Singular CTO Eran Friedman presents SKAN 4 for Google for Startups

How do you know when your mobile app marketing campaigns are working? 

Traditionally this has been fairly easy for mobile marketers on iOS, using the IDFA (Identifier For Advertisers.) However, since iOS 14.5, the IDFA is only available upon request via Apple’s App Tracking Transparency modal, and only about 12% of iPhone owners say yes to Apple’s App Tracking Transparency.

That cuts the amount of data that marketers rely on to optimize ad campaigns by almost 90%. 

So Apple’s recently updated SKAdNetwork framework is now essentially the only deterministic way to measure mobile app marketing success. And that’s why it’s critical for app publishers to deeply understand and fully exploit all the data that Apple does provide for advertisers in the fourth major iteration of the company’s SKAdNetwork framework, SKAN 4.0.

“SKAdNetwork is a framework created by Apple for privacy-preserving mobile app install attribution,” Friedman said in a recent presentation for GameCamp via Google for Startups.

If you’re in the mobile advertising space, you probably know a ton about SKAdNetwork and App Tracking Transparency already. (See this post on SKAN 4.0 strategy to learn more about transitioning from SKANn 3.) If you’re not, however, Friedman explains the broad strokes:

1. Advertisers with apps they want to promote buy ad campaigns from an ad network
2. Ad networks buy space in app publishers’ apps for the ad. Each campaign needs a specific SKAN campaign ID.
3. The ads get shown to people using a publisher’s app
4. The person who sees the ad taps on it, installs the advertiser’s app, and launches it
5. Code inside the newly installed app watches for specific predetermined behavior in the app, and if it sees something like “completed level” or “subscribed” or “purchased,” updates a conversion value. This value is not tied to a specific person or even device, and can only be a number between 0 and 63. (That’s part of Apple’s “privacy-preserving” technology.)
6. After a randomized timer between 24 to 48 hours, the device sends a notification — a SKAN postback — to the ad network to let them know their ad campaign worked.
7. If and only the number of app installs from that campaign surpasses an Apple-set privacy threshold limit, postbacks will include conversion values.

(Note: some of these details are changing in the currently ongoing transition from SKAN 3 to SKAN 4; more details here on what those changes include.)

MMPs like Singular, of course, have been helping marketers make sense of all this. Up to this point under SKAN 3, they’ve been doing 4 key jobs:

1. Collecting all postbacks
2. Enriching them with campaign information
3. Decoding numerical conversion values to human-readable names
4. Modeling missing data, since about 21% of all SKAN postbacks don’t carry conversion values due to Apple’s privacy thresholds

The goal in the fourth step it to replace as much data as possible and emulate, as much as you can, the data that was fully visible before App Tracking Transparency, when 70-80% of iOS installs were measurable via IDFA (the Identifier for Advertisers that now requires App Tracking Transparency consent to acquire.)

“The end result that you want to get from SKAdNetwork is providing a report with the full analytics, end-to-end, of your best-performing campaigns,” says Friedman. “For a specific source, such as Google, you see how much you’ve spent, how many impressions or clicks, how many SKAdNetwork installs you received … without the need for consent, total installs across all users that came from this campaign. 

“From this you can calculate your CPI, your CPM, your CPC, you can even decode back your revenue coming back from your conversion values providing ROI on your campaigns.”

But the real magic, Friedman says, requires modeled data.

SKAN intentionally drops data due to privacy constraints. If Apple gave you everything, instantly, some would be able to correlate SKAN postbacks with specific devices and people. So privacy thresholds in SKAN 3 and crowd anonymity in SKAN 4 delete some of that data, and SKAN applies random timers to postbacks to obfuscate individual actions. The result is that SKAN data on its own has a major problem from a marketing perspective: campaigns can look both expensive and ineffective. Marketers looking to scale good campaigns and drop poor performers lack sufficient compelling data to make those vital calls. 

The solution: modeling SKAN data by enriching it with first-party data, thereby giving advertisers an accurate picture of results while not violating privacy.

That layer — SKAN Advanced Analytics — helps Singular clients achieve close to 90% D7 revenue accuracy: essential insight to optimize campaigns.

Of course, Apple is now updating SKAdNetwork to SKAN 4. 

The software is already released, but there’s a lot of work happening behind the scenes in the mobile adtech ecosystem to make it all work. Singular released a SKAN 4-compatible SDK in November of last year, but there’s a lot more work to be done by ad networks, advertisers, and app publishers.

SKAN 4 will provide richer data in the first postback advertisers can get — assuming high levels of installs per campaign providing significant crowd anonymity — and additional postbacks over time to get more insight into campaign performance over time. Singular’s using these additional signals to enhance the virtual cohorts SKAN Advanced Analytics already provides. The complete SKAN 4 version of SKAN Advanced Analytics, Friedman says, will add new elements to the 4 key jobs above, including long cohorted KPIs, extremely accurate data estimations thanks to SKAN 4’s additional data, and decoded SKAN 4 coarse conversion values.

Watch the video embedded above or on YouTube for a full explanation of both our current situation and the ongoing transition to SKAN 4.