iOS 14 significantly disrupted app attribution by restricting access to the IDFA (Identifier for Advertisers), which ad networks and measurement partners previously used to track exactly which ad drove each install. From iOS 14.5 onwards, users had to explicitly opt in to tracking via Apple’s App Tracking Transparency (ATT) prompt, and the majority chose not to. The result was a sharp drop in deterministic attribution data, leaving many app marketers with install numbers that no longer matched their ad platform reports. This article unpacks what actually changed, how the industry adapted, and what it means for your attribution setup today.
What changed about user tracking after iOS 14?
After iOS 14.5, Apple required all apps to ask users for permission before tracking them across apps and websites using the IDFA. When a user declines, the IDFA is no longer available, which means ad networks and mobile measurement partners cannot tie an install or in-app event back to a specific ad click with certainty. Opt-in rates across most app categories settled well below 50%, which meant the majority of iOS installs became unattributed or only partially attributed.
Before iOS 14, the IDFA was available by default. Advertisers and their measurement partners could match a user’s device ID at the point of ad click to the same ID at install, giving a clear one-to-one link between ad spend and results. That deterministic chain broke for users who declined the ATT prompt. If you have ever looked at your dashboard and wondered why your app install numbers look wrong, or why your tracking numbers don’t match between platforms, this is very likely the root cause.
Apple introduced SKAdNetwork as its privacy-preserving alternative for campaign measurement. Rather than passing individual user-level data, SKAdNetwork sends aggregated, delayed signals back to ad networks. The shift fundamentally changed how attribution works on iOS and forced the entire industry to rethink how performance is measured.
How does SKAdNetwork work for app attribution?
SKAdNetwork is Apple’s framework for privacy-safe install attribution on iOS. Instead of sharing user-level data, it sends a postback directly from Apple to the ad network after an install, confirming that a campaign drove a conversion without revealing which specific user converted. This keeps individual identity private while still giving advertisers some signal about campaign performance.
The postback includes a campaign ID and a conversion value, which app marketers can configure to represent meaningful in-app events such as registration, first purchase, or a specific engagement threshold. However, there are important limitations that affect how useful this data is in practice:
- Delayed reporting: Postbacks are sent after a timer window, typically 24 to 72 hours after install, depending on the version and configuration.
- Limited conversion values: Earlier versions of SKAdNetwork offered only a 6-bit conversion value, giving 64 possible states. SKAdNetwork 4.0 expanded this but still requires careful planning to extract meaningful signals.
- No user-level data: You cannot see which individual user converted, only that a campaign contributed to a certain number of installs.
- Crowd anonymity threshold: Apple withholds postbacks for campaigns below a minimum traffic threshold, meaning smaller campaigns may receive no data at all.
For marketers used to seeing granular, real-time attribution data, SKAdNetwork feels limited. But it is the primary deterministic signal available for opted-out iOS users, so configuring it correctly is important for maintaining any visibility into campaign performance.
What types of app campaigns were most affected by iOS 14?
Performance-driven iOS campaigns that relied on user-level data for optimisation were hit hardest by iOS 14. This includes retargeting campaigns, lookalike audience targeting, and any campaign that used in-app event data to optimise bidding in real time. Without the IDFA, ad platforms lost the granular signals they needed to train their algorithms effectively, which often led to a period where ROAS didn’t match reality and campaign performance appeared to decline even when installs were still happening.
Retargeting campaigns
Retargeting on iOS became extremely difficult after iOS 14. Re-engaging lapsed users requires identifying them across apps, which depends on the IDFA. Without it, the addressable audience for retargeting shrank dramatically for most advertisers. Many shifted retargeting budget toward Android or moved to contextual strategies on iOS.
Lookalike and algorithmic audience targeting
Ad platforms like Meta and TikTok use conversion signals to build lookalike audiences and optimise delivery. When iOS 14 reduced the volume of attributable conversion events flowing back to these platforms, their algorithms had less data to work with. This caused a period of instability where attribution data mismatches between platform reports and MMP dashboards became a common complaint. Campaigns that previously scaled predictably became harder to optimise.
What’s the difference between probabilistic and deterministic attribution after iOS 14?
