For years, mobile app marketing ran on precise tracking. Advertisers could see which ad led to which install, follow each user's behaviour inside the app and feed that data back to platforms to find more people like their best customers.
That changed when Apple introduced App Tracking Transparency with iOS 14.5 in 2021. Apps now have to ask permission before tracking people across other companies' apps and websites, and many people say no. Attribution on iOS moved largely to SKAdNetwork, Apple's privacy-preserving framework, which reports installs in aggregate and with delays, and more recently to its successor, AdAttributionKit.
The result is less detail and more uncertainty. It is not the end of effective user acquisition. It is a shift in what drives it.
What changed for app marketers
Three things became harder at once. Targeting lost some of the individual-level data that powered detailed audiences. Attribution became aggregated and delayed, so it is harder to link a specific campaign to specific users. Optimisation slowed down, because the signals platforms use to learn arrive later and in smaller quantities.
Advertisers who relied most on precise, user-level tracking felt the change most sharply. Those with strong creative, clear value propositions and good in-app experiences adapted faster.
What still works
Creative that explains the app quickly.
With less precise targeting, the ad itself does more of the work of finding the right users. Showing the core value of the app in the first seconds, with real screens and real use, remains one of the strongest levers available.
Well-chosen conversion signals.
Frameworks like SKAdNetwork allow advertisers to define which early in-app actions they report. Choosing actions that genuinely predict long-term value, rather than simply the install, helps platforms optimise towards better users.
App store optimisation.
More people now discover apps through store search and browsing. Strong store listings, clear screenshots and relevant keywords matter more than ever. Apple's custom product pages let advertisers create different versions of their store page to match different campaigns and audiences.
Store search ads.
Advertising inside the app stores themselves, such as Apple Ads on the App Store and Google's app campaigns, reaches people at the moment they are looking for an app, with first-party context that does not depend on cross-app tracking.
A healthy channel mix.
Relying on a single network is riskier when measurement is less precise. Spreading acquisition across several platforms, and comparing them with consistent measures, reduces dependence on any one source.
Measuring with less detail
With user-level attribution limited, app marketers increasingly combine several measurement approaches:
| Platform and framework reports | for day-to-day optimisation within each channel. |
|---|---|
| Incrementality tests | such as switching spend on and off in particular regions, to see what campaigns actually add. |
| Marketing mix modelling | to estimate the contribution of each channel over time. |
| Cohort analysis | inside the app, tracking how groups of users acquired in a given period behave, to judge the long-term value of acquisition. |
No single method gives the complete answer. Used together, they give a clearer view than any one report.
Retention is part of acquisition
When acquiring each user is harder to measure and optimise, keeping those users matters more. Strong onboarding, useful notifications that people have agreed to receive, and a product that delivers on the ad's promise all raise the value of every install. They also improve the signals sent back to ad platforms, which helps future campaigns find similar users.
Growth without the old signals
The privacy shift removed some of the precision app marketers were used to, but the fundamentals of growth never depended on tracking alone. Apps that explain themselves well, show up strongly in the stores, measure with several methods and keep the users they win are well placed to grow, whatever the next change to the rules turns out to be.
