ABOUT: LinkedIn  | Twitter

Thomas’ course on Mobile Dev Memo Academy: Apple Search Ads: beyond the basics

KEY HIGHLIGHTS

🤔 Why there can be a 30 to 70% difference between Apple Search and MMPs.

👀 What LAT is – and how these users are marked by Apple and Google. Why LAT installs don’t get seen in MMPs.

🚧 How different channels handle LAT. Why LAT can sometimes mean that a significant portion of your audience is unreachable on Facebook and Instagram.

⛰ Geowise, OS-wise and vertical wise differences in LAT-on percentages.

🔞 Apple turns on LAT by default for anyone under 18.

🤷‍♂️ How a marketer might find out what percentage of users have LAT turned on.

💸 Why avoiding targeting users with LAT on can be more expensive than targeting them.

👍 How Thomas recommends bidding for LAT on users vs. all users.

📈 How Thomas recommends estimating LTV/retention/monetization of LAT on users in the absence of perfect tracking.

🙅‍♂️ Why attributing organic revenue to LAT on users can potentially open up a Pandora’s box.

KEY QUOTES

What Limit Ad Tracking means

We call them LAT, which are the letters that are the acronym for Limit Ad Tracking. 

That’s an option that both Google and Apple have enabled for users to do for many, many years. But it wasn’t really communicated broadly, where basically, you would have to go to the settings yourself and activate the option, which leads to your advertising number or IDFA to be wiped out to zero. So sometimes we would call the LAT cohort the IDFA zero cohort – it is pretty much the same thing.

LAT and Facebook don’t mix

There is a very little known fact is absolutely critical: Facebook has decided to never show mobile app ads to LAT users. And because the LAT cohort is say approximately 20-30% of the ecosystem that means a huge amount of your potential audience is never going to see ads on Facebook and Instagram.

The conundrum of LAT in Germany

One country is a complete mystery to me: Germany. 

It seems that, culturally, they are more privacy conscious. They talk a lot about how they don’t want to be tracked by Facebook, and the laws are stricter still. It’s a big topic of conversation. But then you look at the LAT rates in that country, and they’re much smaller than the rest of the European countries, which I can’t explain. 

Assessing how LAT affects apps in different verticals

Another one is linked to the vertical. I often give this example—because I think it’s the most extreme—of VPN apps. These are people who are a lot more likely to be conscious about privacy. And in that particular vertical, you can see insane LAT rates; moving from the 20% average and getting to 50-60-70%. So if you’re only advertising your VPN app on Facebook and programmatic, you’re in big trouble because your audience is very limited.

Teen cohorts and LAT

The cohort of teenagers between 13 and 17: 100% percent of them are labeled in Apple Search Ads with LAT on, because Apple doesn’t think that advertisers should use this data, but they still allow other networks to see it. 

Are Android users becoming less privacy conscious?

There is a very bizarre trend on Android: that it’s moving the other way and the rate has gone down. Now it’s very much of an iOS-specific problem. Well, firstly as Apple Search is only on iOS, but with the LAT, in general, the Android rates would be around 2-3%.

I’ve got a bunch of theories, but I don’t have a lot of evidence to show why this is. And it’s very weird because a few years ago, Android used to be like iOS, at 15-20%. And it’s gone down to 3%, which is very bizarre, to say the least.

How LAT affects bids

As soon as you said you wanted users to be over 18 years old, Apple would stop showing the ads to LAT on users. The problem is a lot of advertisers have done that. And that put a lot of pressure on the LAT off users eventually, which is a lot more expensive. Obviously if your half of the market is only bidding on that pocket of inventory, that pocket of inventory becomes more expensive.

Segmentation based on LAT

What I do now is I have three groups: I have new LAT off, returning LAT off and what I call the open group that has a bit of everything – new, returning LAT on, LAT off, that one is the most shaky one but I have really good results in these three ad-groups as a part of my strategy.

