On October 24th, 2022, alongside the release of iOS 16.1, Apple officially launched SKAdNetwork (SKAN) 4.0, marking a significant milestone in the evolution of privacy-friendly mobile ad attribution. SKAdNetwork (SKAN) is Apple's framework designed to allow advertisers to measure the success of their ad campaigns in a privacy-preserving manner. Following Apple’s release of App Tracking Transparency (ATT) in April 2021 with iOS 14.5 - which prevented ad networks from collecting their own user level attribution data - SKAN has become the primary way to measure install ads on iOS.
SKAN 4 is much more marketer-friendly than previous versions, since it provides more detailed insights into where users came from (source identifiers) and what they did (conversion values). When Apple first announced SKAN 4.0 the marketing community was excited about the change, but adoption of SKAN 4.0 has been slow. This article explains why this was the case, why this is changing, and what you need to do over the next few months to take advantage of SKAN 4.0. But first it gives a quick primer on SKAN itself.
Table of Contents
- What is SKAN, and why is SKAN 4.0 so important?
- Why has it taken so long for SKAN 4.0 to roll out?
- What’s new in SKAN 4.0?
- How marketers can implement SKAN 4.0 to improve mobile attribution
What is SKAN and why is SKAN 4.0 so important?
SKAdNetwork (SKAN) is Apple's framework designed to allow advertisers to measure the success of their ad campaigns & optimize them in a privacy-preserving manner. It does two things:
- Measurement & reporting - it lets marketers know how successful their campaigns have been
- Campaign optimization - it lets ad platforms know how successful ad campaigns have been at driving high value conversions
At a very high-level, SKAN works by Apple acting as an intermediary between advertisers and ad platforms, collecting data at a user-level from both, joining these data together, and then sending non-identifying aggregated data back to both parties. For example, the ad network will never know which specific user was high value according to the advertiser, and the advertiser will never know which advertising campaign was the source of each specific user.
The importance of SKAN has grown significantly following the release of ATT in 2021. With ATT preventing ad networks from collecting their own user level data for mobile install ads, advertisers have had to adapt their strategies to this new framework. This shift has had a meteoric impact on the industry, with estimates suggesting that the return on ad spend (ROAS) for iOS ads on Meta decreased by 40% following the release of ATT.
The release of SKAN 4.0 in October 2022 marked a significant development in the industry. This latest version of SKAN offers more granular data to both ad networks and advertisers:
- Multiple conversions - advertisers and ad networks know how valuable a user is (conversion value) at three different points in time (rather than just one in previous versions)
- More information on where the user came from - advertisers and ad networks know where a user came from (source identifier) with much more precision
- Support for web to app install ads - SKAN 4.0 supports apps shown on web that drive app installs, especially in Google Search
Although industry excitement was high, adoption remained low. The next section explores why that was the case.
Why has it taken so long for SKAN 4.0 to roll out?
SKAN 4.0 was announced in October 2022, but even after nearly a year on, adoption remains low at just over 10%. Adoption is dependent on a number of factors:
- Transitioning to SKAN 4.0 "starts" with an ad network supporting it (and ‘signing’ their ads with SKAN 4.0 signature). Even if advertisers had set up everything in October 2022, without ad network adoption they would not have received any of the additional benefits of SKAN 4.0.
- Advertisers have to set up and configure SKAN 4.0, and if you are implementing via a Mobile Measurement Partner (MMP), then they need to support these configuration options. Examples of MMPs include: Appsflyer, Branch, Singular and Adjust.
- Ad networks need to understand these new schemas advertisers have set up and use this to optimize their campaigns. There is additional complexity here as within the same campaign you may have SKAN 4.0 information and SKAN 3.0 information that needs to be handled
- Ad networks (and MMPs) need to update their reporting so that marketers can see all the new information in their dashboards
This chart from the MMP Singular highlights the interplay of these factors:
- Adoption from ad networks was very low up until mid-July 2023
- At that point, Meta announced they were supporting SKAN 4.0 which led to a significant increase in SKAN 4.0 adoption as marketers rushed to take advantage of it
- When a bug was identified which caused conversion value updates to be overridden, resulting in ‘NULL’ conversion values in SKAN 4.0, ad networks quickly shifted back to SKAN 3. This bug has been fixed in iOS version 16.6. We expect Meta and other ad networks to resume rolling out SKAN 4.0 in Fall 2023.
