Mobile attribution in the age of privacy: Challenges and opportunities

By Michal Wagner
atribuição mobile na era da privacidade - quadrado

Mobile measurement and attribution as we know it is about to undergo a major transformation amid a shift towards a more privacy-centric relationship with consumers. 

But compared to some of the other curve balls 2020 threw with COVID-19, when it came to consumer privacy the writing had been on the wall for some time. Between the introduction of Limited Ad Tracking (LAT), new regulations, and the planned deprecation of the third-party cookie, there had been a slow but consistent shift towards protecting personal data.

It all culminated in Apple’s June announcement of the AppTrackingTransparency (ATT) framework, which requires iOS users to opt-in to measurement. This fundamental change, which is undeniably good for consumer privacy, will inherently alter the way we measure and attribute marketing activities.  

For those who don’t opt-in, Apple said that its SKAdNetwork solution would be the primary deterministic attribution mechanism for aggregated attribution. SKAdNetwork has a number of limitations – more on those later – which severely limit data collection and optimization. 

This creates a reality where there are two completely different measurement methods. Those who opt-in will be grouped with the much larger segment of Android users as the Google Advertiser ID (GAID) matching remains unchanged. For those who opt-out (expected to be the vast majority), SKAdNetwork and some other tools we will discuss ahead will be used.

If this sounds complicated, it’s because it is.

So, the question is how do you continue to add value when measurement is going through such a fundamental change?    

In this blog we will discuss what changes are taking place, how we can still measure and why attribution is, more than ever, a must-have in the age of privacy. 


Getting started with mobile attribution in the age of privacy: The complete guide


The age of privacy

In recent years there has been a gradual move towards a more privacy centric mindset. Apple first replaced its Unique Device Identifier (UDID), with the ID for Apple (IDFA) in 2012. A UDID was a permanent device number and sharing it could not be turned off whereas an IDFA could be reset.

In 2016, Limited Ad Tracking (LAT) was introduced, which allowed Apple users the option to opt out of measurement on their devices. As many as 30% have done so in early 2021.

Then in 2018, we saw two major pieces of legislation and regulation. The General Data Protection and Regulation (GDPR) and the California Consumer Privacy Act (CCPA) both aimed to protect data and personal information. 

In early 2020, Google announced it would be phasing out their 3rd party cookie, something that Safari and Firefox had already done. In 2021, the search giant declared that once third-party cookies are phased out, they will not build alternate identifiers to measure individuals as they browse across the web, nor will we use them in their products.

But it wasn’t until Apple’s announcement of ATT and the updates to SKAdNetwork that mobile attribution really felt the tides of change. 

Once enforced in early Spring 2021, according to Apple, iOS apps will have to show users this message:

Apple iOS 14 ATT new message

It is predicted that only 2%-20% of users will opt-in, rendering the IDFA all but obsolete. As a result marketers will have to adjust to measuring via SKAdNetwork and additional tools.  

Types of measurement

To understand more about these changes, we must first discuss the types of measurement that will be available to app marketers when ATT is enforced.

skadnetwork six bit mechanism (types of measurement)

User level deterministic attribution

This method uses multiple identifiers including device IDs (Apple’s IDFA and Google’s GAID), Google Play Referrer, and customer identifiers (i.e. hashed email addresses, loyalty card numbers) to accurately identify users and measure their actions (e.g. a click on an ad, an install, and an in-app event).


SKAdNetwork is Apple’s deterministic attribution mechanism that aggregates attribution data for iOS apps.

It is a privacy-centric framework that aims to measure app installs and campaign performance in aggregate without user level data.

Using a 6 bit system, which allows users to be graded from 0-63, advertisers gain some insight into the quality of users they are converting.

The limitations and challenges of measurement in SKAdNetwork

While we can all agree that privacy is a good thing for both consumers and the mobile ecosystem, these changes do not come without challenges. 

First and foremost, there are several limitations to SKAdNetwork which greatly reduce the quantity and quantity of data available to app marketers. 