How to use organic metrics to understand Apple Search metrics

The other thing I invite you to do is actually start splitting your organics into organic LAT off and organic LAT on, on your internal systems, because you want to understand if the behavior differs.

So you don’t understand the Search Ads user because organic comes as a whole and it’s a lot of different behaviors: maybe it’s featuring, maybe it’s search, maybe brand search, maybe it’s virality. It all contributes to this bucket, but it gives you a rough idea, and secondly what would be the quality of LAT on.

Then you could measure the behavior of this cohort of organic and against search ads.

FULL TRANSCRIPT BELOW:

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

Thomas: Thanks for having me. It’s been a long time. We wanted to have this chat together. Very happy to be here.

Shamanth: Indeed, indeed. It’s been a while since we’ve talked. And the world has changed a lot since we’ve been talking. I also have been very keen on having you because I’ve certainly followed your Twitter for a long, long time. For our listeners, that’s @Thomasbcn, we’ll link to that definitely. Learnt so much from your Twitter, learnt so much from your writings, from everything you have posted online. 

For all of those reasons, I’m excited to have you Thomas as we’re going to dive into an area that is very, very much your specialty, among many other things. Actually there are very many things you’re very, very good at. But I think one particular area that you’re very particularly experienced in this, Apple Search. And we’re going to talk about a very tricky dimension of Apple Search, which is really measuring it accurately. 

Because for the layperson, it might look like there’s just no way of getting truly accurate data. So can you talk about what are some of the key reasons why Apple Search reported metrics can differ so massively from what an MMP does report?

Thomas: Yeah because it always does and it’s a bit of a nightmare, to be honest. I’m looking into this channel in particular, like, it’s a personal interest of mine – but also because it is very different from other channels in terms of its size, in terms of how you manage it, but especially in terms of measurement and maybe that’s why I want to cover a little bit more detail into that.

So in terms of measurement, there are always discrepancies between your sources of data, whether you’re looking in the Facebook interface, MMP, your own BI, even when data is linked, it’s never fully matching. And I think that marketers have accepted these sorts of small differences, because there’s no way to fully solve them. But we’re used to acceptable discrepancies, like maybe, your Facebook account is going to differ by 5% or like I tend to say, if it’s under 5% at attribution level, it’s probably acceptable. It could be slightly more depending on your settings, window and so on.

But typically on such as the difference would be around 50% and it can vary quite a lot,  let’s say, between 30% and 70% difference, which so we’re talking about – one is the double than the other, so that’s the reason why it is a nightmare. 

The reason it’s happening is there are multiple factors here at play. I’ve published a video on Appsflyer blog that details a little bit more those reasons. And at some points, Apple has actually communicated about it. They’ve been staying a bit silent for the last couple of years. So, the first one you need to understand is that the definitions of Apple and your MMP are different. What Apple calls an install is when the package is installed on the device, but there’s a bunch of users who actually are never going to launch it and the MMP can only see it after the first open when we pull in an install and MMP is actually the first open. I don’t think the difference is massive, but it can lead to at least 5% difference. And I think that’s a bigger case if your app is very big, like if it’s a one gigabyte app, it’s more likely that it’s going to take a while to get on the device and so that can happen. I don’t think it’s a major reason for discrepancy, but it does exist. 

And re-downloads are dealt very differently between Apple and MMP, especially since Apple calls re-downloads a bunch of different things. I cant talk more about that because it’s a bit lengthy. And in the early days of Search Ads, I could confirm from various sources that there was a problem with the API between Apple and MMPs, there was a latency that made a few installs escape the connection there, I don’t know if that problem has been solved. It’s obviously pretty complicated to get Apple to recognize that their tech system is working the way they’re supposed to, but I knew it used to be a problem of up to like 5- 8%. I want to believe that they solve it in a large way, but I don’t know.

And then we’ve got the LAT people, the Limit Ad Tracking people, which among all these factors are a pretty big one, I’d say probably the biggest vis-a-vis redownload factor. So all these small factors of 2,3,5% they add up, of course, but the two big ones, they’re re-attribution and these LAT people.

