ATT Opt-In Rates: The Picture So Far And The Ugly Truth Behind Why The Numbers Vary So Widely

Alex Bauer, head of product marketing, Branch

Data-Driven Thinking” is written by members of the media community and contains fresh ideas on the digital revolution in media.

Today’s column is written by Alex Bauer, head of product marketing at Branch.

Have you seen a tweet, blog post or news article in the past few weeks about “ATT opt-in rates,” perhaps accompanied by apocalyptic commentary about how that number proves we’re nearing the end of advertising as we know it?

Or perhaps you’ve come across someone hopefully noting that the rate actually might not be quite as bad as many were expecting.

So … which is it? Well, this is the mobile ecosystem’s newest hot debate.

Now that iOS 14.5 is finally out the door – and Apple’s new AppTrackingTransparency policy along with it – everyone is breathlessly digging for clues about the impact. Will users choose to opt out of “tracking” in overwhelming numbers as many observers have predicted? Or will the reality be more complicated? (Spoiler alert: the reality is, indeed, more complicated.)

Let’s investigate why the numbers you’re seeing vary so widely, how they’re being calculated, and get into ways you can use this data to inform your own decisions.

What is the ATT opt-in rate?

The most obvious metric to tell this story is the percentage of users that select the “Allow” option when they see Apple’s new AppTrackingTransparency prompt in an app:

Screen grab from Apple's promotional video for the AppTrackingTransparency framework

Source: Apple

A very low opt-in rate would mean users are buying into the strong anti-tracking message coming from Apple, while a higher rate would indicate a more nuanced story. For example, a higher opt-in rate could mean that users are evaluating these requests on an app-by-app basis and may be more receptive when developers provide context for why users should consider granting ATT consent.

Almost immediately after Apple’s 2020 Worldwide Developers Conference wrapped, people started making predictions about what the opt-in rate would be. Some of this prognostication has been based on user surveys and some on tests run by individual app developers. But a lot of it was just gut calls.

In the weeks since the release of iOS 14.5, those predictions are finally giving way to real data. However, the range of numbers being shared right now is so wide that the picture is even more confusing.

To be fair, there are many real-world factors that can impact an app’s ATT opt-in rate, including the app’s category, geography of its primary user base and – of course – how the app developer has optimized its presentation of the ATT prompt, just to name a few.

But a major contributor to the current confusion is differences in how these “opt-in” metrics are being calculated. There are several methodologies in use right now, each of which tells us something valuable about the impact of iOS 14.5, but these numbers cannot and should not be compared on an apples-to-apples basis.

And yet they are.

Because of this, many observers are perplexed by the different “opt-in rates” they’re seeing and, unfortunately, that confusion makes it easy to draw incorrect conclusions.

Why do ATT opt-in numbers diverge so widely?

The equation for any ATT opt-in rate will be intuitively simple for anyone with a basic background in algebra: authorized / (authorized + denied). The complications begin when deciding which events qualify for the calculation and in the method used to collect them.

Here’s a breakdown of the various ATT opt-in numbers you’re likely seeing – and what you need to know about what each one signifies.

ATT opt-in metric 1: Percentage of opted-in iOS user base, across all apps.

Worldwide daily opt-in rate after iOS 14.5 launch, according to Flurry

Source: Flurry

Where have you seen it? This is the top-line number being reported by major platforms, including from Branch and the widely cited iOS 14.5 opt-in rate report by Flurry. Because the percentage is quite small, this is also the metric that many privacy advocates – and most mainstream media articles – have trumpeted as overwhelming evidence that users hate “tracking.” 

What is the current range of numbers? Depending on the source and region, this opt-in rate has fluctuated between 4% and 13% thus far.

How is it calculated? This metric looks at the ATT status for active app users on the date in question. This means a user who interacts with 10 different apps on a given day will be included in the calculation 10 times, with each reflecting the user’s ATT status for that app.

