The Signal You Can Not See
For a decade, mobile marketers knew exactly who installed, from where, and whether they paid. Then Apple flipped a switch. ATT did not just remove a field — it removed the certainty the whole acquisition machine was built on. Three years on, most teams are still guessing, and the ones doing well stopped guessing and started measuring differently.
Q: What actually changed after ATT, and why does it still hurt?
Before, you got a device-level IDFA and a clean path from ad to install to purchase. After a user opts out — and most do — that path goes dark. Apple's replacement, SKAdNetwork, returns only coarse, delayed, aggregated data with no user-level trail. Your dashboards did not just lose a column. They lost the ability to say which campaign actually caused the install. That silence is why UA teams still feel blind in 2026.
Q: Is SKAdNetwork enough, or do you need a better plan?
For blind optimization, no. SKAN tells you a campaign worked-ish, a day or two late, in broad buckets. It cannot tell you which creative, which segment, or which user. Serious teams now run three things on top: media mix modeling for the big picture, incrementality tests to prove a channel caused installs, and first-party data to close the loop on their own users. SKAN is the floor, not the strategy.
Q: What is first-party data, and why is it the real asset now?
It is everything you collect directly from a logged-in, known player: what they did, what they bought, what they did not. Because it lives on your side, no platform can take it away. The studios that pushed players to accounts and tracked server-side events own a measurement signal Apple cannot dim. The ones still leaning on ad-network panels are renting their eyes from someone who is slowly going blind.
Q: How do you measure UA spend when you cannot see the conversion?
You stop asking "which click" and start asking "did the region move." Incrementality is the honest version: hold out a small geo or audience, run the spend elsewhere, and compare. If the held-out group did not move and the targeted one did, you caused it. Media mix modeling does this across everything at once. Neither is as tidy as the old attribution, but both survive a world with no IDFA — and they are what the survivors actually use.
| Old world (IDFA) | New world (post-ATT) |
|---|---|
| User-level tracking | Aggregated, coarse data |
| Last-click attribution | Incrementality + MMM |
| Rent the signal | Own first-party data |
| Tidy dashboards | Honest ranges |
You will not get the old clarity back. The teams winning post-ATT are the ones who made peace with fuzzier numbers and built the systems to live inside them.