Shamanth: Yeah. And as you said, the difference can be as huge as 50%. For a lot of channels, if it’s like 5-10%, you can just say, well take an adjusted MMP number but if it’s 50%, it can be enormous. And re-downloads and LAT, you said are the biggest sources of this difference – and re-downloads are understandable. Can you talk about LAT? Yeah, give us more context about it.

Thomas: Yeah, that’s a core reason for the difference. I mean, if you solve the LAT problem, it is not going to solve the whole discrepancy, but it’s definitely a very big one. And I say 50% is an average, but I’m seeing accounts where the difference is actually 70-80%. So, Apple will declare 100 installs & I’ve got 20 installs in my MMP. Oh, we’ve got a problem here. Yeah, so the LAT people, maybe I should say, what they are, since people are not too aware.

We call them LAT, which are the letters that are the acronym for Limit Ad Tracking. 

That’s an option that both Google and Apple have enabled for users to do for many, many years. But it wasn’t really communicated broadly, where basically, you would have to go yourself to the settings and activate this option, which leads to your advertising number or IDFA to be wiped out to zero. So sometimes we would call the LAT cohort the IDFA zero cohort – it is pretty much the same thing.

So we don’t – advertisers and MMP can’t rely on this advertising number to tie campaigns to results. Users have to take that step by themselves – like it’s never by default that this option is activated.

Obviously this behavior can change a little bit like and maybe we will get on to that. But yeah, this is challenging for Search Ads, because Apple always reports LAT on installs in its interface and you never see them in your MMP. They actually come to your app, but they’re marked as organic, like Apple is not passing the data at all to your MMP. So you can assume that they’re all coming in the organic, you can also extrapolate some data to verify if it’s true, or at least partially true, entirely true. And this is very specific. I mean, LAT users exist by themselves, not on channels. The thing is different channels have decided to deal with this problem differently.

If we take SDK networks, such as AppLovin or Chartboost, or I don’t want to name them, then I’d have a long list but you see right? They do show ads to this user, but the MMP manages to attribute most of them through other techniques. Not the IDFA is zero, you probably have a normal cohort of LAT attributed to that channel, it’s not really a problem in this case. 

There is a very little known fact that for me is absolutely critical to know – it is that Facebook has decided to never show mobile app ads to LAT users. And because the LAT cohort is say 20-30% approximately in the ecosystem, that means that a huge amount of your potential audience is never going to see the ads on Facebook and Instagram.

It’s not that these people don’t get ads – like some people actually on the user side, not advertiser side, a lot of people have wrong assumptions about things like this, “oh! if I activate this, I’m not going to get ads anymore.” 

No, it’s just you’re not going to get relevant ads targeted at you, but you’re still going to get a lot of ads. And in the case of Facebook, in particular, what happens is Facebook is showing you web ads on mobile, right? So you would have ecommerce, DTC, any kind of lead stuff, lead generation stuff. So you still have the ads, but you never see games and apps, advertised to you if you have the option on. 

And the reason I’m saying it is for one, it’s not a problem for the advertiser. Because in terms of measurement and tracking, just don’t have them, but I think they should know it, because you probably want to reach that cohort elsewhere. Maybe on Apple Search Ads, such as maybe on an SDK network, maybe you want to run web campaigns to another landing page, and then to your app just to hold back. I mean, if you’re advertising on Facebook and Instagram, which happens quite often for early stage advertisers, you’re missing out on 20-30% of your potential audience, which is absolutely massive. Really, really big.

Shamanth: Yeah. No, that is a huge number, as you said, and I didn’t realize they were not targetable at all, not targeted by Facebook at all. And of course, they are targeted on Apple, but as you said, they just show up as organic. So it sounds like the best ways to target these LAT users is to go via some of the SDK networks that use presumably fingerprinting to identify these users or via Apple.

