Hello everyone!
My name is Hrysha Kobyzhskyi. I’ve spent four years working in a marketing agency, where I became the SMM Lead and developed communication strategies for brands on social media. Then, I transitioned into IT as a Marketing Manager and, for the last year and a half, I have focused on user acquisition for mobile iOS apps in the Guru Apps business. Our apps are used in over 180 countries worldwide.
Moving to a new profession was a real challenge that required mastering new tools and approaches. One of the key aspects was learning SKAN attribution — an essential element that allows effective analysis and traffic management. In this article, I’ll share my thoughts on traffic attribution for mobile iOS apps across the most popular platforms (Meta, TikTok, Google). Additionally, we’ll discuss the latest changes in this field and predict future developments.
The main goal for any marketer is to efficiently attract users and ensure a return on investment. Paid advertising on popular platforms like Meta, TikTok, and Google is widely used for this purpose. However, launching an ad campaign is just the first step. It’s crucial to deeply analyze the results and make data-driven decisions.
One of the main challenges while working with iOS apps is event attribution. It’s complex and requires a tailored approach since there are no straightforward solutions.
Magical Events and Where to Find Them
User acquisition for mobile iOS apps on advertising platforms can generally be done in two ways:
Direct purchase through SKAdNetwork (when the user clicks on an ad and goes directly to the App Store).
Web2App traffic (when the user clicks an ad, lands on a webpage, and then proceeds to the App Store).
The difference is that with web2app you can immediately see events (purchases, installs, etc.) in the dashboard, which leads to quicker decision-making. With SKAN, events are recorded with a delay of 24 hours and even more. So, why use SKAN if web2app is more convenient?
Here are a few reasons:
SKAN is designed specifically for advertising iOS apps. The algorithm targets an audience already interested in the app, allowing you to attract more motivated users with higher LTV (lifetime value), as the practice has shown.
SKAN also helps diversify traffic. When web2app results don’t meet expectations, SKAN can compensate for this since the auctions in SKAN and web2app are different. For example, during Black Friday, when the web2app auction is overheated.
Long Story Short, Why This Situation Exists:
In 2021, Apple introduced App Tracking Transparency (ATT), which limited access to user data. ATT prohibited sharing the IDFA (Identifier for Advertisers), significantly complicating user behavior tracking.
How It Used to Be: Every user had a unique IDFA, allowing advertisers to track all actions before and after an installation. Events were transmitted in real time, and all information was immediately available.
What Has Changed: Personalized attribution was replaced by aggregated attribution, without the ability to track individual user actions, making ad optimization more difficult. Now, events are transmitted with a 24 to 72-hour delay.
To help advertisers evaluate campaign performance under the new rules, Apple introduced the SKAdNetwork framework. It provides aggregated data without revealing personal user information, allowing marketers to track campaign effectiveness, though with certain limitations. Most advertisers are now working with the latest versions of SKAN 3.0 and SKAN 4.0, which we’ll discuss later.
All platforms, including advertising giants like Meta, TikTok, and Google have begun adapting to the new conditions by integrating attribution solutions. This allows them to gather the maximum amount of data for optimization without violating ATT.
What Meta Offers
Meta has quickly integrated SKAN to provide advertisers with information. However, this brought certain limitations, affecting the effectiveness and scalability of purchases:
There is a limit on the number of active campaigns for an app. Initially, you could create 9 campaigns with 5 ad sets each. Later, the limit was raised to 18 campaigns.
SKAdNetwork has a 24-hour timer for “event anonymization,” after which Meta receives the installation information, but the exact time of the event is unknown, leading to a delay in processing and complicating analysis. Additionally, some events may be lost due to privacy thresholds.
Meta understood that this wasn’t enough for advertisers. Therefore, in 2023, they announced Advanced Event Measurement—a new event measurement protocol for mobile apps that compensates for SKAN’s limitations. It allows tracking almost real-time events due to statistical modeling. Additionally, advertisers gained the ability to choose attribution windows—1, 7, or 28 days post-click. This greatly simplified the work for marketers, as decisions could now be made faster and more efficiently. A small tip: the system doesn't always model events accurately, so it's better to verify this data with a mobile measurement partner if the results seem suspicious.
An MMP (Mobile Measurement Partner) is a platform that helps mobile apps collect, attribute, and aggregate data and events. It allows for a comprehensive assessment of the effectiveness of ad campaigns. In analogy, MMP is like a fair football referee making decisions regarding attribution. Some of the most popular MMPs include AppsFlyer and Adjust.
TikTok's Solution
TikTok has also adapted to the changes, although it has its own limitations and specifics compared to Meta. You can create up to 15 ad campaigns per app, each with 5 ad groups. Similar to Meta, due to SKAN’s framework, there is also a 24-hour timer for receiving installation information. In my experience, events in TikTok's ad dashboard appear faster compared to other platforms.
