At last year’s WWDC, Apple dropped the news that developers would have to obtain an explicit user’s consent in order to use IDFA. Since then, UA Managers were obsessing about the effectiveness of their campaigns in this “new normal” while Ad Monetization Managers obsessed about eCPMs and potential drop that would follow. The concerns were only amplified after Facebook published a white paper that claimed that the effectiveness of their Audience Network algorithms depended so much on personalization that without it, publishers lost 50% of their ad revenue.

Another question that arose in the past six months and was extensively debated in the industry was what % of users would consent to tracking. The bets varied from 10%  to 90% for consent. After all the discussions and some tests, it seems that many now agree that the majority of the users (70% or more) will not allow IDFA.

Given the above, I thought it would be useful to share some data, insights and practices from my experience of managing ad monetization stacks for many of the companies out there. Here are some questions to be answered:

• What % of users are already IDFA-less?
• What are the eCPMs for IDFA and IDFA-less users?
• Is the eCPM for no-IDFA users increasing?
• Is the gap in eCPM between IDFA-less and IDFA users getting smaller over time?
• What are the actions to be taken now?
• What are the actions to be taken once Apple starts enforcing ATT?

I’ll share findings based on my work on three different f2p mobile games. The scale of the games is significant (100K+ DAUs). Ad revenue is meaningful, although they are generating revenue primarily through IAPs. The United States is the #1 market, and users mostly brought through paid UA activity. All games are using rewarded videos and interstitial ads.

What % of users are already IDFA-less?

Although ATT is not enforced, users have the option to restrict developers from accessing their IDFA by following this path: Settings >> Privacy >> Advertising >> Turn on Limit Ad Tracking. I checked these numbers at the beginning of November and in the beginning of February. Here’s what I’ve got for US only:

Two things are worth highlighting here: first, the number of IDFA-less users is very different depending on the game. Since November, all three games have seen some increase in % of IDFA-less users but the biggest growth comes with the game that had the lowest number to begin with. Second, keep in mind that this is still without the ATT prompt. Another thing worth mentioning is that, although these numbers are on average higher for the US-based users, the differences (compared to global user base) are not dramatic.

What are the eCPMs for IDFA and IDFA-less users?

Based on the findings #1, I decided to focus on Game #1 and Game #2. These games are using MAX mediation by Applovin which conveniently announced some IDFA-related functionalities early in September. This enabled me to easily check eCPMs depending on IDFA.

The data shows that eCPMs are much lower for IDFA-less users. The difference in eCPM is dramatic, almost 50% on rewarded video and around 40% on interstitials.

Is the eCPM for IDFA-less users increasing?

In order to answer this question, I’ve compared the eCPMs for both IDFA and IDFA-less users from beginning of November last year and beginning of February this year. Of course, the comparison is not ideal (seasonality effect, UA activity, etc.) but it still might be some indication of a trend.

The simple answer is yes, eCPMs for IDFA-less users did increase from the beginning of November until now. However, we also see that the eCPMs of IDFA users also increased. Taking a closer look, it does seem that increase on IDFA-less is bigger than on IDFA, as the table below shows:

In the case of these two games, it’s true that eCPMs for IDFA-less users have on average increased more than eCPMs for IDFA users. Also, Fyber published some data based on their marketplace and claims that eCPMs for IDFA-less traffic is increasing.

Is the gap in eCPM between IDFA-less and IDFA users getting smaller over time?

Based on the data shown in #3, it appears that the gap is somewhat smaller (beginning of February VS. beginning of November). In November, eCPM of IDFA-less users was equal to 56% of IDFA users. Beginning of February, it was 63%.

What are the actions to be taken now?

There are some things that I would consider mandatory and that I would recommend to all my colleagues to put on their checklist:

• Updating apps to Run on Xcode 12 with AppTrackingTransparency
• Updating the network’s SDKs and adapters for iOS. Here are some minimum-required version for commonly used networks:
• Facebook Audience Network 6.2.1
• AdMob 7.64.0
• UnityAds 3.5.1
• ironSource 7.1.0
• Applovin 6.13.4
• Vungle 6.8.0
• Fyber 4.3.12
• Tapjoy 12.7.1
• AdColony 4.4.0
• Include all your networks’ IDs into your info.plist file. ironSource compiled a convenient list of pages that contain the IDs for each of the commonly used networks. Pay attention that Apple updated its guidelines around these IDs and now it’s required for all the letters to be in lowercase. As of now, all networks except Vungle have followed this requirement. For those using Fyber marketplace, they created a useful generator for the IDs.

Apart from this, an additional step that can be taken is separate mediation optimization (depending on the IDFA status). It doesn’t necessarily mean a huge amount of time for you and it might be good for the performance. Here are some tips from my experience:

• If you are using Facebook Audience Network and AdColony, turn them off for IDFA-less users because they are not serving ads to them (and if you are not using bidding, this can mean removing as many as 10 instances from your waterfalls)
• Increase the number of instances for other networks
• Consider lowering the floors (especially for AdMob)
• Here’s one that I was personally surprised to see:
• In my IDFA waterfalls, ironSource and Vungle networks were bringing on average, around 8% of total ad revenue each (rewarded and interstitials combined)
• In my IDFA-less waterfalls, I noticed these networks work particularly well. After some optimization, ironSource and Vungle are taking now ~34% of total ad revenue EACH (rewarded and interstitials combined)

There are also some more general things that can be considered for potentially softening the blow we will suffer once ATT goes into full swing. These require additional effort on all sides (product, ad monetization, development):

• Developers that have only rewarded video might want to consider adding interstitial ads as a valid option.
• For those using both rewarded and interstitials and have strong economies, offer wall or even rewarded surveys might be a thing to consider.
• Those more driven by ad revenues, increasing the frequency and caps for their current ads, integrating banners, native ads or more experimental formats (such as audio ads or AdMob’s rewarded interstitials or app-open ads) might be the way to open up for more ad revenues.
• Consider adding ad networks with non-performance demand (more focused on brands and awareness campaigns in general)

What are the actions to be taken once Apple starts enforcing ATT?

• Keeping an eye on the metrics is a good starting point. Those who haven’t separated waterfalls per IDFA before might want to do it at that point, depending on what they see in the data.

• If the eCPMs start going dramatically down, adding a new network into the mix might increase competitiveness.
• The same goes for fill rates – if they start suffering it’s an indication that a new network might bring some benefit and reduce that opportunity cost.
• Regularly be in touch with your Account Managers in order to be in the loop of the latest improvements, SDK updates and tips and tricks on how to improve your IDFA-less performance. Some mediations, such as MAX by Applovin, LevelPlay by ironSource and MoPub by Twitter have introduced options to run different waterfalls based on IDFA and moving forward we can probably expect more features going into that direction.

That’s it. We are now weeks away from “early Spring”. Let’s see what happens next and if this time next year we will be discussing the performance of GAID and GAID-less users.