Apple’s privacy changes in 2021 upended the mobile app advertising market.
With the release of AppTrackingTransparency (ATT), developers could only use IDFA, Apple’s mobile ad ID, for ad targeting and measurement if users give permission.
But iOS user opt-in rates hover around 29%, according to mobile measurement platform Adjust.
Although gaming apps, specifically, fare a bit better at 36%, the low overall opt-in rate makes it much more difficult for app marketers to track behavior and target new users.
ATT’s impact has been especially hard on game app developers, which depend on ads placed in other mobile games, within apps and alongside search results in app stores to grow their user bases.
According to a survey of mobile app devs that use Liftoff’s marketing and retargeting platform, 64% said that ATT has had a negative impact on their user-acquisition campaigns.
Rising costs, lower revenues
Customer acquisition costs (CAC) are also on the rise.
The average cost per install on iOS has grown to $3.80 as of Q4 2022, according to AppsFlyer, up 88% from Q1 2021, the last quarter before the rollout of ATT.
Impulse, a mobile app that features brain training games, saw its CAC increase by 40% in the first few months after ATT went into effect, said CMO Dariia Opanasiuk, while its in-app purchases fell short of pre-ATT benchmarks.
As a result, Impulse “experienced a significant drop in revenue” in ATT’s immediate aftermath, Opanasiuk said.
To recoup lost revenue, app marketers have had to experiment with new marketing channels and buying approaches.
Before ATT, Facebook was “the Holy Grail of user acquisition,” said Dom Davies, an independent mobile gaming consultant and former senior business development manager at AppLovin.
“Facebook is now the smallest channel for all of my providers,” Davies said, “because buying effectively on Facebook is extremely expensive.”
Fixing broken modeling
To adapt, app marketers have been using probabilistic data and methods to reach opted-out users.
Still, replacing a deterministic signal like the IDFA, which can be easily tied to a specific user, with probabilistic signals, which cannot be tied to a specific user, has made targeting and attribution much less efficient.
And ATT wrecked the models that many developers use to determine which users yield the highest return on ad spend (ROAS) or have the highest potential lifetime value (LTV).
“The period when ATT went into effect was almost the only time we were forced to [scale back] our plans for expected LTV and ROAS,” Opanasiuk said, noting that it was only a two or three month period.
Post-ATT, marketers had to adjust to building their predictive modeling using aggregated attribution data provided through Apple’s SKAdNetwork (SKAN). But the early versions of SKAN failed to deliver attribution data quickly enough for marketers to pivot their campaigns in response.
“Previously, with IDFA, we could see the effectiveness of each creative we tested very quickly,” Opanasiuk said. “With [SKAN], the data became aggregated and began to come with a delay.”
Plus, the variability of SKAN’s reporting left a lot to be desired in terms of actionable attribution data.
Some platforms provide this data at the ad set level or “even worse” at the campaign level, Opanasiuk said, which isn’t specific enough for marketers to support sustainable growth.
“The key task for us,” she said, “was to create a marketing analytics system and implement alternative methods of traffic acquisition, such as Conversion APIs at different platforms.”
SKAN4 plans
Impulse’s monetization struggles persisted until it was able to figure out how to achieve positive CAC margins by running SKAN-compliant campaigns, Opanasiuk said.
The emergence of attribution alternatives like Facebook’s Conversions API helped somewhat by allowing developers to collect data about conversion events (like app installs) directly from a platform’s ad servers. Without access to a device ID, it’s much harder to use third-party tracking cookies like Facebook’s pixel to track impressions served to specific users and subsequent conversion events.
With fewer users sharing their IDFA, SKAN is the only immediate alternative for tying ads to app installs on iOS, which offers a lucrative audience pool. Mobile advertisers generally consider iOS users to have the highest potential LTV.
Although early versions of SKAN didn’t deliver attribution data quickly enough, the latest version, SKAN4, has taken steps to correct the delay in reporting by offering three postback timing windows that can be adjusted by marketers based on campaign criteria.
Getting multiple postbacks when users install an app allows marketers to gather valuable post-install information about user behavior, such as if they make in-app purchases. The ability to determine which ad impressions led to app installs and conversions means marketers can do more effective LTV modeling.
SKAN4 also supports more granular attribution data, including the ability to convey which campaign creatives led to an install or an in-app purchase.
More detailed postback reporting allowed Impulse to build more effective predictive modeling, bringing its CAC margins back to acceptable levels, Opanasiuk said.
But while helpful, getting more data points within each postback has also forced app marketers to rely more on their programmatic tech partners to make sense of these new aggregated targeting and attribution signals.
“Now we buy a lot through our programmatic direct partners and a few DSPs who can support MMP [mobile measurement platform] postbacks,” Davies said.
That means DSPs rely more on their partners to include the most important signals for customer acquisition within SKAN’s postbacks.
“We’re running SKAN campaigns, but we’re reliant on what signals we get from the advertiser integration, be that SKAN or the MMP,” said Levi Matkins, CEO at LifeStreet, which operates a DSP for mobile apps.
More contextual, more ad volume
Because ATT and SKAN have made user-level targeting on iOS much more complicated, some game app marketers are doing more contextual advertising. But making contextual work on mobile comes with its own set of problems.
“The IAB contextual categories that we get on mobile apps are completely unhelpful,” Matkins said.
So LifeStreet has invested in building its own contextual modeling.
“We’ve spent a fair amount of time building various app mapping models that help us understand similarities between apps and the themes that they have in common,” said Luke Matkins, CTO at LifeStreet (and Levi’s brother).
For example, if a marketer is having success advertising within a Candy Crush-esque “match three” game, LifeStreet’s modeling will analyze App Store preview screenshots to find similar games by looking for graphical elements that are common to that genre.
And since ATT made it harder for app marketers to target the highest-value impressions to maximize their ROAS, some are instead casting a wider net for their customer acquisition campaigns by serving more impressions.
But just because direct marketing has gotten less efficient “doesn’t necessarily mean return on ad spend is worse,” Levi Matkins said. The same can’t be said for the user experience.
“Typically, we are buying more impressions at lower prices to drive similar or slightly better results,” he said. “Unfortunately, people are probably seeing more ads on their phones right now when they’re interacting with apps to offset that signal loss.”