Additionally, less than a year ago, TikTok launched its own attribution solution — Self-Attributing Network (SAN). SAN integrates closely with SKAdNetwork, enabling TikTok to optimize ads based on its own data, supplementing the aggregated data from SKAN. The main advantage for advertisers is more accurate and faster event tracking, allowing them to manage ad campaigns more efficiently.
Another upgrade came with the release of the Advanced Dedicated Campaign (ADC). This new campaign type allows real-time event tracking in the dashboard if you have SKAN 4.0. For SKAN 3.0, events can also be tracked in real-time, but only through an MMP. These data will not appear in the dashboard.
Google’s Adaptation
Google has integrated SKAN and was the first to introduce Conversion Modeling as part of its adaptation. This technology uses machine learning to predict and model conversions that could have been tracked using traditional methods but cannot be directly captured due to SKAN limitations. Additionally, events that are tracked later are attributed to the day they occurred, meaning that data of one day changes over time, and the full picture is typically received within about two weeks.
For SKAN launches, only one campaign type is used — a Universal App Campaign. Here, we cannot select narrow targeting or placements and must upload both text and photo/video creatives. The system automatically distributes traffic across placements (Google Search, YouTube, partner networks), based on where the target audience is. A broad audience and the inclusion of all necessary creative elements in the settings (static and video) allow the algorithms to gather more data and find the best combinations for better results.
Laying Out the Tarot Cards
The key for a marketer when working with SKAN is to correctly interpret the data received and make decisions based on them, despite the limitations.
To achieve this, you can use the following techniques:
Define your own cost coefficient based on dashboard price vs. real price. For example, the price in the dashboard for a specific geo is $50, but the total cost is $25. Therefore, our reference point will be a cost coefficient of 2, meaning that if we want to get a total price of $20, we need to reduce the dashboard price to $40. Importantly, this exercise can only be done when our geo works exclusively with SKAN and no other channels are in rotation.
At the start of the launch, focus on top-funnel metrics (CPM, CPC, CTR). If these significantly exceed your benchmarks, you can stop the campaign without waiting for core events and avoid ineffective spending. With Meta AEM and TikTok ADC, this is easier, as events appear immediately (although there can also be delays at the start).
Assess the effectiveness of SKAN campaigns over a certain period (3-7 days). Data for a single day may be somewhat random since events arrive in no particular order. For example, your target price is $30. On one day, you may see a price of $22, on the second day $37, on the third $24, and on the fourth $32. The average price is $28.75, so everything is fine. Evaluating day by day individually could lead to incorrect decision-making.
Regularly compare dashboard data with MMP data. For SKAN MMP it is the source of truth. Cross-checking helps ensure that events are being correctly reported in the ad dashboard and that the picture matches reality. If there are anomalies, it's worth double-checking the integration setup with the MMP and whether data is being correctly transmitted.
What’s Next?
Platforms are striving to improve the quality and speed of event attribution. But, anyway, there are still limitations. Much of it depends on Apple and what features they allow with new versions of SKAN.
SKAN 4.0 was aimed to improve and fix some gaps. In this version, advertisers could receive three postbacks (0-48 hours, 3-7 days, 8-35 days), allowing user action tracking at different stages of app interaction. Before this, in SKAN 3.0, there was only one postback in the 24-hour interval. The privacy threshold has been changed because previously, data might not be sent if the number of installs didn’t reach a certain threshold to protect user privacy.
However, the number of innovations turned out to be quite numerous and complex: changes in algorithms, limited compatibility, multiple postbacks, hierarchical conversion structures, etc. As a result, advertisers couldn't fully understand how to effectively use all of these features together. Furthermore, different platforms integrated the new framework at different times. This has led to a situation where some advertisers stuck with SKAN 3.0, while others switched to SKAN 4.0.
Under the circumstances of increasing user privacy demands, the market is discussing the upcoming SKAN 5.0, which Apple has announced at WWDC 23. It is expected to be easier to integrate and at the same time, it’s going to be more efficient in terms of attribution and optimization. Also, at WWDC 24, Apple announced AdAttributionKit — an additional framework that will provide wider possibilities for attribution and retargeting and will be able to work with third-party app stores (SKAN works only with the App Store). Both frameworks will coexist and will not affect each other’s operations.
Conclusions
A mutual dependence has emerged between Apple and advertising platforms. Everyone understands that for advertisers to spend their budgets, conditions must be created where they can control the effectiveness of their spending. For this, higher attribution accuracy is essential. This is why we can see a gradual simplification of conditions for SKAN purchases, and I predict that in the future, referring to Apple’s announcements and additional solutions from Meta, TikTok, and Google, attribution will become more accurate, offering greater opportunities for direct app acquisition. If you're interested in working in this field and eager to grow, we'd be happy to welcome you to our team!