What does it mean and why is it useful? This metric reflects what you can think of as the live opt-in status of iOS users, meaning that it stays up to date as users opt in and out of ATT for individual apps over time. It helps to quantify how much AppTrackingTransparency is reducing the overall “tracking capacity” of iOS apps, compared to pre-ATT launch.

What are the gotchas? This metric does not provide useful insight into how users are responding when presented with the ATT modal. Also, while the numbers we’ve seen so far help quantify the overall ecosystem impact, this is likely not what many app advertisers want to understand when they’re looking for an ATT opt-in rate. They’d rather know what is going to happen to measurement for their app, if they decide to show the ATT prompt. 

Because of how Apple’s ATT framework returns data, answering this second question is more challenging and will require some additional analysis work.* During the next few weeks, we’ll likely see improved estimates for this number.

*Why is this a challenge? Users can opt-out of ATT device-wide. If that happens, the ATTrackingManager function returns a “denied” status for that user even when the app hasn’t implemented the ATT prompt at all. Since this denied status is identical to what is returned for users who see the ATT prompt and decide to opt out, there is no simple way to separate the data and the result is a lower reported “opt-in rate.” Fortunately, it’s possible to correct for this by only including data from apps that are also reporting ATT statuses of “authorized,” and we’ll hopefully see refined numbers that incorporate this logic soon.

ATT opt-in metric 2: Percentage of users who select “Allow” when shown the prompt.

Overall cross-category ATT opt-in rates, as per AppsFlyer

Source: AppsFlyer

Where have you seen it? The broadest study of this metric appears to be coming from AppsFlyer, although the sample size (around 550 apps) is still significantly smaller than the platform-wide reports seem with Flurry’s metric. If you’ve heard any anecdotal reports from individual app developers, this is most likely the number that they are referencing.

What is the current range of numbers? Because this metric is often being shared at the individual app level, it literally runs the gamut. Reports are floating around of opt-in rates that range anywhere between 1% and 60%. However, according to AppsFlyer’s data, the global average is around 37%.

How is it calculated? When the “requestTrackingAuthorization()” method is called to display the ATT prompt, a developer can immediately capture the ATT status after users make their selection. (If the user has disabled ATT at the device level, a status of “denied” is returned by default).

What does it mean and why is it useful? This is the number that most closely answers the question of how many users decide not to share their data after seeing the ATT prompt. That makes it useful for measuring user sentiment toward ATT, and the impact of the optimizations developers make to their ATT presentation.

What are the gotchas? Unfortunately, this data is challenging to get at scale, because collecting the data requires some additional instrumentation work at the individual app level. Although it’s a helpful indication of user preferences toward tracking, it does not reflect the full impact of ATT on the app’s measurement capacity.

ATT opt-in metric 3: Percentage of ad impressions containing an IDFA.

Where have you seen it? If you’ve seen opt-in rates that originated from an ad network (for example, in an article like this one from VideoWeek), this is the metric being cited.

What is the current range of numbers? Depending on the ad network and region, the rate appears to range between about 15% and 30%. 

How is it calculated? This is the number of individual ad impressions that contain an IDFA, which is only possible after the user has opted into ATT for the app that is showing the ad. A single app can generate multiple ad impressions for the same user, each of which will be included seperately in the calculation.

What does it mean, and why is it useful? Unlike the first two metrics, which are closely aligned to opt-in rates in advertiser apps, this metric almost exclusively reflects the publisher side of the equation. That makes it useful for apps looking to get an idea of what impact ATT will have on their monetization strategy.

What are the gotchas? Since this metric can reflect opt-in data from multiple ad impressions per user session, it’s not an accurate way to demonstrate how many users select to not share their data when they see an ATT prompt.