Deterministic attribution matches an install to a specific ad interaction using a unique identifier, such as the IDFA. Probabilistic attribution uses statistical inference, combining signals like IP address, device type, operating system, and timing to estimate which ad likely drove an install. After iOS 14, deterministic attribution became unavailable for users who declined the ATT prompt, making probabilistic methods the fallback for a significant portion of iOS traffic.
The key difference in practice is certainty. Deterministic attribution gives you a confirmed match. Probabilistic attribution gives you a likely match, which introduces a margin of error. This is one of the main reasons iOS install tracking becomes inaccurate when opt-in rates are low: a large share of installs get attributed probabilistically, and some are not attributed at all, showing up as unattributed installs in your MMP dashboard.
Unattributed installs are not necessarily lost installs. They are installs where the attribution chain could not be completed with sufficient confidence. Understanding this distinction helps you interpret your data correctly rather than assuming your campaigns stopped working. If your app installs dropped suddenly in reports without a corresponding drop in actual downloads visible in App Store Connect, attribution loss is almost certainly the explanation.
How did mobile measurement partners respond to iOS 14?
Mobile measurement partners (MMPs) such as Adjust, AppsFlyer, and Branch responded to iOS 14 by building frameworks that combine SKAdNetwork data with probabilistic modelling and aggregated reporting to give advertisers the most complete picture possible under the new privacy constraints. Each platform developed its own approach, but the common thread was moving away from user-level data toward modelled, aggregated insights.
AppsFlyer introduced its Privacy Cloud and SKAdNetwork solution to aggregate and model conversion data. Adjust launched its own SKAdNetwork management tools alongside its probabilistic attribution engine. Branch similarly adapted its measurement infrastructure to handle the split between opted-in users (where deterministic attribution still works) and opted-out users (where modelled data fills the gap).
These tools help reduce the gap between what your ad platform reports and what your MMP reports, but they do not eliminate it entirely. Some degree of discrepancy is now a permanent feature of iOS measurement. The practical implication is that marketing numbers that don’t add up perfectly are not necessarily a sign that something is broken; they often reflect the structural limitations of post-iOS 14 attribution.
Should app marketers still worry about iOS 14 attribution today?
In 2026, iOS 14 is no longer a new disruption, but it remains the baseline reality for iOS attribution. The privacy framework Apple introduced has not been rolled back, and SKAdNetwork has continued to evolve. App marketers who have not fully adapted their measurement setup are still likely experiencing attribution data mismatches, underreported ROAS, and difficulty scaling iOS campaigns with confidence.
The more relevant question today is whether your current attribution setup is properly configured to work within these constraints. Key areas to review include:
- ATT prompt placement and copy: The timing and framing of your opt-in prompt directly affects your opt-in rate, which determines how much deterministic data you collect.
- SKAdNetwork conversion value mapping: If your conversion values are not mapped to meaningful in-app events, you are not extracting the signal SKAdNetwork can provide.
- MMP configuration: Ensure your MMP is set up to handle both SKAdNetwork postbacks and probabilistic attribution correctly, with modelled data filling gaps where needed.
- Cross-platform benchmarking: Use App Store Connect data as a ground truth to sense-check your attributed install numbers and identify where the gaps are largest.
If your marketing numbers still don’t add up or you are seeing persistent discrepancies between ad platform data and your MMP, the issue is rarely the platforms themselves. It is almost always a configuration or methodology gap in how attribution is being handled across your iOS funnel.
At Wuzzon, we work with MMPs including Adjust, AppsFlyer, and Branch on a daily basis and help app teams set up proper attribution frameworks that account for the realities of post-iOS 14 tracking. If you want to understand where your iOS attribution gaps are and how to close them, explore our app growth services or speak to one of our specialists directly.
Frequently Asked Questions
How can I improve my ATT opt-in rate to get more deterministic attribution data?
The single biggest lever is the timing and framing of your ATT prompt. Showing the prompt after a user has experienced value in your app — such as completing onboarding or reaching a key milestone — significantly increases opt-in rates compared to triggering it on first launch. You can also add a pre-permission screen that explains in plain language why tracking benefits the user, such as seeing more relevant ads or supporting a free app. Small copy and UX changes here can move opt-in rates meaningfully, and since each opted-in user restores full deterministic attribution, even a 10–15% improvement compounds across your install volume.
What is the best way to set up SKAdNetwork conversion values so they actually provide useful data?
