April 26, 2021 will go down as one of the most important dates in the history of digital marketing.
On this day, Apple began enforcing its App Tracking Transparency (ATT) framework in iOS14.5, requiring apps to actively gain user consent for tracking via an opt-in mechanism, ushering in a new privacy era.
While it was certainly a welcomed move as far as consumer privacy was concerned, ATT created a massive challenge for mobile app marketers. Suddenly, data was mostly limited and aggregated, turning user level-driven measurement and optimization upside down.
How did ATT impact the number of downloads driven by marketing? Is remarketing still a viable marketing activity? How did budget allocation change, and what has been the effect on consumer spend in apps?
To address these questions, our annual Top 5 data trends report analyzed over 60 billion app installs in 2021, allowing us to predict what 2022 will bring as the industry continues to adapt to the new privacy paradigm.
Click on the following links and back to navigate in the post:
All data used in the report is measured by AppsFlyer, the marketing measurement and experience platform. This is done via integrations with leading media partners, as well as its proprietary SDK implemented in its clients’ apps. All data is anonymous and aggregated.
1) Installs: Marketing measurement capabilities largely retained following initial post iOS 14.5 plunge
Much has been said about how COVID-19 significantly accelerated digital transformation in 2020.
Adding to the digitalization trend, our data shows that a new normal has also been created in the world of mobile apps, with the entire install and usage baseline elevated in 2021, as we can see in the following chart.
A year-over-year comparison has 2021 installs exceeding 2020 by 19%. This is quite a feat considering 2020 with its lockdowns and social distancing policies that generated a 33% leap compared to 2019.
Enhanced digital consumption of mobile devices in 2021 was a great starting point for the mobile app space.
But on April 26th, Apple’s iOS 14.5 changed how the [marketing] game was played.
As seen above, overall app installs climbed, which can be easily measured. The problem of magnitude was, initially, the difficulty to measure non-organic installs (NOIs) and dedupe data from multiple sources:
- SKAdNetwork (SKAN): Attribution performed on the device itself by Apple’s privacy-centric attribution mechanism.
And traditional attribution performed for:
- ATT-consented users: Attribution based on ID matching. It’s worth mentioning that although consent rates were higher than anticipated, the majority of users could not be attributed with ID matching as it requires dual consent (more on that further down).
- Non-consented users: Attribution based on AppsFlyer’s aggregated models – Aggregated Advanced Privacy framework for paid media or probabilistic modeling for owned media.
- Older versions users: Those who have yet to upgrade to iOS 14.5+.
To dedupe, we used the highest value of NOI per media source and campaign in each measured week, which explains why the overall figure is lower than the sum of traditional and SKAdNetwork data.
Initially, traditional attribution took a hit, dropping 25% by June 21st and then a further 5% by mid September. On the SKAN side things weren’t looking better early on either, with slow adoption and implementation among advertisers and media companies alike.
While SKAN traffic started climbing in July, traditional attribution remained low. Over time, however, innovation started kicking in as we invested significant resources in alternative and privacy compliant measurement.
Clearly, winning credit for NOIs this year has been tough for iOS marketers, with an overall 5% drop (while Android NOIs increased 15%).
In the last couple of months, traditional attribution picked up speed, rising by 17%. With the continued adoption of SKAN, measurement capabilities have been largely retained with the number of attributed NOIs nearing pre-iOS 14.5 levels.
Gaming adopts SKAN but takes a hit due to its reliance on marketing
On the vertical front, we can see that Gaming apps, known for their data savviness and agility, have adopted SKAN much faster than non-gaming apps with 67% of NOIs vs a mere 30%.
But a view from above shows that data limitations and measurement shifts have taken a toll on Gaming apps. Based on analysis of over 100 billion installs since January 2020, we can see a 6% YoY drop in overall Gaming app installs in 2021, while at the same time non-gaming apps enjoyed a 25% jump.
Because of the massive competition games face and the fact that branding plays a much smaller role in this vertical, Gaming is far more reliant on marketing. In fact, Gaming app marketers have often treated non-organic users as of higher quality than organic users because they excel at user acquisition that’s based on user-level signals.