It’s important to note that even at the ‘peak’ of SKAN 4.0 adoption in July 2023, not all ad platforms had adopted SKAN 4.0, and even those who did lacked full support. For example, Meta only optimized their campaigns on one conversion ‘postback’ and Google have only just announced bidding on these conversion values across any SKAN version!
What’s new in SKAN 4.0?
SKAN 4 introduced a range of new features that make it more valuable for marketers. We will go through each of the new main features in the following sections, but the chart below from appsumer shows how SKAN has evolved over time.
At the heart of SKAN is a commitment to privacy. Previous versions had ‘privacy thresholds’ but these have been replaced in SKAN 4.0 with crowd anonymity. This feature ensures user privacy by only providing data when a certain number of users from a specific campaign install the app. The more users who install an app, and therefore the more difficult it is to identify unique users, the more data Apple sends back to advertisers and ad platforms.
Crowd anonymity in SKAN 4.0 is implemented through a tiered system (from Tier 0 to Tier 3). Reaching a higher tier leads to more data being sent back to you, allowing for more detailed insights into where the user came from (source identifiers) and what they did (conversion values).
So, what are the thresholds for these tiers? Based on observations from AppsFlyer, the thresholds are as follows:
- Tier 0 to Tier 1: Just a few installs per campaign, per day are required. This means that even small campaigns can benefit from some level of detailed reporting.
- Tier 1 to Tier 2: A few dozen installs are required. This level provides more detailed data, making it suitable for medium-sized campaigns.
- Tier 2 to Tier 3: Reaching this tier requires a significant increase in the number of installs. This level provides the most detailed data, but is only achievable for large-scale campaigns.
A postback contains information on where a user came from (source identifier) and what they did on the app conversion value (CV). These reports provide valuable insights into the effectiveness of ad campaigns. SKAN 4.0 introduces a significant change to this process by allowing up to three postbacks per user (up from 1), each covering a different time window.
The first postback window in SKAN 4.0, covering the first 0-2 days of user activity, is particularly important. During this window, you can receive detailed information about user actions, with the ability to group users into 64 different buckets that help you identify how valuable a user is, referred to as the 'fine-grained' conversion value. However, to access these 64 buckets, you must reach Tier 2 of Crowd Anonymity. If you only reach Tier 1, you will receive a less detailed 'coarse-grained' conversion value, categorizing users as High, Medium, or Low value. If you have very few installs per campaign per day, a conversion value of 'NULL' will be returned.
The second and third postbacks provide a longer view of user activity, offering insights into user behavior over time. However, these postbacks only provide coarse-grained conversion values (High, Medium, Low).
Some key points:
- It's important to note that you can define what High, Medium, and Low values in each postback window. For example, a user completing a level in a game may be considered Medium value in the first few days, but if they have only completed one level by day 35, they probably aren't a Medium value user anymore.
- Another key point is that a user must use your app within each window for the postback to be sent. For example, the user must open the app within days 3-7 to trigger the second postback.
SKAN 4.0 also introduces 'lockWindow', a feature that allows app developers to stop measuring user activity in order to receive the postback as soon as possible. This can be done for each postback within its activity window. Note that locking a window does not start the next window. This feature allows for more timely data, enabling quicker optimization decisions.
In previous versions of SKAN, the source of a user's journey was identified solely through a campaign identifier, which was limited to a 2-digit value. This provided a basic level of insight into where the user came from, but lacked the granularity that marketers often need for detailed campaign analysis.
SKAN 4.0 introduces a significant enhancement to this feature by allowing additional source information (configured by the ad network) to be added as you ascend the 'Crowd Anonymity' tiers. This expanded source identifier provides a more detailed view of the user's journey, offering valuable insights for campaign optimization.