Chief among these challenges are:

Lack of ROI/ LTV data

SKAdNetwork only delivers one postback that is based on data signals from early on in the funnel, such as installs. There is little or no postback data from in-app events.

Since the majority of LTV data is derived from in-app activity, app marketers, between the timer limit and the postbacks, lose several indicators that help measure the LTV/ROI of their campaigns.

Timer limit

SKAdNetwork has a timer mechanism that begins when the SK function is called for the first time (usually the first launch). However, after a 24 hour period the timer expires and the data is locked in and the postback to networks is sent. 

The problem here is that advertisers often require more than a 24 hour period to understand post-install user activity to derive the value of the users. Workarounds such as timer extensions do exist, but these too come with their own limitations.

There is therefore a tradeoff between time and data. If you keep the timer limit to 24 hours then you lose out on potentially valuable post-install data. But, if you extend the timer then you lose your ability to react and limit your ability to optimize campaigns effectively. 

SKAdNetwork timer limit

Reimagining remarketing

In iOS14, many users will opt out of ad tracking, which means they won’t have an IDFA. Without this identifier, remarketing becomes far more complex and challenging. 

App owners will need to use other identifiers to encourage users to opt-in, such as requesting an email address or another form of identification. These identifiers are considered first-party data and therefore adhere to the new privacy standards.

Using first party data collected via your owned media (website, emails etc) helps to “extend” the user experience and ensure the relationship between customer and app is maintained..

Each of the above limitations has a workaround so you can still glean insights, and add value. However, the combination of these factors means there is fewer data for advertisers and attribution partners to work with. 

So, while these limitations present challenges, they also offer the opportunity to innovate.

Help is on the way

Attribution in the age of privacy is the art of generating a full picture out of fragmented and limited data. With each piece we are able to build a more complete picture. Let’s look at some of the solutions which can help fill the gaps.


Incrementality measurement uses a control and test mechanism to help marketers understand the true value of their campaign with data-driven insights. In particular it helps to identify which cohort of users would not have converted without marketing efforts. 

In SKAdNetwork, where data is limited, incrementality can help fill the gap and add confidence to the success of a campaign, providing an important additional layer of intelligence.

This will be particularly relevant to remarketing campaigns where data in the SKAdNetwork data is at a minimum.

First party data

As mentioned above the use of first party data is permitted by Apple under the new privacy restrictions if done transparently and with full user consent. 

Marketers therefore have an opportunity to maximize the value gained from first party data and should double-down on the information they collect. 

This data can then be used for more direct-to-consumer engagement. For example, a delivery app could send a push notification with suggestions or promotions for dinner based on previous orders.

Machine learning and predictive analytics

By leveraging machine learning algorithms, digital marketers are able to understand trends and user behavior and apply these to predictions of how valuable a user may be over time. 

Because SKAdNetwork limits the quantities of data, marketers do not have the luxury of time to assess the efficacy of a campaign. Therefore the ability to be able to predict the success of a campaign early on, and optimize accordingly becomes an invaluable tool.  

Web-to-app flows

Web-to-app flows take the user from a webpage through to a corresponding app. The mobile web is an important first touchpoint as potential customers research and learn about a brand and what it can offer them. 

The ability to use this touchpoint and direct users to an app helps to create a positive and easy user/ app onboarding experience. Similar to first party data, web-to-app flows are able to connect the user journey without the use of the IDFA as they include ad networks and/ or owned digital properties, which make them a critical tool in a post iOS14 world.

web-to-app flow

The bottom line

App marketing in the age of privacy is more complicated and challenging.

That’s why having a trusted attribution partner is more important than ever to overcome the limitations of SKAdNetwork and offer alternative measurement solutions for a holistic picture. 

Michal Wagner

Michal is a Content Writer at AppsFlyer. She has extensive experience writing on subjects ranging from marketing, digital transformation, machine learning, and telecoms tech. Together these give her a wide perspective on matters relating to Mobile Attribution and Marketing Analytics. Michal's passion lies in taking a complex subject and making it easily accessible to the reader.

Follow Michal Wagner

Ready to start making good choices?