Alex joined Branch as a Developer Advocate in 2016, and helped build the company's early developer relations and long-tail adoption strategies. He works closely with teams across the organization to help shape Branch's place in the mobile ecosystem, writes the Mobile Growth Newsletter, and tweets regularly about mobile-related topics at @alexdbauer.
Jun 01, 2021
By now, the entire mobile industry knows that iOS 14.5 came out a month ago. According to Apple, that means the AppTrackingTransparency policy is now in effect. Time obviously isn’t standing still either: iOS 14.6 was released last week, and the first beta of iOS 14.7 is already available. Add in the barrage of mainstream news headlines proclaiming the end of mobile advertising as we know it, and you’d be forgiven for believing that some fairly dramatic changes are taking place.
Yet, somehow, the music hasn’t stopped. What happened to the long-promised apocalypse?
We’ve been monitoring the data closely over the past few weeks, and there is a clear explanation for this unexpected period of calm. And unless Apple has changed its mind about the whole AppTrackingTransparency initiative, this calm is also not going to last forever.
Here are the four key things you need to know:
1. Adoption of iOS 14.5+ is still really low. That means ATT and its associated technical limitations only apply to a small fraction of iOS users so far (20.6% as of publication).
2. Apple hasn’t started enforcing the policy component of ATT yet. While Apple rejected a few app updates right after the launch of iOS 14.5, almost every example so far has been a case of helping developers align their ATT prompt implementation with the self-reported details on their app’s privacy “nutrition labels.” In other words, Apple has taken care of honest mistakes…but they haven’t yet begun auditing for real malfeasance.
3. The “ATT opt-in rates” we’re seeing don’t all mean the same thing, and are easy to misinterpret. Theoretically, these metrics should indicate how app users feel about “tracking.” In reality, the numbers floating around are so widely divergent (depending on the date and the source, they’ve ranged anywhere from 4% to 60%+) that it’s easy to find data to support almost any spin you want.
4. ATT opt-in rates are vanity metrics that do not reflect the true impact of ATT. Even if these metrics were more consistent, they’re at best a distraction. The only number that really matters is how often attribution is still available…and no one has been sharing that data (spoiler alert: until now — keep reading).
Let’s take a look at each of these to try and untangle the real story.
There are always two phases to the release of a new iOS version:
Apple generally doesn’t move past phase one until they’re really sure that the new release has no issues. In the case of iOS 14.5, there were a few ATT-related bugs in the initial version, and a pretty serious unrelated security issue with WebKit. Apple patched one of the ATT bugs and the security issue with iOS 14.5.1, but still held off pushing users to update from iOS 14.4.2.
Looking at past iOS 14.x releases, we can clearly see that iOS 14.5 and 14.5.1 never moved past the first phase. However, there’s a clear group of updates that took off after about a week, and iOS 14.6 is ripe to hit a similar inflection point any day now.
All of the ATT opt-in numbers floating around right now (which we’ll discuss in the next section) are for only one side: either the advertiser or the publisher app. But in order to use the IDFA for attribution, the user must consent via ATT on both ends.
To get an idea of the true impact of ATT, we need to compare the rate of installs attributed to ads on iOS 14.5+ with the baseline from previous iOS versions. This metric is the real deal — the elephant in the room that no one is talking about yet. And as you’d expect, it’s much lower:
As of writing, the rate of successful attributions is slightly above 6.5% of the pre-ATT baseline. This means that compared to how things used to be, device-level insights are now missing for 14 out of every 15 ad-driven installs.*
However, there are two important caveats to this:
* For the mathematically-inclined, this metric is calculated by taking a sample of apps known to be showing the ATT prompt and running ad campaigns, and then comparing the difference between the attribution rate (attributed installs / all installs) for versions of iOS < 14.5 and > 14.5. For example, if 10 out of every 100 installs are attributed on iOS < 14.5, and only 1 out of every 100 is attributed on iOS > 14.5, the metric would be 10% (indicating that data for the other 90% of ad-driven installs is now unavailable due to ATT).
While the attribution rate discussed above is the most important metric, the number of ATT opt-ins is also useful because it (theoretically) indicates how many users give consent to “tracking” when asked, and what overall end user sentiment is toward ATT.
In reality, there has been a lot of confusion about these numbers, and what they mean.
The next section will take a dive into the various public ATT opt-in metrics, how they’re being calculated, and why this leads to so many discrepancies. But this is a case where looking at the raw data is actually helpful too, so let’s begin with a few basic breakdowns of ATT data we see in Branch traffic:
This graph shows ATT status for users of apps showing the ATT modal (bottom) and apps not showing it (top). As a reminder, these are the definitions for each status, adapted from Apple’s documentation:
Here is the same raw data for apps showing the ATT prompt, broken down by country of traffic origin and top app verticals:
Finally, here is a comparison of how widely implemented the ATT modal is across verticals:
Here’s a quick explainer of top-line numbers getting most of the attention over the last few weeks.
ATT opt-in metric 1: Percentage of opted-in iOS user base.
ATT opt-in metric 2: Percentage of users who select “Allow” when shown the prompt.
ATT opt-in metric 3: Percentage of ad impressions containing an IDFA.
While all of these opt-in metrics are interesting, they don’t tell the whole story for two reasons:
Because of these last two points, any rate calculated with the equation of authorized / (authorized + denied) might help us understand user sentiment toward ATT but doesn’t tell the whole story of how it’s impacting mobile marketing overall.
A more insightful metric — and the one we’ve been recommending to Branch customers who want to understand the true opt-in rate — is a “tracking allowed” rate calculated as authorized / (authorized + denied + restricted + not_determined).
We’ve been looking at this rate from Branch data, and it’s obviously somewhat lower than the other metrics…but still high enough to raise doubts about the media narrative that “all users hate tracking,” and continuing to grow gradually as more users see the ATT prompt over time.
The restricted rate (users who cannot be shown the ATT prompt for various Apple-defined reasons) was artificially high for the first few weeks, presumably due to a bug on Apple’s end. It abruptly on May 18, suggesting that Apple finally fixed this issue:
The device-wide opt-out rate indicates how many users have enabled the device-level ATT toggle. This is calculated by looking at the rate of denied statuses for apps that are not showing the ATT prompt. It’s decreased slightly over the past few weeks, perhaps reflecting the preferences of users who update quickly.
There are a few key inflection points to watch for over the next few weeks:
Ultimately, one thing is clear: obsessing over these data points doesn’t change the reality that Apple has used AppTrackingTransparency to fundamentally shift the narrative around mobile user privacy: IDFA-based, device-level attribution data is now far scarcer. SKAdNetwork will help fill the gap, but going forward, app marketers will need to find new ways to drive growth and measure results. Branch is here to help.
To help you fuel cross-channel and cross-platform mobile growth, our team works hard to deliver the most current, relevant resources.