At the lowest end of crowd anonymity (Tier 0), you will receive only campaign information. However, once you reach Tier 2, which requires a few dozen installs, you start to receive additional source information - up to four digits and the source-app-id. This allows for a more granular understanding of where your users are coming from, enabling more precise targeting and optimization of your ad campaigns.
Support for Web to App
One of the most significant additions in SKAN 4.0 is the support for web-to-app attribution. This feature allows for the tracking of user activity from web ads to app installs.
While this feature is currently only supported by Safari, it's important to note that Safari holds a dominant position as the mobile browser of choice for a significant majority of iPhone users, with a market share of 93.65%.
This new feature fills a significant gap from previous versions of SKAN. Prior to SKAN 4.0, all install ads that appeared in Google Search were not compatible with SKAN, limiting the scope of campaign tracking. With the introduction of web-to-app attribution in SKAN 4.0, these transitions can now be accurately tracked and attributed, providing a more comprehensive view of campaign performance. Although, just like with the other features of SKAN, this relies on the ad network adopting SKAN 4.0, which Google has not yet completed.
Further, we have seen advertisers drive web ads to a landing page and then use their MMP provider’s probabilistic attribution functionality to track users from their own website to app install. With iOS 17 changes reducing the ability for MMPs to probabilistically attribute users, this measurement approach will become less reliable driving more adoption of direct app install ads.
How marketers can implement SKAN 4.0 to improve mobile attribution
There are two primary options for implementing SKAN: using a MMP or doing it yourself. If you go native then you need to tell ad networks what your conversion values mean and maintain code which calculates them. This option might be tricky to update, since there’s always the possibility you’ll have multiple app versions in the wild at the same time. If you decide to go with an MMP, then you can just send them your conversion value definitions and the MMP will do the rest.
The complexity of SKAN has significantly increased with SKAN 4.0, so you will need to define four conversion value schemas instead of just one:
- Full granularity model if you surpass crowd anonymity thresholds for postback 1
- Coarse conversion model for postback 1 if you don’t surpass crowd anonymity thresholds
- Coarse conversion model for postback 2
- Coarse conversion model for postback 3
While this may seem more complex than previous versions, the following guidance can help streamline this process:
1. Focus on what qualifies as a Low, Medium, or High-value user in each window
- This is the lowest common denominator across all three postback windows, so normalizing back to this is crucial
- Create a low, medium, and high conversion schema for each window. What classifies as a high-value user in the first 0-2 days will differ from a high-value user by day 35
- Remember, a user must return to the app in each window and perform the action in that specific window. Consider whether your Low value could simply be that they returned
- Unless you have a specific reason, we recommend keeping the default postback windows and consider using 'lockWindow' in a later iteration of your implementation
2. Drill into the first postback window
- Monitor your campaign postbacks to see if you are reaching the crowd anonymity thresholds that get you the granular data for the first few windows. If not, consider consolidating your campaigns to meet these thresholds
- Note that while Meta supports SKAN 4 postbacks, they do not optimize on the 2nd/3rd postback, so getting the first postback right is essential
- For conversion windows, ask: How do you most effectively represent value within 2 days of install? If you haven’t already, carry out a segmentation exercise and identify early indicators of high Lifetime Value (LTV) and, if you have the resources, invest in predictive Lifetime Value (pLTV) modeling so you can understand how actions lead to individual revenue bands
Understand how SKAN fits into your holistic marketing measurement strategy
Attribution is just a single piece of the measurement puzzle. Mobile app advertisers with a strong iOS presence should also be considering Media Mix Modeling and Geo-Testing to validate their last-touch attribution that SKAN provides
SKAN 4.0 offers marketers many more insights into the performance of their advertising campaigns vs earlier iterations of SKAN, while still safeguarding user privacy. However, keep in mind that SKAN 4.0 adds significantly more complexity than previous versions, and therefore we recommend a gradual approach to implementation, with continuous testing to ensure optimal setup.
Mammoth Growth has deep experience with MMPs like (Appsflyer, Adjust, Singular and Branch), and we are closely monitoring the evolution of SKAN as more marketers and ad platforms adopt Apple’s framework. Connect with our team and let’s discuss your mobile attribution goals.