Apple has released iOS 14.5 on Apr 26, 2021 along with exciting features: new emojis, unlocking iPhone with Apple Watch, different voice options for Siri, updates in Apple Maps, and more¹. As exciting as the new features sound, you probably did not come here to learn about the new emojis, instead a discussion on new privacy practices might be a more relatable topic if user acquisition is a part of your everyday life. There are some important changes in iOS 14.5 that have a significant impact on the mobile advertisement world and the success of all players in the game, which we’ll uncover. However, before deep diving into them let’s start simple and clarify some important concepts.
User tracking in mobile marketing: RIP IDFA?
One of the terms you might be hearing often or have become familiar with is IDFA. If you heard about cookies, the information saved by your web browser, you may think of IDFA or Identifier for Advertisers as cookies for mobile devices. IDFA is a unique and random device identifier assigned by Apple. Its most significant difference from browser cookies is that it has a longer lifetime; it remains constant until the user decides to reset it.
IDFA has been the backbone of mobile marketing because it enables advertisers to track and gain information about users and their activities without giving away users’ sensitive data such as their personal information. With IDFA, advertisers can deliver more customized ads to users. Therefore, it helps advertisers to optimize the campaigns and campaigns’ budget to achieve better results and hit their KPIs while the users see more relevant ads when browsing through their mobile apps. It also helps mobile measurement partners (MMPs) to make more reliable attributions, which means, for example, they can identify which campaign led the user to install the app with a high level of accuracy. With the iOS 14.5 update, IDFA still exists, however, mobile apps need to ask explicit consent for IDFA from users.
A brief history of limited ad tracking
LAT or Limit Ad Tracking was introduced with the release of iOS 10 in 2016 and deprecated with the iOS 14 update. LAT was a feature that allowed users to opt-out of personalized ads. By default, disabled LAT meant that the user could be tracked to deliver customized advertisements. When LAT was enabled, the user’s IDFA appeared as a string of zeros, which may seem to make it impossible to accurately attribute. In order to overcome this challenge, most of the mobile measurement partners developed techniques such as probabilistic attribution, which uses statistics and machine learning to attribute anonymized IDFA values. What enabled probabilistic attribution to work well was that when LAT was disabled by default, most users did not bother to change their settings to enable it. So a large portion of the users still had their traditional IDFAs (i.e. data from a large number of users was available to predict how the small number of users are behaving). With the new set-up, the default is no-IDFA and users have to specifically allow apps to access their IDFA. The industry experts estimate only about 11 to 36% of users will actually do this, which means only a small number of users’ data will be available to make a prediction on how a large number of users are behaving.² ⁴ ⁵
With probabilistic attribution, another difference is that the attribution window (mostly up to 24 hours) is usually shorter than the window of other methods. Details of how user data can be received and tracked after the deprecation of LAT with the iOS 14 update⁶, are explained in the AppTrackingTransparency (ATT) framework.
What is the AppTrackingTransparency (ATT) framework?
AppTrackingTransparency or ATT is the framework used to determine user privacy preferences starting with iOS 14. In other words, ATT is used to provide the status for application tracking.
According to Apple,
“Tracking refers to the act of linking user or device data collected from the application with user or device data collected from other companies’ applications, websites, or offline properties for targeted advertising or advertising measurement purposes. Tracking also refers to sharing user or device data with data brokers.”³
Apple specifically states that,
“Placing a third-party SDK in the application that combines user data from the application with user data from other developers’ applications to target advertising or measure advertising efficiency, even if you don’t use SDK for those purposes.”
⁷ is an example of tracking (such as MMP SDKs).
Developers must use the ATT consent prompt to request permission to track the user and access the IDFA by specifying a string that explains the purpose of the track in the prompt. If the user doesn’t give permission, IDFA will be all zeros (similar to the LAT enabled case before it was deprecated). Developers can display an ATT prompt whenever they want but IDFA will only be returned once the user gives the permission. The most important change with iOS 14.5 is that applications need explicit permission to track users. The prompt can be shown for once and the decision of the user on sharing their IDFA will be final unless they uninstall and reinstall the app again. As a result of this huge change, app owners need to carefully build their consent flow to find the right time to show the prompt if they are willing to collect IDFAs for their marketing campaigns.
What is SKAdNetwork and what are the setbacks?
StoreKit Ad Network or SKAdNetwork is an API that allows advertising networks to attribute application installations to a particular campaign by receiving a signed postback from Apple. SKAdNetwork helps advertisers measure how many installs resulted from a campaign and the success of the campaign while maintaining user privacy. Ad networks must register with Apple to receive postback calls. App developers don’t need ATT permission to use SKAdNetwork. Granted that it might be a part of the solution to the non-trackable IDFA, SKAdNetwork comes with its own challenges such as no real LTV, postback delay and limited granularity.
SkAdNetwork measures conversion value, number of installs and post-install data in a very limited way. Conversion value enables application developers and UA managers to understand the quality of the user they acquire. It is defined by 6-bits (64 different values) and can be shared with advertising networks via postbacks. Since it’s a part of the limited information advertisers can get from SkAdNetwork, conversion value must be carefully configured to measure post-install activity for more efficiency in campaign optimization.
SkAdNetwork doesn’t share real-time postbacks, postbacks are built on a timer mechanism and it works with a minimum of 24-hour delay. The 24-hour timer restarts each time the application updates the conversion value with a value greater than the previous value. When the timer expires, the conversion value is final and subsequent attempts to update the conversion value have no effect. SKAdNetwork sends the install notification postback to the advertising networks within 0-24 hours after the timer expires. In other words, postbacks are delayed at least 24 hours and the delay can take even more time and the postback only contains the final conversion value. Also, each ad network can receive only one install postback.
Limited granularity is another setback advertisers must deal with when using SkAdNetwork. Postbacks do not include device-level and creative-level data and data is limited to campaign-level and 100 different campaigns per network.
What has been the impact of iOS 14.5?
After the rollout of iOS 14.5 advertisers and user acquisition teams have been experiencing a major impact on their businesses.
To name a few:
- The observed CPI is usually higher than the bid for iOS 14.5 campaigns. Do not be surprised when you see your CPI is three times higher than your bid in the ad network.
- Networks find it harder to estimate the conversion rates (or install probabilities) for user segments, hence the auction mechanisms fail to target the right users compared to before iOS 14.5.
- As a result, the cost of running campaigns has increased since with iOS 14.5 release it has become harder to target personalized ads.
- More installs have been classified as organic installs. Probabilistic attribution models of MMP’s are great but they are not perfect. A larger number of installs cannot be attributed accurately to a particular campaign, hence leading to installs being attributed as organic.
- ROAS optimizations need to adjust themselves to work with the SKAN revenues (coming from delayed postbacks) intelligently.
- To make effective immediate campaign optimization, use of predictive models is a must. Delayed postbacks and limited granular data make UA managers’ jobs harder than ever before. However, by leveraging the insights deducted from historic data, AI based automation can improve the quality of campaign optimization.
These are unprecedented times for all user acquisition folks, that’s for sure. Luckily, some tools adjust the changing landscape and help you maintain your strength in the market. To unlock the power of optimizing campaigns with predictive models check out UAhero, the AI-powered UA platform made for boosting the profits of your games.
UAhero does not rely on user-level data and performs its optimizations across multiple campaigns and ad networks so that the impact of poor performance on any of the ad networks is minimized. Its automation capabilities also minimize the impact of the delay in receipt of postback data and make the life of advertisers easier. Working with multiple MMPs provides additional insight on how attribution models of the MMPs are working in the New World Order for improving the effectiveness of your campaigns.