Thomas : I would still add that the web is actually a third very good arm there. So why I mention the SDK networks specifically is because programmatic is heavily reliant on the IDFA. So typically you’d see almost zero percent LAT if you’re going on an open programmatic network – and in an SDK network you dont. So that’s a big difference between networks. And so you’ve got Apple, you’ve got the SDK networks. And then you’ve got everything you can do on the web, like, I’m working with a lot of advertisers that are going back to the web Google Ads, keywords that you would have them there. I work with a bunch of apps: we’re heavily advertising on Outbrain and Taboola, so you also have the 20-30% cohort there so that the web would be the third one to complete your list.

Shamanth: So these would be, you would direct users to a web page, and then those would be directed to the App Store.

Thomas : Absolutely, in some cases, you can actually hack your way through and make the platform believe that they’re going to the web page, but the user is multi-redirected and ends onto the Appstore. So it’s usually not authorized and it doesn’t work really well, for a bunch of reasons when you hack your way through that. So yeah, basically through landing pages, which is typically something you’d see on native content ads like Outbrain, Taboola, where there’s a piece of content, an extra step in the funnel between the ad and the store.

Shamanth: Yeah, that all makes sense. And you said, typically, it’s 20-30%, which is huge. But you did say that there are apps where there it’s like 70% on, and you did express that look, you just had to manually go in and turn it on. So are there users for whom it is off by default, what’s the reason that the range is so huge?

Thomas: Yeah, there’s a bunch of different reasons. So first, I’m going to be very specific so I don’t get called for bullshit is I didn’t say there’s 70 LAT sometimes. I say 70% difference between Apple and the MMP. I can’t remember the vertical where the LAT was really 70% across the whole board. 

So those LAT range, that would vary a lot – first, because it depends on where you advertise your app. Like if you have very little organic activity, and you heavily use Facebook and programmatic, then it’s very likely your LAT is going to be extremely low, because you’re just not reaching them ever. Then you’ve got ads that rely a lot more on because they have a past existence on the web or so on. So first is a little bit like the context of your app. I don’t think that’s the biggest one. Two massive one – one is geographical and I think that’s related to culture, but also maturity about using your phone. And typically the US would be one of the markets, where the range is among the highest. 

I talked to Apple recently, they’re not sharing it publicly, but they told me that in the last quarter in the US, the average rate was 30%. Well, you can go to your MMP or analytics and check – what’s the rate? If you’re seeing that you’ve got 10,15 or 20% in the US, you’re probably not reaching a bunch of the audience – like in the US 30%. I’d say on the worldwide level, it’s probably around 20%. Singular published a massive data set recently, I can’t remember what was the exact number, but I think it was 20-25%. 

I would have to check. And then it depends on the country like for example, Russia and Brazil have extremely low LAT rates. I don’t know if culturally they don’t care about privacy tracking, or I don’t know the reason. Both Singular and Apple have confirmed with me that Russia & Brazil are complete outliers.

There’s one country, there is a complete mystery to be, which is Germany. 

I mean, I lived in Germany for a couple of years. It seems that culturally there are more privacy conscious aware. Like they talk a lot about, they don’t want to be tracked by Facebook and the laws are stricter, still, like, it’s a big topic of conversation. But then you look at LAT rates in that country, and they’re much smaller than the rest of European countries, which I can’t explain. I just have zero hypothesis for that.

But it’s an interesting thing to observe. You know, as always, what people say vs what they do, like, there’s often a gap here – so the country’s a huge one. Again, I think this Singular study, it discloses a bunch of other countries.

Another one is linked to vertical. And I often give this example because I think it’s the most extreme, which are the VPN apps, which of course, there are people who are a lot more likely to be conscious about privacy. And in that particular vertical, you can see insane LAT rates, like moving from the 20% average or getting to 50-60-70% is really insane. So if you’re only advertising your VPN app on Facebook, and programmatic, you’re in big trouble because your audience is very limited. 

