Fixing iOS attribution

Can you fix iOS attribution?

• How do you get true organics in iOS user acquisition?
• How do you manage measurement for networks doing SKAN and those who kinda aren’t?
• How do you separate out results by channel, owned, earned, and paid?
• And how do you bring together measurement data based on SKAN, ad partner modeled numbers, IDFA, first-party data, UTM-tracked link data, Apple Ads Attribution API, and other sources?

Plus, importantly, can you do all that without losing half your SKAN conversion value payload?

In short: is it possible to fix iOS attribution? And, perhaps, achieve something even roughly similar to pre-ATT days? Maybe … hit play and keep reading!

We all know the problem: multiple competing iOS attribution methods

By default, this isn’t happening right now on iOS.

“You want to see in a single place: just show me how many installs is every channel driving? What’s my ROAS for every channel?” says Evyatar Ram, Singular’s VP of product. “It’s all over the place: some of the data’s here, some of the data’s there, some of the data’s cohorted, some’s not cohorted, some’s available immediately, some’s available three days later.”

Summing up, he adds:

“It’s painful.”

The reality is that different channels are using different attribution methods, resulting in different answers to the same question. You could — and frequently do — have a single user and a single install that is attributed to different ad networks or even different channels by different mechanisms.

Unified measurement: fixing iOS attribution

The good news is that this problem is now largely solved, thanks to Unified Measurement.

Singular’s always had a competitive SKAN solution with SKAN Advanced Analytics. But while it was the best available option in the iOS attribution space, it didn’t solve every problem: it didn’t dedupe every single install from every source, and too much showed up as organic rather than a result of a paid campaign. There have been different attempts to solve this problem in the mobile attribution space (SSOT is one). But SSOT burns half your SKAN conversion values, and Singular was unwilling to sacrifice so much post-conversion engagement and revenue data to build what essentially still results in a partial solution. 

That’s too high a cost, especially because it falls victim to censored data due to Privacy Thresholds in SKAN 3 and Crowd Anonymity in SKAN 4, eroding valuable signal even more. 

The long-term solution is Unified Measurement that reliably combines SKAN data with MMP tracker data in a way that is de-duplicated and accurate, says Singular product marketing manager Kelsey Lee.

“You know the real number of installs that are driven from every channel and you know your actual number of installs … as well as revenue cohorts and also events.”

The solution has been live for months for Singular clients, and the results have been impressive, with clients seeing numbers like 31% improved accuracy, a 43% boost in campaign ROI thanks to better organic versus paid attribution, or a 19% reduction in double-counted installs. Exact results per client depend, of course, on specific media mix, ad partner selection, and scale, but the bottom line is that accuracy is up across the board.

Almost as important, Unified Measurement offers 35-day cohorts: much longer than the 7-day cohorts offered by some MMPs, and much more indicative of the true long-term value of marketing campaigns. 

35-day cohorts is something that just isn’t possible anywhere else.

The solution: UA measurement using the best from each attribution methodology

Unified Measurement takes the best from SKAN, which for all its faults is a true deterministic attribution method that is on-device and close to the user. Then it adds the best insight from multiple other forms of measurement, including first-party data, network data, MMP tracker data, and more to form the best possible iOS attribution solution.

An important part of that is allowing SKAN to retain as many bits as possible for your conversion data.

“Preserving what’s going on in your conversion model is ultimately going to make your reporting better at the end of the day,” Lee says. “And that means we’re also not subjected to any Privacy Thresholds or Crowd Anonymity … we don’t have to deal with censorship issues, because we’re not relying on the SKAN postbacks.”

The result is really good data for growth marketers, even if it’s not exactly a one-and-done silver bullet for iOS attribution for all time.

“It is the best solution, I think, that is available in the market,” Ram says. “But I think that it is an evolving space and I don’t think it’s sort of this one solution that you now will be set for the next 3 years. We have a lot of ideas of how to improve it.”

High on the list: media mix modeling, or MMM.

Also, as it gains more adoption, SKAN 4.

The same basic methodology will work for Privacy Sandbox on Android as well as any additional privacy initiatives that may occur.

“If you’re running iOS marketing campaigns, you can now operate effectively,” Ram says. “But we’ve also built a framework for the future which is going to basically future proof ourselves for other privacy changes down the road.”

More in the full episode

As usual, check out the full episode for more insight on the effective iOS attribution.

You can always watch Growth Masterminds episodes on our YouTube channel, or in your favorite podcast platform (Apple, Spotify).