You’ve probably heard the chatter in the mobile marketing world lately around the IDFA, a core piece of the tracking puzzle within the iOS ecosystem. And there’s good reason for all the attention — Apple’s upcoming move to increase users’ privacy by setting the IDFA to “off” by default will force major changes in the mobile tracking and attribution spaces.
What’s an IDFA?
The Identifier for Advertisers (IDFA) is a unique identifier assigned to a mobile device, and used to allow advertising companies to track users across the mobile app ecosystem. Basically, it’s the data point that has powered much of the retargeting and app install attribution industries up to this point.
What’s happening with the IDFA?
Apple recently announced that iOS 14 will include a major change for app-based tracking. Specifically, they will be switching the IDFA to be off by default on iOS devices. Previously it was on by default, and users had to manually change their tracking settings to turn it off. The difference between defaulting to on and defaulting to off (aka opt-out vs opt-in) is massive — which means that the IDFA will be effectively dead once iOS 14 is released (estimated to be in September or October of this year).
Who does this affect?
Since the IDFA is related specifically to apps, this change is only going to be relevant to advertisers that are tracking app installs and/or in-app actions. Unlike some of Apple’s previous moves to enhance consumer privacy, this change will have no effect on anyone concerned solely with web-based tracking and attribution.
In addition, the IDFA is only useful when you’re tracking a user from one app to another. Even today, mobile websites do not have access to a user’s IDFA, which means that mobile web-to-app tracking and attribution is not experiencing any loss here — the IDFA has never been of use in that scenario.
All that to say: when we talk about “losing” the IDFA, we are talking about the subset of partnership programs that are tracking installs and in-app conversions, and who see a substantive portion of their traffic coming from other apps.
How will this change affect Impact’s tracking?
Here’s the short version:
We don’t expect this change to significantly affect our ability to track installs and in-app actions, and attribute them to the correct partner. That said, anyone who is handling their mobile tracking through an MMP will want to confirm with them as well.
Here’s the longer, more nuanced version:
Attribution is the process of determining that one event, such as a click, is connected to another event, such as an app install. In order to make that connection, an attribution provider needs some way to link those discrete events together and reliably say “this event caused that event.” There are two main ways of attributing installs and conversions on a mobile device: deterministic matching and probabilistic matching.
Deterministic matching is the more accurate method, but it relies on a consistent identifier like the IDFA to know for sure that a given user a) interacted with a partner or advertisement and then b) installed an app or completed an in-app event. For instance, if IDFA #12345 clicks a tracking link, and then later the same IDFA installs an app, it’s clear that the same user took both actions.
Probabilistic matching, on the other hand, takes a large number of data points into account, runs it all through an algorithm, and makes educated guesses about which click events are linked to which install and conversion events. For instance, if a user at IP address 220.127.116.11 clicks a link to the Candy Crush app store listing at 10:48am EST, and then a user at the same IP address installs the Candy Crush app at 11:01am EST, it can pretty easily be inferred that the same user took both actions.
Impact uses both deterministic and probabilistic attribution methods. We use deterministic because it’s the most accurate, when you can get the right identifiers. And we use probabilistic because it works great for all the times when deterministic identifiers aren’t available — especially when the time between the connected events is fairly short. And to provide some context, we currently see anonymized IDFAs around 25% of the time, give or take. That is, for ~25% of app-to-app conversion paths, there is already no IDFA, so probabilistic is the only method that will work.
Let’s say for the sake of example that those numbers will invert post-iOS 14, and at that point the IDFA will only be present 25% of the time (industry predictions range from less than 10% to 30%). That means that — while still occasionally helpful — deterministic matching will be much less reliable.
In short, the probabilistic arm of our attribution system will be pulling more weight. And to be perfectly honest, the drop in deterministic matches is probably going to have a negative impact on our match rates. That’s simply the industry-wide truth and anyone who says differently is selling something. That said, we’ve already seen our probabilistic matching system perform admirably and we’re confident that it will continue delivering great results even as it takes on a greater workload.
What about Google?
Google has not yet announced a similar move to their corresponding identifier, the Google Advertising ID (GAID). However, Apple and Google have been doing the consumer privacy tango together for a while now, and it would surprise no one if Google followed Apple’s lead here and made the GAID opt-in as well. The difference is that the Android ecosystem offers another tool for install attribution: the Google PlayStore Referrer parameter. This parameter can be appended to links and used to piece together conversion paths and thus attribute an install to the media which drove it.
So when Google nigh-inevitably makes the GAID opt-in, it will hurt the retargeting industry just as much as the IDFA change will, but it presumably won’t have much of an effect on the attribution industry. In essence, Google has a more precise way to target “those creepy ads that follow you around the internet” without catching other innocuous industries like attribution in the crossfire as Apple has.
How is the industry responding?
There’s no shortage of responses from all corners of the industry. Here are a few highlights from some of our integration partners:
- Branch discusses the change and suggests ways for advertisers to prepare
- Appsflyer prepares for an IDFA-less world and wonders whether privacy and good UX are mutually exclusive.
- Adjust proposes a novel way to retain the IDFA without sacrificing user privacy: the Attribution Hash
How should I prepare for the loss of the IDFA?
If you’re tracking your app in Impact, you should first make sure you know who is handling attribution for you. If you’re tracking with Impact directly, then our notes above about the increased importance of probabilistic matching will apply to you. You can also reach out to your CSM — they’ll be happy to answer any questions you may have.
If you’re working with an MMP, you’ll want to check in with them directly. Chances are, they’ve already put out a blog post or three, but we recommend following up directly to understand how they’re planning to maintain their level of service once iOS 14 becomes widely adopted.
Stay in the know about all the latest changes — get in touch with a growth technologist to find out more at firstname.lastname@example.org.