There’s another case that is very particular, but it’s one I work very closely with, which are the kids apps. And that’s specific to Apple. It’s not about the ecosystem itself, like it would be normal. But on Apple Search ads, Apple self categorizes as LAT, even though it’s not activated on the device – anyone who’s under 18. So Apple decided that they’re not going to show ads to people under 13, for good reason.

But

The cohort of teenagers between 13 and 17, 100% percent of them get labeled in Apple Search Ads as a LAT on, because Apple doesn’t think that advertisers should use this data, but they still allow other networks to see it. 

And what that means is if you operate an app that is a particularly big area for teenagers, or under age in general, some social networks have a higher rate because of that – on Search Ads again, and this does not affect your necessarily organic and other cohorts. But on Search Ads, in particular, if you’ve got a very young audience, it’s definitely more particular.

One of the apps I work with on Search Ads, specifically for kids, it’s very common to see a LAT rate of 50-60%. But I believe it’s only 20-30% of people who activate the option and then we’ve got a maybe 30% cohort of Apple, detects them for us and blocks tracking there. So vertical, countries, and it also varies over time and it’s going up so roughly we can say that the group is growing by about 50% in the last two years, moving from 20 to 30%.

There is a very bizarre thing on Android – that it’s moving the other way and the rate has gone down. No, that’s very much of an iOS specific problem. Well, firstly as Apple such is only on iOS, but in the LAT, in general, the Android rates would be around 2-3%.

I’ve got a bunch of theories, but I don’t have a lot of evidence to show why this is. And it’s very weird because a few years ago, Android used to be like iOS like, 15-20%. And it’s gone down to like 3%, which is very bizarre, to say the least.

Shamanth: Interesting, what’s your strongest hypothesis about why Android is different?

Thomas: Yeah, I’m not sure I want to go on the record with that one, because I think Google has a dark pattern to force you out, even if you put the setting. Until I have evidence, I’ll leave it there for people to go and investigate. So I’ll keep you posted.

Shamanth: We will keep an eye on your Twitter and as soon as it’s up, he will have your back for this second episode. So yeah, I know you spoke about percentages of LAT users. How might a marketer find out what percentage of users are actually LAT users?

Thomas : Well, I mean, if you never look into it that’s definitely something that – it’s not that you have to be obsessed about it. But first if you operate on Search Ads you should, because it’s a huge gap. The easiest way because you don’t need any tooling is within the Apple Search Ads interface you can see out of 100 installed 30 had the LAT on – because Apple discloses to you. You might have to add a few columns to the download report by default, but there, what I do recommend is you actually check up either in your MMP, they do provide them or directly in your analytics because you can isolate it, by saying show me the cohort of IDFA zero. 

So to understand a little bit the channels but also to understand like, as I said, before, were you actually missing out on that cohort, because Facebook had such an important share that maybe you pushed very hard there? And you might want to consider if your rate of LAT is extremely low, and it’s not related to your vertical, it’s probably there are some actions that you can think about, which are usually a little less percentage of that – there are other ways to do that. So definitely have a check on it. Don’t obsess with it. But it’s good to know in general, and I think it’s a must know for Search Ads in particular.

Shamanth: Certainly, certainly. So what you’re saying is just because the IDFA comes as zero, that is tracked at MMPs, that should be visible within the MMPs.

Thomas: Yeah, I mean, if you export the raw data, you would see them because the IDFA would be a field. I can’t remember how it’s displayed on the Appsflyer dashboard. I know for sure that on the Adjust dashboard, there is a column that shows the LAT rate there, but it doesn’t even stop at the MMP. I mean, if you’re looking at your own BI or analytical channel, you can isolate this cohort because you just go and say, okay, show me the cohort of IDFA zero versus anything that’s not zero. And you’d get down there to whether it’s in Amplitude, Mixpanel or your own BI. You just have to surface that field, but the flag is there. I’m not sure about Firebase there, but pretty much every tool should be able to surface that.

Shamanth: That makes a lot of sense. Now that LAT is real, how do you adjust your bidding strategy to account for LAT?