But wait, there’s more

As if all of this weren’t complicated enough, there are two additional – and important –considerations to keep in mind with all of the above metrics:

  • There is a cohort of users with an ATT status of “restricted.” This status indicates that a user is not allowed to make ATT selections at all. While Apple lists an array of situations for when the restricted status is expected, including child iCloud accounts, device management profiles or brand-new iCloud accounts, there are also widespread reports of a bug in iOS 14.5 and 14.5.1 that is incorrectly enforcing this status.
  • Users who have not seen the ATT prompt are de facto the same as opted-out. Whether this is due to the app developer choosing not to implement ATT, or simply delaying the presentation of it for optimization purposes, users who have not seen the prompt cannot be “tracked” for as long as this is the case.

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 to help us understand the true opt-in rate would be a “tracking allowed” rate calculated as authorized / (authorized + denied + restricted + not_determined). So far, none of the numbers we’ve seen reflect this bigger picture.

It’s also critical to remember that to be useful for most purposes, ATT requires user opt-in on both sides. As Eric Seufert describes in his article, “ATT opt-in rates are irrelevant,” the full impact of ATT on the iOS ecosystem is more nuanced than any single percentage number is able to show.

So, where do we stand?

Ultimately, any ATT opt-in rate provides a useful point-in-time look at user preferences. Used responsibly, these numbers can serve as a benchmark for advertisers and publishers working through the disruption caused by iOS 14.

But used irresponsibly, especially this early in the adoption cycle, citing any flavor of ATT opt-in rate can provide ammunition for narratives that may not truly exist – and that’s true for both extremes of the user privacy debate.

And yet one thing is clear.

Obsessing over this data 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 and, going forward, app marketers will need to find new ways to drive growth and measure results.

Follow Alex Bauer (@alexdbauer) and AdExchanger (@adexchanger) on Twitter.

Enjoying this content?

Sign up to be an AdExchanger Member today and get unlimited access to articles like this, plus proprietary data and research, conference discounts, on-demand access to event content, and more!

Join Today!


  1. Micah Donahue

    Fantastic analysis, thank you for digging deep on it.

    I have a Q around Facebook specifically. FB has said they’re opting out all users until/unless they opt-in. Have you heard if this auto-opt-out applies to both 1) tracking and 2) excluding those users in reporting results like conversions? Or to only one of those? Thanks!

  2. Hey Micah! Glad you found it helpful 😃.

    With Facebook, it’s not really a case of what *they* can do, because “opting out all users until/unless they opt-in” is just simply the way ATT works. Though in Facebook’s case, I believe they are treating a user as opted-out until that user opts in on BOTH the Facebook and Instagram apps.

    The answer to your question is unfortunately a bit complicated: an opt-out requires Facebook to exclude the user from “tracking,” which also includes conversion reporting the way it’s traditionally been done. However, SKAdNetwork doesn’t require ATT consent, and provides some very limited conversion reporting even when the user hasn’t opted in.

  3. Question, on Flurry’s you state “This means a user who interacts with 10 different apps on a given day will be included in the calculation 10 times, with each reflecting the user’s ATT status for that app.”.
    Yet the Flurry’s graph label states: “Opt-in rate = number of devices that opted-in / (number of devices that opted-in + number of devices that opted-out)”
    Is then app level equal to device level? To me these are not the same…
    Thanks in advance.

  4. Thanks Alex for the thorough research and well written summary.

    At Kayzen we have also provided data on the various ATT statuses. I can confirm, the data we provided is indeed not on a user level, but on an impression level. Hence, if say users who opt-in to tracking see more ad impressions (publisher side) than users opt-out of tracking then the opt-in ratio may look better than it would on a user level.

    The challenge here is to track down the users who opt-out, because you dont have a device ID (anymore). However, you can still track them through IDFV (limitation: you will count users across 2 ore more apps of different developers several times). We havent done that analysis yet because only about half of the ad requests in the market that do not contain an IDFA yet contain an IDFV.

    For everyone who is interested to get more date on this (from us) PM me on linkedin.