Measurement

What SKAdNetwork 4 changes in postbacks and conversion values

Apple announced SKAdNetwork 4 at WWDC with up to three postbacks, a four-digit source identifier and coarse conversion values. Here is what is confirmed, what is not, and what to plan now.

4 min read

At WWDC this week, Apple announced SKAdNetwork 4, the next version of its install attribution framework for iOS. Release is expected later in 2022 alongside the new operating system, and full documentation had not been published when the first write-ups appeared. Treat the details below as the announced design and expect some to be refined.

Nothing here changes how you run campaigns today. It does change what you should be designing for, and it gives you a few months to do that.

Three postbacks, but only the first is detailed

SKAdNetwork today sends one postback per install. Version 4 can send up to three, tied to conversion windows of roughly 0 to 2 days, 3 to 7 days and 8 to 35 days after install. Only the first postback carries the fine-grained conversion value, the six-bit number you already use. The second and third carry a coarse value.

The practical meaning: you can finally see some signal about what users did in the second week and the first month, which matters for any app where the valuable action is a trial conversion or a renewal rather than a day-one event. But the later signal is blunt. It won't be a second fine-grained value.

Coarse values and crowd anonymity

The new coarse value takes one of three settings, low, medium or high. What you receive depends on how many installs a campaign generates, which Apple calls crowd anonymity. As described at the announcement, there are three levels in place of the single privacy threshold used today, and the data you get gets richer as volume rises. When installs are too few for a fine-grained value, a coarse value still provides some signal.

That should change how you design conversion values. With only a handful of installs per campaign, you may only ever see low, medium or high. Decide now what each of those should mean for your business, for instance:

  • Low: installed but no meaningful engagement.
  • Medium: completed onboarding or started a trial.
  • High: made a purchase or reached your core retention action.

Those are examples, not a recommendation to adopt. The point is that the coarse mapping should be defined deliberately and should line up with how you value users.

A four-digit source identifier replaces the campaign ID

Today the campaign identifier is two digits, which limits you to 100 values. Version 4 replaces it with a source identifier of up to four digits. Developers can treat it as hierarchical, for example campaign, then location, then placement. The number of digits returned depends on privacy thresholds: at low volume you receive fewer digits, and the full four appear only as install volume passes the privacy thresholds.

This is good news for planning, with one catch. A structure that puts the most important grouping in the leading digits means you will always get that grouping, even when volume is low. Put the dimension you most need first, which is typically the campaign or channel, and the finer detail last. Your ad networks and attribution provider will define how they use these digits, so ask them before you design anything.

Web ads can be attributed to installs

Apple also described attribution for ads on websites that link directly to an app's App Store page. That gives advertisers using web placements a route to attribution that did not exist before. How well it works in practice will depend on implementation details that Apple had not fully documented.

What to do now

  1. Do not rebuild yet. Version 4 is not retroactive to earlier iOS versions, so SKAdNetwork as it exists will keep running alongside it for some time.
  2. Draft a conversion value map for both modes. One for the fine-grained first window and one for the coarse values. Keep them aligned so a user's classification is consistent.
  3. Draft a source identifier plan. Decide which dimension goes first and how many campaigns you really need.
  4. Ask your attribution provider and networks for their timelines and whether they will support the new postbacks at launch.
  5. Plan to test. Apple says postbacks can be tested in Xcode, which is the safest place to confirm your mapping behaves as designed.

The main takeaway is that version 4 gives you more windows into user behavior, not sharper ones. Spend the summer on a clean mapping from your revenue events to low, medium and high, and a source identifier plan that survives low volume.

Sources

The Lingviny brief

Platform changes and what to do about them, for people who buy Apple, Meta and Google traffic.