Thomas: So here we’re talking specifically with the context of Search Ads because elsewhere, there’s not much you can do anyway about it, except being conscious about it. So, on Search Ads, most of the people at first they were like, whatever and then they realized that there was this discrepancy. So the first tactic was to actually say, “oh, these people are not tracked. I don’t want them because they’re not going to count on my result. So I’m going to go with users without the LAT on” – on Apple you do that by selecting age criteria. 

So as soon as you say I want a user over 18 years old, Apple would stop showing the ads to LAT on. The problem is a lot of advertisers have done that. And that put a lot of pressure on the LAT off eventually, which is a lot more expensive.

Obviously if your half of the market is only bidding on that pocket of inventory, that pocket of inventory becomes more expensive. I do have internal data from Apple, I’m not supposed to share that shows that the LAT on cohort is a third cheaper to acquire on Search Ads, it’s very tricky because you can’t track it directly in MMP, but it’s also a lot cheaper, I think it would be a shame to miss out on it.

In games, it’s even cheaper than a third. I think that was around 40%. I can’t remember the exact geography scope here. Let’s say a third cheaper. So I’ve got a bit more detail on where I’m running a workshop on how to deal with LAT nightmares on the MDM Academy in a couple of weeks. And so, basically, there’s a strategy of, “Oh, I don’t want these guys,” which has to be conservative. 

There’s the strategy of, “I’ll take them all and make some extrapolation,” which is what I’ve been doing in the early days, and I know a bunch of people doing it like this. I don’t particularly like this extrapolation, but at least you’re targeting them and you’re getting them and I’m going to do more analysis, about how to extrapolate this from Search Ads from organic and so on. There’s what I do recommend now and I’ve practiced for, it’s been a long time now, but whatever, is to actually have those ad groups like targeting LAT off only and then targeting everybody, you can’t target them specifically. 

So you would have to have – because there is no overlap issues on Search Ads as like we have on Facebook and other networks, basically multiply your ad-group, your keyword or whatever by two and have a bid for the LAT off because they are more trackable, and more expensive – and have another big level for the other one. So you do have to run a bunch of extrapolation on this, but so maybe in a fictitious case, I’d be up to $1 per click for the general population and I’d be to $2 for the LAT off. This has an extra benefit which is on the cheaper group, it’s not 100% LAT on. So you’re also getting a bunch of LAT off for cheaper than you would have gotten them, if you only target them in LAT off. It’s also a way to expand your audience, and to actually lower down the prices by splitting in this way. It makes the structure a bit more complex like this a bit more burdened structure, but in the era of automation I dont think that is a problem for people.

Shamanth: Yeah, that is that’s very smart to essentially have a two pronged strategy where you’re targeting both and like you said, there’s no downside literally.

Thomas : There is a little bit of downside because there’s noise in the LAT on cohort. So you’re never entirely sure, and you have to run extrapolation about the performance of that group. But I still think it’s a lot more upside than downside and I’m gonna even add something I really recommend to actually apply this to also new users versus returning users. 

So for a long time, I had for every single keyword, I would have 4 ad-groups, like, new LAT on, new LAT off, returning LAT on, returning LAT off – I realized you only need three ad groups because in your LAT on, Apple says they don’t want to know if this user is a new or returning users and they target both, even if you put new users only right, which there is a contradiction in here because they do report if they’re re-download or not. They can’t tell me at the same time that they know because they’re reporting it, but that they don’t know they don’t want to anyway. It’s sort of an internal Apple thing there.

But basically, what I do now is I have three groups. I have new LAT off, returning LAT off and what I call the open group that has a bit of everything – new, returning LAT on, LAT off, that one is the most shaky one but I have really good results in these three ad-groups as a part of my strategy.

Shamanth: Right, that I think is very granular and very sophisticated, that makes a lot of sense. For clarifying the actual tactical execution – to target LAT off, LAT on users, all you do is target 18 plus, is that, right?