However, reliance on user level data also means that they are more exposed to changes in how data is measured and optimized. With the shift towards aggregate measurement, overall install numbers in Gaming have dropped – mostly the result of a decline in marketing-driven installs.
What’s in store for 2022
As the industry continues to adapt to the new privacy norm, the ability to measure, attribute, and optimize marketing activities will continue to improve on iOS (and on Android when Google introduces its own privacy measures).
Improvements will be driven by better and more sophisticated models, increased usage of predictive analytics, acquired expertise in SKAN – particularly optimizing conversion values (see further down for more on the topic), and all around innovation across the ecosystem.
We can only hope that Apple will add more privacy compliant data to SKAdNetwork, particularly much-needed organic data, GEO level information, deferred deep linking, and data from web-to-app and remarketing campaigns.
2) Budgets: Apps spent $78-$83 billion on user acquisition in 2021 – up 40% YoY
User acquisition budgets in 2021 will reach $78-$83 billion, with the range driven primarily by budget estimations in China (see end of article for more on our methodology). The figure represents a 40% YoY increase driven by a 50% jump in Android and 26% rise in iOS.
While spend increased on both platforms, the reason behind the growth in each OS was entirely different.
As explained above, iOS saw fewer NOIs this year (-5%). At the same time, iOS also experienced a significant rise in effective cost per install, which drove budgets upward. Ad prices increased by 20% to as much as 50% across nearly all categories since the enforcement of iOS 14.5!
What this means is that marketers were able to acquire and get credit for far fewer users for the same budget they’ve invested in 2020, or – had to increase their spend for the same amount of users.
Why did media become so expensive on iOS?
For starters – basic supply and demand. The prices for a much lower supply of consented users leaped as demand for users with full data granularity has skyrocketed for obvious reasons.
But it appears that the main cause for this jump in cost has been the inefficiency of major media sources during a time of transition to target the right people and estimate the right impact.
It’s harder to make small segments of high-value users, so marketers need to use broader targeting, which also makes it less relevant.
On top of that, networks can’t optimize toward in-app signals yet, although this capability is being developed with solutions such as Google’s GBRAID and Facebook’s AEM; until they do, their targeting and optimization abilities are negatively affected.
In parallel, demand for Android with its full data granularity surged with NOIs rising over 40% YoY, especially in Gaming where NOIs leaped 50% compared to non-gaming at a still impressive 36%. Interestingly, global Android CPI in Gaming was down 11% this year (it did increase 23% in North America, but dropped 24% and 15% in Southeast Asia and Japan & Korea, respectively), while non-gaming CPI increased by almost 20%.
As shown in our latest Performance Index that includes the first SKAN ranking, budgets have not only shifted to Android, they’ve also shifted among media companies. TikTok Ads came out on top of the SKAN Index while Facebook ranked 2nd.
The social giant also lost its dominant position in iOS among consenting users to Apple Search Ads (ASA). The fact that ASA has its own API and is able to maintain full data granularity has attracted more budgets to this platform.
What’s in store for 2022
User acquisition budgets will continue to rise with more NOIs on both platforms. However, the cost of media won’t increase as much in iOS this year given the estimations that the inefficiencies on the network side will be largely solved.
Total industry UA spend methodology
The formula combines three buckets:
- Budgets measured by one of the major mobile measurement partners (industry wide extrapolation was made based on 3rd party estimates)
- The non-attributed market or marketing driven installs that were not measured by one of the major MMPs (about 10% according to estimates)
- Spend in China where measurement is a far greater challenge due do its fragmented app store space; we used App Annie’s total of 96 billion installs in 2020, AppsFlyer’s share of NOI and CPI, and eMarketer’s estimate of China’s 27% share in global mobile ad spend.
3) Re-engagement: Paid remarketing innovation drives iOS rebound while owned media rises by 45%
After a steady climb in recent years, paid remarketing efforts by mobile apps plummeted once iOS14.5 was released. This was expected – without IDFA, targeting and personalization, which are the basis of remarketing, cannot work.
