, the prompt, or do the prompt in a browser.
So, immediately, there is no opportunity for IDFAs to be included in a big chunk of Google’s inventory. Then you have YouTube, which is primarily an app. Google decided not to prompt for ATT in YouTube in the app, and I think most people know. But if you don’t, in order to collect the IDFA, you have to be able to get user consent in both the publishing app and the downloaded app.
Even if somebody did consent in the downloaded app, they had no prompt to consent on YouTube. So, there are no IDFAs there. There are no IDFAs on search, and there are no IDFAs on YouTube. So that really leaves AdMob, and then it’s just a function of whether publishers are prompting.
SHAMANTH
And I think that’s a very important dynamic, as competitor channels like Meta, like you pointed out, because Meta does have an audience network, but that’s a far lower percentage of their inventory as compared to Google’s AdMob, which I think that definitely makes so much sense.
You’re right that not many people know, recognize, or are clear about this. So, thank you for explaining that.
ASHLEY
One of the things I was going to mention is that SKAN4 was released earlier, or maybe mid-year in 2023, but Google didn’t really adopt some of it until October.
Up until October 2023, the SKAd network also wasn’t reporting on web-to-app traffic. So you’re going back to the idea that Google search is all web. I mean, some people are going to say, “Oh, what about the Google search app?”
That’s nothing meaningful. And so you have all of this traffic. I cannot tell you the number of times, especially with not so much with games, games didn’t tend to tap into Google search inventory quite as much, but if you had like an e-commerce app or, in some cases, FinTech or all of those other categories that relied heavily on search.
Up until October, they weren’t seeing any of their search traffic counted in SKAdNetwork. If that’s what the advertiser was using to look at performance, they would say, “Well, Google’s awful.” It wasn’t until that switch that we really started to see all of those conversions funnel through, and it started to make a meaningful difference in terms of how people measured.
SHAMANTH
Tell me about SKAN4 and how SKAN4 changes things on web traffic.
ASHLEY
To be fair, Google wasn’t as impacted by some of the changes with iOS 14 and ATT. I think, like some of the other networks, namely Meta, who we saw, you know, even in their earnings, really be substantially impacted by the changes as a result of it.
Maybe that’s why Google didn’t get quite as much publicity on how their network was really different and how it was impacted by the changes that happened with ATT. But for SKAN4, what most people talked about was that we’re going to get multiple postbacks. And now we have fine grain and coarse grain, and that’s what a lot of people focused on.
And then, sort of buried deep in there, most of the comms that I read or articles were, “Oh, and by the way, there’s web-to-app tracking now.” Well, that was huge. You know, we celebrated that so much at Google. It was really meaningful. That was enormous for us.
Again, a huge percentage of Google’s iOS traffic comes from search. Most of those searches are done in a browser on your mobile device, so for the longest time, those weren’t tracked. SKAN4 is changing that now, keeping in mind that you can only track those conversions from the web to the app in Safari, which is, by and large, the largest browser on iOS devices.
But if you’re somebody like me, who uses Chrome as my primary browser on my iPhone, my conversion still would not be tracked in SKAdNetwork. There are just these really fine details. I think that people don’t always realize how Google is impacted or different from the other networks.
SHAMANTH
Right. As you pointed out, Web to app was an afterthought for the vast majority of people. One reason was that it is Safari only. I certainly remember telling folks, “Oh, it’s Safari only.” But you’re right that Safari is significant on iOS.
If you were getting zero attribution in the past, anything non-zero is a big win. And that’s pretty much what I imagine happened. Would it be fair to say that contributed to better visibility on Google and thereby, performance on Google post-SKAN4?
ASHLEY
I would hope so. I was only at Google for a few more months after that release, but I did see several advertisers, especially the ones who were serving 99 percent of their traffic on search, really after that change because every time we would go into conversations with them, we’d have to remind them, don’t forget that all of your traffic is missing in SKAN because of this, because it’s not like they’re looking.
In most cases, they’re separating out what network they’re delivering to. They’re just looking at their app campaign holistically. So the fact like it wasn’t being called out that they were only being served on search. And so anyway, I did start to see a lot of folks after that change come back and say, “Oh, okay, wow.”
We need to reevaluate our investment strategy now that Google is capturing these conversions. The playing field is a bit leveled, and we need to allocate dollars accordingly.
SHAMANTH
Certainly. Another piece of the puzzle really tends to be Firebase. I’ve heard somewhat anecdotally from Google teams that the presence of Firebase correlates with stronger performance.
So, what have you seen in terms of Firebase’s impact on performance? And how did this differ from Android versus iOS?
ASHLEY
On Android, I think there’s no contesting the fact that bidding on Firebase events and having Firebase will improve performance. So, I have no qualms about saying that to anybody or recommending that they make that switch.
That’s not necessarily the case for iOS. To be honest, in terms of performance, it’s not going to really see any significant uplift, but there’s no harm in having it. So, at that point, I sort of viewed it as a developer’s choice on iOS. The performance isn’t improved much, as some of the data that Firebase is able to gain and, therefore, improve performance on the campaigns has to do with device IDs, which they obviously still have on Android.
But per our conversation, it has not been on iOS up to this point. That’s really why you don’t see the same level of performance boost with Firebase on iOS. However, I think this speaks for Android products that Google rolls out and iOS. Firebase will sort of be the gatekeeper for any of those new features and products.