Thomas : Yep. So if you at first thought that any kind of demographic criteria would change and switch to LAT off. I’ve come to realize that location doesn’t exclude LAT at all. I think gender does it. I don’t use gender a lot. But like, let’s say you’ve got an app that is very gender specific. If you always target only women, or only men, then you will never get the LAT off either so it’s either age or gender. It doesn’t apply to location. A lot of the apps were looking at ad return on ad spend anyway, and are interested to target 18 plus because usually, they convert better too. I This might not be true for ads(monetization) actually. Anyway, so the structure would be, you have one group where you leave all kinds of users, which is the default option on Apple. So if you don’t touch anything, you have what I call the open group. And then the two other groups, you would select, I want 18 plus user new and 18 plus user returning.

Shamanth: Right and new and returning, you can segment at the setup.

Thomas: So at the ad group level, so it’s not something you can do at the keyword or campaign level, but ad group level when you go and when you create it, it’s there. But if you go and edit the ad group, you have a type of users, so that would be new and returning. There’s also one that’s called a user of your other apps, if you’ve got a portfolio, and then you’ve got the demographic that is gender, age and location. And, all of these happen at the ad group level. That’s why before I would say what I do is I create three ad groups with the same keywords inside because of the difference in how the settings at the ad group level are done.

Shamanth: Right, and I can see how that will let you go after LAT on users. So you’re not excluding them from your targeting. And they get cheaper that makes a lot of sense. But they are still not coming into MMP, they’re still not showing up. They still show up as organic in your MMP. How do you think about estimating their LTV or monetization or retention profile of these LAT users, if at all?

Thomas: Here, I think there are like two main ways of doing it. The first one, which is sort of the easy one that marketers are inclined to do. I’m promoting it myself and other people too, which is basically extrapolating based on the LAT numbers. 

So let’s say you know that the tracked cohort has generated a $100 and that there was 20% of LAT users in this mix, you would add this 20% and say, oh, it’s probably, I’m gonna take $20 out of the organic cohort and attribute it to my Search Ads cohort like the LAT on. And this methodology is okay. And it’s very easy. But I think it’s a little bit shaky in the sense that one, I’ve seen cases where it looked like that was wrong. There may be edge cases and all but I became a bit suspicious of my own method on some edge case. 

And I was like, then taking revenue from the organic is not a great practice, because then you open a Pandora’s box for marketers where everybody is going to go and pick their revenue from the organic bucket, which can lead to some bizarre situations like the Uber case that was a disastrous recipe. And so it still is still a decent one to have an idea. So I do use it as an input, but I don’t report on it, like what I would do is I factor it and typically when I manage a whole team with people with different channels, I would tell them, I’m only going to look at the MMP result. But I’m going to assign you a lower ROAS goal than another channel, because I know there is an effect. And maybe it’s not the full 20%, but I’m going to give you 10-15% leverage. 

One of the things here is one of the problems of this method is assuming that LAT on has the exact same behavior as LAT off. While that might be the case, and Apple once communicated that they do, they removed this sentence from the documentation, and I’m inclined to think that it’s not true. First because of the under age cohort, which there’s no way the 13 to 17 years old cohort is behaving the same as  the 18+ cohort, I’m not buying it. And then for whatever reason, I believe people go and self opt into that option, they have to have a very specific profile – which I’m not saying they’re spending more or less, it’s really hard to determine. But it’s not hard to determine. It’s just it varies. I’ve seen apps for whom this cohort is the better cohort. And apps for it’s a worse cohort. 

So, the other one I invite you to do is actually start splitting your organics into organics LAT off and organic LAT on, on your internal systems. One – because you want because you can understand if the behavior differs. So you don’t understand if the Search Ads user because organic comes as a whole and it’s a lot of different behavior. Maybe it’s featuring, maybe it’s search, maybe brand search, maybe it’s virality, like it’s all contributing to this bucket, but it gives you a rough idea – the other one, and that would be like sort of on quality. 

So you would measure the behavior of this cohort against organic and against search ads.