From March to July, the number of remarketing conversions plunged by almost 35%, despite the fact that adoption of version 14.5 was slow (only 27% in June).
But then, as seen in the NOI trend above, innovation kicked in – this time on Google’s side. The search giant introduced GBRAID – an aggregate and privacy compliant identifier designed to optimize re-engagement campaigns.
Because Google is the largest player in terms of sheer volume within the remarketing pie, the fact that it introduced a new way to measure single handedly increased the number of remarketing conversions by 12% from July to September, and another 14% in October as the holiday season kicked off (last year, iOS remarketing increased almost 10% in October vs. September).
As with NOIs, Android capitalized on iOS challenges, increasing its number of attributed remarketing conversions throughout the entire year by a total of over 45%.
Owned media on the rise as an alternative
With the challenges of remarketing in iOS, it is no surprise that the use of owned media to re-engage existing users through push notifications, email, and in-app messages has jumped almost 45% since April (compared to only 17% lift in Android).
Furthermore, owned media gives apps full data granularity within a 1st party data environment (privacy measures seek to prevent the passage of data to 3rd parties, not within a company’s owned properties).
More and more apps are using integrated audience segmentation and marketing automation tools to help app marketers efficiently re-engage users. This allows marketers to create tailored messaging, and even exclude users who have already responded to a campaign from seeing similar ads across other owned channels.
What’s in store for 2022
Re-engagement will continue to drive increased usage, both on the owned media 1st party data side, and the paid remarketing front – where conversions will continue to rebound as other networks are already working on solutions.
4) Revenue: Gaming reliance on marketing highlights its monetization challenges in the privacy era
An analysis of total consumer spend (IAP or in-app purchase revenue) in apps shows just how challenging iOS14.5 has been for Gaming apps.
As we only compare apps that were live throughout the entire measured period, revenue in Gaming usually doesn’t increase by much as many games have a limited shelf life.
However, a comparison between platforms has Gaming IAP plunging 38% between April and September compared to a minor 13% drop in Android.
As discussed in trend #1, unlike non-gaming apps, games are heavily reliant on marketing and their ability to optimize based on user-level data signals. For most users, targeting and optimization have been affected in their absence, leading to a drop in overall Gaming revenue from IAP.
The same trend is seen in IAA (in-app advertising) revenue on the publishing front, which is the other side of the Gaming UA coin. Between April and September, iOS revenue dropped by 15%, while Android revenue jumped 45% during the same time frame.
Similar platform trends were seen among non-gaming apps that are driven mainly by organic traffic. We can see an increase in May, and then a relatively flat revenue trendline.
With October marking the start of the Q4 holiday shopping season, and November in full speed with Singles Day, Black November, and Black Friday sales, we can see the impact of Shopping – the largest non-gaming category – on the overall trend.
What’s in store for 2022
Gaming apps will find a way to adjust to the available data and measurement methodologies. They have always been the fastest learners and are certainly the best at data-driven optimization.
This time around, however, things have been more difficult as their reliance on user-level data, and also the fact that the networks they rely on are also in the midst of their own learning curve.
5) iOS 14.5+ deep dive: ATT opt-in rates & SKAN conversion value mapping
The ATT prompt has single handedly changed an entire industry and created quite a dilemma for mobile apps. Because it’s an opt-in mechanism and a simple tool to gain user consent, apps can choose not to show it. In such a case, app marketers would settle for SKAdNetwork attribution and probabilistic modeling.
The main reason why some apps chose not to implement ATT is out of concern that the prompt with its somewhat off putting language can drive churn and disruption to the user experience.
But the reality is that the benefits of showing the prompt far outweigh the benefits of not showing it.
Supporting this claim, we can see that six months after the release of iOS14.5, almost 65% of apps implemented ATT and we expect this number to increase to 70%-75% next year.
Why do most apps show the prompt?
Well, we’ve seen ATT opt-in rates consistently reach 46% during the last few months. This means that in nearly 1 in every 2 instances when a user actually sees a prompt, the ‘Allow’ button is tapped.