The key is mapping your conversion values to events that reflect real business outcomes within the early post-install window, typically the first 24–72 hours. Rather than using a single binary install event, define a value schema that encodes progression — for example, distinguishing between a user who registered, one who completed a tutorial, and one who made a first purchase. SKAdNetwork 4.0 introduced coarse and fine conversion values across multiple postback windows, giving you more flexibility to capture later events like day-3 or day-7 revenue signals. Work with your MMP to design a schema that balances granularity with the crowd anonymity threshold, ensuring you receive postbacks even from smaller campaigns.
Why do my Meta or TikTok campaign numbers still not match my MMP dashboard even after configuring SKAdNetwork?
Discrepancies between ad platform reports and your MMP are now structurally expected and do not necessarily indicate a technical error. Ad platforms apply their own modelling and view-through attribution logic, while your MMP uses a different methodology to deduplicate and attribute installs. The gap is widest for opted-out users, where both sides are using probabilistic or modelled data that may reach different conclusions. The practical approach is to stop trying to reconcile the numbers to zero and instead use each data source for its intended purpose: ad platform data for in-platform optimisation decisions, and MMP data as your single source of truth for cross-channel performance comparison.
Is Android attribution affected by the same iOS 14 privacy changes?
Android was not directly affected by Apple’s ATT framework, but it is undergoing its own privacy transition. Google has been phasing out the Google Advertising ID (GAID) through its Privacy Sandbox initiative, which introduces a similar shift toward aggregated, privacy-preserving attribution signals on Android. The timeline and technical implementation differ from SKAdNetwork, but app marketers should be actively monitoring these changes and ensuring their MMP is prepared to handle Android attribution under the new framework. For now, Android still offers more granular attribution than opted-out iOS users, but treating it as a permanent safe haven from privacy changes would be a strategic mistake.
What should I use as a reliable benchmark if my MMP install numbers are incomplete?
App Store Connect is your most reliable ground truth for total iOS install volume because it reflects every download regardless of attribution status. By comparing your total attributed installs in your MMP against the download figures in App Store Connect, you can calculate your attribution gap and understand what percentage of your installs are going unmeasured. This gap is also useful for sense-checking ROAS: if your MMP is only attributing 60% of installs, your actual cost per install and return on ad spend figures need to be interpreted with that gap in mind. Regularly benchmarking these two data sources together is one of the most practical habits for maintaining measurement accuracy on iOS.
Can I still run effective retargeting campaigns on iOS after iOS 14?
Effective iOS retargeting in the traditional IDFA-based sense is largely no longer viable for opted-out users, which represents the majority of most iOS audiences. However, there are practical alternatives worth exploring. For opted-in users, deterministic retargeting still works normally, so maximising your ATT opt-in rate directly expands your retargetable audience. For opted-out users, contextual retargeting strategies — such as targeting based on app category, content type, or behavioural cohorts without individual identifiers — can partially substitute. Many advertisers have also shifted retargeting investment toward Android where addressability remains stronger, or toward owned channels like push notifications and email where no third-party tracking is required.
How do I know if my current MMP setup is actually configured correctly for post-iOS 14 attribution?
A few diagnostic checks can quickly reveal configuration gaps. First, verify that SKAdNetwork is enabled in your MMP and that your app’s Info.plist includes the correct SKAdNetworkIdentifier entries for every ad network you are running. Second, audit your conversion value schema to confirm it maps to real in-app events rather than being left at default settings. Third, check whether your MMP’s probabilistic attribution is active and correctly weighted for opted-out traffic. Finally, compare your attributed install volume against App Store Connect downloads to quantify your attribution gap — if the gap is larger than 40–50%, it is a strong signal that your configuration needs attention. If you are unsure where to start, working with a specialist partner who operates across multiple MMPs daily is the fastest way to identify and close the gaps.
Related Articles
- 6 reasons every ad channel claims credit for the same app install
- Why did my numbers change without any changes made?
- What is attribution tracking in mobile app advertising?
- Should you advertise your app before or after launch?
- How do you advertise an app using Google App Campaigns?
- Should you advertise your app globally or locally first?
- How do you advertise an app in the Dutch market?
- The complete app store optimization checklist for iOS and Android
- What is app store keyword research?
- 6 app store metadata optimization tips for iOS and Android
EN 
NL 