I basically have three cohort Search Ads LAT off, organic LAT off, and organic LAT on and I compared them in terms of onboarding, conversion, retention, renewal, LTV and so on. The other one is to actually monitor over time the size of this cohort. 

And typically what I recommended is stop targeting LAT on for a week or two, and look at how your organic two cohorts are evolving and then you can start making extrapolations on how much I’m really not seeing in my MMP. And that’s where you start seeing that it’s not exactly the extrapolation you’re seeing from the dashboard. That’s why I became a little bit more cautious with the first method, and I used the secondary method. I recommend people to use both. It’s a little bit complex. I guess if you operate at a small scale, you can just forego this. If you operate at a very large scale, that would be a mistake not to run this analysis.

Shamanth : Indeed, yeah. And just the fact that the IDFA says zero means there’s some ways to measure it, even if it’s not what Apple is mandating that you do, right?

Thomas: Yeah, of course. So you’ve got a bunch of people in this cohort of organic LAT on. We’ve got people who came from Search Ads and you cant track them. You’ve got people who came just like the other organics. And I want to believe because of how fingerprinting is done, that you also have spillover from the web and SDK network activity, and then you’ve got the normal organic. 

So this cohort is a little bit of a bizarre cohort by itself. It’s definitely not only Search Ads. It’s still a very interesting one to monitor and to be honest, I don’t see a lot of people putting much attention there. The few cases where I’ve seen them do it, we got a lot of very good learning. And even if you don’t learn anything, at least you validate that your assumption of not looking at them was the right one.

Shamanth: Indeed, indeed. And as you said, this LAT bucket is at least 20-30% in many cases. So it’s definitely worth investing time and effort into.

Thomas: Yeah, so it’s 20-30% in the nature, in the ecosystem. For many apps I work with, it’s actually less, but I would be like, somewhere between maybe around 15%. The reasons are, if you use a lot of paid activity, it’s more likely to be less. And then it depends on your geographic scale. If you only work in the US, then it’s likely to be bigger, if you’re very strong in emerging countries it’s likely to be smaller. So usually, if you go today to an MMP tool, and you look at the total of all channels, on iOS only, it’s going to look like probably somewhere between 10 and 25%. Again, that’s iOS only. If you forget to put the iOS filer, you’re gonna go to an analytics tool and say “hey dude, that’s like 6% or 7%. Why bother with that?” Yeah, it’s just you’re mixing a 20% cohort on iOS with a 2% cohort of Android. So definitely split the OS when you’re running this analysis.

Shamanth: Indeed, indeed, Thomas. I think this has been incredibly insightful about a topic that isn’t quite as much spoken about as we would all like it to be talked about.

Thomas: It’s a little bit geeky and nerdy. Yeah, to be honest, I don’t like it because of its own complexity. Although it does pose an intellectual challenge that I do like, I do look at it just because it’s not very debated, because the networks wouldn’t educate you about it, because the platform doesn’t mention it. And because I think it brings light on the paid organic relation in many many different ways.

I mean, UA people shouldn’t focus on the paid cohort, you have to understand the impact overall, what’s the blended impact? And I think understanding the LAT is, is a way to start understanding this relationship a lot more. I’m happy to educate a lot of people around this topic. I’ve given several talks. And we’ve gone in pretty much depth today, and I’m pretty sure we’ll have to keep doing it. There is a lot to be found too. So, I wish I would also be challenged by other people so that I can learn because there’s a lot to discover that there’s very little information about it. And I’m happy to share, if people get confused, they can ping me – we’ll have a chat on one of those Slack groups, Twitter, LinkedIn. I’m happy to chat about this issue because I discover stuff all the time chatting with more developers about it.

Shamanth: Absolutely, absolutely. And we will link to your Twitter. We will link to your upcoming course on the Mobile Dev Demo Academy and we will link to that in the promo for this episode just as well Thomas.

Thomas: Thanks very much, I appreciate it, Shamanth.

Shamanth: Absolutely. It’s been an honor having you. Thank you so much, Thomas.

Thomas: Thanks.

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!