From a pure user experience perspective, that number is quite promising and shows that many users are willing to accept tracking in return for a more personalized experience.
Although there’s certainly plenty that can be done to increase opt-in rates, when we include restricted users and users who enabled Limit Ad Tracking (LAT), the overall rate shrinks as Apple does not allow the prompt to be shown to these segments.
More importantly, since it takes two to tango, ID matching can only work if there is dual consent on both the publisher and advertiser side, which means IDFA attribution rates are much lower.
The bottom line is that IDFA is still here, and although it’s only present in a much smaller cohort of users, this group is highly valuable for benchmarking, modeling, and extrapolations on non-consenting audiences.
Conversion value mappings show gaps between Gaming and non-gaming
The main area marketers have been struggling with this year was figuring out how to work with Apple’s SKAdNetwork conversion value mechanism.
More specifically, how to make the most of limited data; because it is the only post-install data that can be connected to their campaigns. Despite limitations, it is paramount to get it right.
What we can learn by analyzing data from AppsFlyer’s Conversion Studio is that Gaming apps are laser-focused on revenue, and as such – it’s a model that’s being involved in most conversion value schemes.
Their most common configuration is Events & Revenue with 36%-48%, with the exception of Hyper Casual apps that prefer events only configurations (38%). A Revenue-only configuration is also very popular among gaming apps, with 18%-35% opting for this schema.
On the non-gaming side, Events-only is the most configured option, with Shopping at 40%, Finance at 56%, Health & Fitness at 58%, and Entertainment at 57%.
What’s in store for 2022
Opt-in rates will continue to improve as users will come to grips with poor user experience they get from untargeted ads, while brands will get better at optimizing their prompts and providing a clear value exchange.
Still not entirely convinced? A new report by Gartner predicts that opt-in rates will increase by over 150% by 2023.
As for conversion values, with time, knowledge, testing, and training, marketers will fine tune their mapping and schemes to make the most of the data that is available to them.
To sum up
2021 privacy changes have been great for end-users and the important cause of safeguarding their data, yet not-so-great for marketers.
But – things are looking up as we head into 2022 and beyond:
- Continued learnings of SKAdNetwork, new Apple features, ad network solutions and innovation – will maximize SKAN’s value
- Opt-in rate optimization helps leverage device IDs from consenting users – where applicable – for the purpose of modeling and benchmarking
- Privacy-preserving innovation will retain measurability (e.g. machine learning models, incrementality etc.)
- Privacy-preserving data collaboration within the ecosystem based on Data Clean Room technologies will introduce a new and improved way of enabling marketing activities that will check all boxes:
- User-level driven insights in aggregated format for optimization
- Enhanced user experience
- Safeguarded user privacy
BONUS: YoY growth per category and country shows Finance is on tear
A category comparison shows us yet another angle of just how different iOS and Android are in 2021.
Changes in overall demand for apps varied significantly between the two platforms in almost every category, especially Dating, News, Shopping, Finance, Health & Fitness, and Gaming.
In terms of pure growth, as we’ve shown in our annual State of App Finance Marketing report, Finance is on a tear with the highest rate in iOS and 2nd highest on Android.
Fintechs are everywhere, helping people to manage their personal finances, invest in stocks, manage digital wallets, payment methods, cryptos, leading a global revolution on the relationship between people and their money.
Social and Shopping have also shown impressive growth, while Gaming lagged behind because of losses on iOS (see above).
Developing markets drive growth
A look at the world map shows many hyper growth markets with a 30% growth rate or higher in the developing world across Latin America (Peru, Mexico, Colombia, Brazil), the Middle East (Iraq, Algeria, Egypt), Africa (Nigeria, South Africa), the Indian subcontinent (Bangladesh, Pakistan, Nepal) and Southeast Asia (Philippines, Indonesia).
Growth in the US, on the flip side, lagged behind compared to other markets, but still managed to drive increased demand for apps (+10% YoY), while installs in another mega market – India – grew by 14%.