How to be strategic about attribution after iOS 14.5
Paul H. Müller, Key Advisor to the CEO, Adjust, Apr 2, 2021.
Important SDK update for iOS 14.5
Today, Apple started enforcing restrictions on SDKs related to AppTrackingTransparency (ATT). Our SDK was one that was flagged because it had code that Apple indicated as being in violation of their guidelines. This code was in the SDK to collect information for our fraud prevention suite. We’ve since released SDK 4.28 which addresses this concern and all apps that have submitted with it have been approved. We recommend all clients update to SDK 4.28 before submitting any app.
We take consumer privacy very seriously and support Apple’s new changes. It’s important for us as a leader in attribution and as an expert in this space to arm you with pointers on how to think about app submissions, SDKs, and attribution going forward on iOS.
Transparency and honesty are key values for us. Therefore, we wanted to highlight a bit more on what happened today.
Spoofing of our SDK presents a significant challenge to our data integrity and is a potentially expensive attack on our customers, so we created code that would collect information that better allowed us to find such behavior. Apple saw these symbols and flagged them because they were similar to symbols being used in other SDKs that together could be used to create a persistent ID, even if a user didn’t consent. We’ve posted a list of all symbols that we know of, that were flagged by Apple here.
We never used these symbols for this purpose, however some were in our SDK. As this code was only used for research, we’ve removed it from the SDK and recommend that you update your SDK to SDK 4.28 because it’s important that all apps and the SDKs in them comply.
Separately, we are testing an ATT consent flow in our Adjust SDK. We have found that if you reference ATT but only ask for consent in OS 14.5+, then a reviewer may reject you the first time since they expect the consent flow on all operating systems. If you then resubmit with information that your consent is only present on OS 14.5+, you’ll get approved. To save time, we suggest that you add this note to the original app submission.
The key methods of attribution going forward
Moving beyond the events of today, we wanted to outline our view on the key methods of attribution going forward:
SKAdNetwork: As we all know Apple’s system provided attribution mechanism has limitations, however it should be very accurate in last-click attribution of aggregated installs.
Deterministic Attribution: If a user consents to sharing IDFA both on the media source your ad appears and within your app once they download it, IDFA will be present on both sides and we can match deterministically as long as you run your click through Adjust tracking links.
Imprecise or Probabilistic Attribution: We can use some basic device information to create a best guess at which advertisement drove which install. While less accurate, this level of data will provide you with enough information to analyze creatives and build some efficiency models around your media spend. Because this method is not deterministic, nor is it persistent, we believe it is acceptable within the policy and you have the option to enable it.
In order for continued success, you’ll need to understand how to use the above strategies together or separately and the trade-offs of each. Below, we highlight some plans and best practices that we have collected from many conversations with advertisers:
Some advertisers will only use SKAdNetwork. Adjust supports this, but we believe this will limit efficiency. While last click attribution will be precise, this could hinder the ability for advertisers to analyze creatives, run web or influencer campaigns, and get long-term return reports and cohorts against the media spend. Further, the lack of longer term visibility decays the ability to optimize the app user experience for new installs.
SKAdNetwork and Deterministic:
Some advertisers will only use SKAdNetwork and deterministic attribution. To do this they will run Adjust attribution links as before, and leave Probabilistic Attribution off. If a user consents to share IDFA on both the app serving the ad and their own app, attribution will be precise, as will long term user attribution, ROAS and cohorts analysis, and creative reporting as it has always been on these users. Unfortunately, because only a small percentage of users will consent on both sides, you will lose visibility into detailed attribution on a majority of your users, leaving you with similar limitations referenced above.
SKAdNetwork, Deterministic and Probabilistic:
Some advertisers will use all three of the above methods. This poses complexity but the most opportunity. The complexity is that if a user starts with IDFA shared on the app that shows the ad, and then doesn’t consent to share IDFA with the advertiser upon download, that install will face duplication. This duplication will occur because it’ll get SKAdNetwork attribution and some percentage of the time it could also get probabilistic attribution. Further, if a user is attributed deterministically, that same user may also have gotten a SKAdNetwork install as well from another media source or the same source that couldn’t be de-duped. Bottom line: duplication becomes an issue here. However, this method provides access to much more visibility into creative data, complete support for web and influencer campaigns, as well as ROAS and longer term cohort maturity analysis, leading to more insights and ultimately better apps.
Developers, MMPs, and all players in the app ecosystem are working to determine the best way to continue to get the data they need while adhering to Apple’s new policy. We believe the third path of combining all three attribution methods has the most potential in achieving this goal.
As SKAdNetwork rolls out and advertisers choose their preferred method, we’ll also be building more tools to better measure duplication rates more accurately. As we better understand duplication rates, advertisers can price their media buys properly and still buy profitably while having enough insights to optimize their buys and their products.
We believe that this industry should have transparent and open conversations around these topics and on strategies for how to deal with potential loss of insights and decays in performance. There will always be evolutions and changes in our growing industry and if we come together as partners, we can develop a better industry-wide understanding that not only helps one another, but also moves our industry forward. To that end, please look for us to post more blogs, host industry forums, and deliver webinars that address these topics. We welcome our customers, our partners and all MMPs to join.
Please reach out if you have any questions to your Account Manager, Customer Success Manager, or email@example.com.