So, if you want to get into one of them, it is like the on-device measurement, on-device conversion measurement that Google has on iOS. That is only available through Firebase. And so I think as you start to see more of Google’s products roll out on either OS, Firebase is going to be a requirement in order to implement them.
So, at some point, it probably makes sense to add it
SHAMANTH
What’s on-device measurement?
ASHLEY
Google’s way of measuring on-device is for apps that collect email addresses or phone numbers. You do have to have one of those user identifiers or personal PII. It’s their way of mapping those users with some sort of action that happened on Google’s network in a way that is like Apple’s privacy-compliant.
All of the measurements and connections happen on the device. It stays on the device. It doesn’t leave. And they’re able to say, “Okay, do we think that something happened here?” And then, they’re able to model out those conversions with greater accuracy.
SHAMANTH
So, if I understand correctly, it’s on the device. It’s aggregate, or maybe it’s not aggregate right because it’s individualized, but it’s on-device, so it’s private, and therefore, it’s used to inform the modeling of the model conversions in some ways.
ASHLEY
We have seen that it’s not a huge increase in performance, but several case studies that Google has released show that when advertisers have the ability to run ODM and opt for it, they do see a modest increase in performance.
SHAMANTH
To switch gears a little bit, I know we talked a bit about the different kinds of inventory on Google versus other channels. What’s also interesting from what we talked about earlier is that the inventory on iOS tends to be very different from Android. Can you talk to us about that and how it can impact user acquisition strategy?
ASHLEY
I mentioned this in an article that I wrote on sort of helping people navigate their iOS campaigns with Google but if you’re not running video in a substantial way, then you are really limiting the amount of traffic and inventory coverage that you have on iOS. So, again, going back to it, kind of like if you can see my hands here, but if we look at Android, there’s a lot more traffic or a lot more inventory sources that Google can tap into.
You have Search, you have AdMob, you have YouTube, but then you also have Play. And that’s the big differentiator there. Play is incredibly meaningful in terms of the Android inventory that’s available on iOS. You basically have all of those minus Play. So you have to search AdMob and YouTube. Well, it’s the amount of inventory that can be captured that that video on iOS Is a lot more substantial than you would see on Android.
So, I think on Android, maybe you still need to focus on video, but you have more opportunities for image assets for coverage and text assets. Again, think about what you see in the Play Store. On Android, you have AdMob and YouTube, which are predominantly video.
The other thing that I would like to highlight is that there are a lot of differences. I don’t even know how many different types of video placements there are on YouTube. But if you’re just thinking about it, you’ve got bumper ads, which are six-second videos, and you have skippable and non-skippable videos on YouTube. You can do it if you’re familiar with it. When you’re watching YouTube or something, sometimes you get two back-to-back ads. Well, those tend to be a little bit shorter. Maybe they want them to be like 10 to 15 seconds. And if they’re only getting one at a time, maybe that’s a 30-second skippable ad.
There are all these different variations, and then ad mobs have a whole other slew of types of formats, like portrait versus landscape, et cetera. In short, it’s just critical to have a super strong video strategy if you’re going to be running iOS campaigns with Google.
SHAMANTH
Again, that’s not something that’s obvious to many people. I think a lot of folks just think, okay, Android is kind of similar to iOS, just different devices. But you’re so right that there’s just a day and night difference in the kinds of inventory you can expect. And that makes so much sense.
ASHLEY
One other quick note on that. I see a lot of people just copying and pasting what they have over on another network, let’s say Meta, and then just saying, okay, I’m just going to put it over here on Google. And the quality of the inventory is different, too. So, I see a lot of people using UGC or like TikTok-style videos.
Google does have a placement for those. They have YouTube shorts, which is kind of the equivalent. So there’s that opportunity, but only to have that one style of content that may work really, really well on TikTok and Meta. But on Google again, you’re having to remember that.
if I’m about to watch a clip from SNL as an example. And that’s going just to be a YouTube video. It’s not in the short. You know, I’m not scrolling; I’m going in, and I’m choosing to select and watch that video to see UGC content in that format, which might be, but it doesn’t necessarily fit. So, in that case, you might need higher production-quality videos for the YouTube inventory, which you don’t necessarily need to do on the other platforms.
SHAMANTH
That’s such a great point because again, oftentimes, with black box algorithms, it’s easy to just forget about what’s underneath and what some of the pieces are that we can actually see, understand, and contextualize.
You’re right that there are Shorts, but there are also very different formats even within YouTube. I think the same is true even on Meta, which is that if you have an ad for that same ad, it may not perform as well on Instagram feeds or Facebook feeds. I think it’s just so important to keep that context in mind, even if the algorithm is ostensibly a black box.
ASHLEY
Right. Absolutely.
SHAMANTH
Ashley, this has been incredibly instructive. And as I said at the outset, I love and appreciate the way you really deconstructed what’s seen as a monolithic black box and just really unpacked some of the things that are very controllable and also just to understand what’s behind the workings of that.
This has been great. This is perhaps a good place for us to wrap up, but before we do that, can you tell folks how they can find out more about you and everything you do?
ASHLEY
I run an app consultancy and media agency called Candid Consulting. You can find me at CandidConsultingGroup.com, and feel free to reach out to me. I would love to chat about these types of issues that you’re having. We help with those types of issues, whether you just want more information or are interested in offloading some things to my team. So that’s where you can find me.
SHAMANTH
Thank you so much for being on the show.
ASHLEY
Thank you so much.