App attribution on iOS in plain language: what you can still measure and how to decide

SKAdNetwork, ATT, postbacks: the words sound technical, but the underlying question is simple. What do you still know, what do you no longer know, and how do you make good decisions with less data?

App attribution on iOS comes down to one shift: you no longer know exactly which user came in through which ad, but you still know how many installs and early events a campaign produces. SKAdNetwork gives you aggregated data with a delay, and that is enough to steer well, provided you redesign your decision process around it instead of mourning the old precision.

What exactly changed with ATT and iOS 14?

An identifier used to travel with every install, the IDFA, which let platforms tie an install to an exact ad and user. With App Tracking Transparency, Apple flipped that logic: apps must explicitly ask permission to track, and a large share of users says no. For that group, the user level link between ad and install is gone. Not delayed, not fuzzier: gone.

For founders of apps and subscription businesses, that felt like flying blind. But blind is the wrong word. You went from a microscope to a pair of binoculars: you no longer see individual users, but the broad lines per campaign are still perfectly visible.

What is SKAdNetwork in plain language?

SKAdNetwork, since evolved under the name AdAttributionKit, is Apple's answer to how advertisers can measure without tracking users. It works like this: not the ad platform but Apple itself registers that an install came from a campaign, and then sends a postback, a small message with aggregated information, to the platform. It states which campaign caused the install plus a coarse conversion value that says something about what the user did in the first period.

• You get data per campaign, not per user: no profiles, no individual level customer journeys.
• Postbacks arrive with a delay, so your dashboards run days behind reality.
• The conversion value is limited: you choose upfront which early events to encode into it, and that choice determines what you can see later.

The most important decision sits in that last line. Because the conversion value can only carry a limited amount of information, you have to choose which events deserve the space: a trial start, a registration, completing onboarding. Pick the events that best predict whether someone becomes a paying user later.

What can you still measure on iOS?

More than most founders think. Installs per campaign remain visible. Your chosen early events come through via the conversion value. And inside your own app you still measure everything: activation, retention, trial to paid, revenue per cohort. The only thing missing is the hard link between that internal data and the individual ad. So the craft is laying both worlds side by side: the aggregated campaign data from the platform and your own product data, and checking whether they tell the same story.

On iOS you no longer measure users, you measure directions. If you can steer on directions, you have enough.

How do you decide with less data?

The biggest mistake is pretending nothing changed: optimizing per ad, daily, on numbers that arrive delayed and rounded. That is chasing noise. The better approach is deciding slower and coarser, but on more reliable signals. Judge campaigns at the weekly level instead of the daily level. Compare cohorts: what did the group that came in last week do, compared to the group before? And use your early events as a compass, because they arrive on time while revenue only becomes visible weeks later.

Also work backwards from your own data. If you know what share of your trials converts to paid on average and what a paying user is worth, you can set a maximum price per trial for each campaign. Then you steer on a signal that is fast and visible, with the certainty that the economics underneath hold up. That is not perfect attribution, it is something better: a decision model that can handle incomplete data.

What does this mean for your campaigns and creatives?

Less measurability makes the contents of your campaigns more important, not less. Keep your account structure simple, because every extra campaign fragments an already delayed signal. Consolidate budget into fewer campaigns so each one has enough volume to stay readable. And shift your testing to where you can still measure sharply: the creative. Hook rates, watch time and click behavior still come in real time, so creative tests remain your fastest learning instrument on iOS.

In practice we see that app businesses which accept this scale calmer and better than businesses trying to reconstruct their old dashboards. The data got coarser, but the game stayed the same: strong creatives, a clear offer and a funnel that carries people from install to paying user.

Conclusion

iOS attribution is not a mystery but a different measuring stick: aggregated, delayed and coarse, yet workable for anyone willing to steer on trends, cohorts and early events. The real challenge is adapting your decision rhythm to it and structuring your campaigns so every signal counts. That is exactly the kind of work we do as a paid social partner for app and subscription businesses: structure, signal and creative volume in one system. Not sure whether you are steering on the right numbers? Book a call and we will gladly take a look with you.