It’s no secret that after Apple introduced App Tracking Transparency (ATT), mobile app marketers significantly decreased spending on iOS campaigns. As we reported in our recent whitepaper, before ATT, two-thirds of total client spend was associated with iOS platforms but that changed dramatically as Android spend increased 361.5% while iOS decreased 45.2%. That being said, over a year into IDFA deprecation we are noticing changes — as of Q1 2022, just 13% of our clients are running Android-only campaigns.

If you are still a bit iOS-shy but know you need to get back to running Apple campaigns, you may be wondering what’s still working on iOS. So were we! Let’s explore what you need to know about iOS campaign success.

Our Director of Growth has prioritized testing SKAN, direct media buying, and DSPs, allowing our existing clients to fulfill their KPIs with trusted/consistent sources while diversifying their supply mix and protecting overall supply in a potentially unreliable iOS environment. So let’s explore the SKAN experiments we are running for clients.

SKAN experimentation

The good news is that because probabilistic matching is widely used, clients across all verticals are still spending on iOS. However, when it comes to testing SKAdNetwork (SKAN), few clients are willing to take on the challenge. Just four clients are testing the SKAN waters, representing social media apps, sports betting, and fintech. 

This inconsistent mix of verticals may seem like an unlikely pairing, but we do see a trend appearing. If an app relies on more complex down-funnel events, marketers are likely more reluctant to test SKAN since the only consistent event you can rely on is install and you likely have to completely change most of your in-app event KPIs.

For SKAN-only campaigns, we have to completely readjust the way we’re looking at events. One of our largest clients, for example, still wants to see a down-funnel event (such as a sign-up or registration) but is less strict with their day 7 retention KPIs due to some limitations of SKAN. Meanwhile, another large client has a more advanced type of SKAN conversion modeling where each value is mapped to a unique event flow.

FeedMob is still in the process of launching various campaigns, so we are eagerly awaiting results that tell us which values we start getting back and if that aligns with what we’re seeing on the MMP side (clients still have user-level probabilistic on).

How shifting conversions are changing pricing

Because of this shift in priority, there has been a huge shift to CPM instead of CPI/CPA. This was predicted by some of the most prophetic minds in our industry, but in FeedMob’s case, it’s mainly due to a shift in prioritizing different supply types (DSPs), which have a strong representation of partners who support and are becoming experienced with SKAN.

So far, we haven’t seen a huge change in conversion rates since user-level probabilistic is still allowed and being exercised by most of our clients. However, it’s been important during this time of uncertainty to make sure we’re prepared for what comes next. The mobile world is changing quickly and we’re continually prepared to change with it.

Preparing for a post-probabilistic future

When user-level probabilistic is enabled, goals and events continue to work more or less as they did prior to ATT for FeedMob — but the future of probabilistic is uncertain. When only aggregated advanced privacy is enabled, we start losing the ability to granularly track performance across specific goals. We can still track clicks, installs, and events, but we completely lose the ability to connect these to one another through a unique identifier (i.e. click ID).

If only SKAN is available, obtaining in-app events has been a huge challenge. There is a 24-hour daisy chain you need to adhere to and if, for example, you have a FinTech client like where one of the main KPIs is getting a direct deposit at 30 days, there’s just no way that’s happening. Install remains the only constant between SKAN and MMP campaigns.

Deterministic Attribution

SKAdNetwork Attribution
(PIE events: no 24-hour time-lapse)

SKAdNetwork Attribution
(PIE events: with 24-hour time-lapse)

Almost all the advertisers the FeedMob team is working with are using, at the very least, aggregated attribution and even more commonly, user-level probabilistic attribution. Until recently, if you chose not to run mobile UA on iOS, you were missing out on swaths of traffic your competitors are benefiting from at a privacy-level that Apple has, thus far, allowed.  However, it’s possible (and has been a subject of rumors) that Apple may find its own way to make probabilistic obsolete.

More recently, Apple announced new changes to SKAN that should make it possible for ad networks and developers to better measure ad performance while still preserving user privacy. Changes are afoot. From hierarchical source IDs aimed at increasing the ability to optimize campaigns in a privacy-friendly way to hierarchical conversion values that give developers more information about small campaigns (and help app marketers calculate ROAS by offering multiple conversions at defined time windows).

Experimenting with SKAN can only help ensure app marketers are prepared for a possible future SKAN-only scenario, and as Apple makes new features available, there has never been a better time to start testing.