Measurement
iOS ad attribution after ATT: what you can actually measure
ATT did not make iOS attribution disappear. It made measurement less granular and more dependent on aggregated signals. A sound setup does not try to identify every user. It connects permitted platform data with product events and supports decisions at campaign and cohort level.
Why the old report is no longer a single source of truth
Marketers once expected a nearly direct line from click to purchase. Today, many users decline tracking, postbacks may be delayed, and available detail depends on volume and privacy rules. Ad platforms, MMPs, and product analytics will naturally report different totals.
A difference is not automatically an error. Each system answers a different question and may use a different attribution window. Trouble starts when a team compares totals without agreeing on definitions first.
What ATT and AdAttributionKit do
AppTrackingTransparency governs access to identifiers and tracking across apps and websites. The permission prompt should explain the purpose in plain language. A refusal cannot be bypassed with fingerprinting or another hidden matching method.
AdAttributionKit lets ad networks receive cryptographically signed, privacy-preserving conversion postbacks. It supports install and re-engagement attribution, but it is not a user-level activity log. Lower-volume campaigns may receive less detail.
The four layers of a useful setup
- Product: registration, trial, purchase, renewal, refund, and revenue.
- Store: installs, sales, and subscription data from App Store Connect.
- Attribution: permitted device-level signals and aggregated postbacks.
- Ad platforms: spend, impressions, clicks, and modeled conversions.
Keep the raw figures separate, then build an agreed management view. Forcing the systems to match creates false confidence rather than better measurement.
How to judge campaigns with incomplete data
Build cohorts by install date, country, channel, and broad campaign group. Read event cost alongside changes in organic acquisition and total revenue. For larger decisions, geo experiments or holdout groups can help when volume is high enough to reveal a meaningful difference.
Modeled conversions can guide bidding, but financial forecasts should be reconciled with payments and renewals recorded by the product. Track the gap between systems as a metric of its own.
Common mistakes
- Treating an ad platform report as exact financial accounting.
- Changing attribution windows mid-test and comparing the periods directly.
- Optimizing for an event that arrives rarely or unreliably.
- Mixing new users with returning users.
- Looking for an ATT workaround instead of improving experiments.
The costliest issue often sits upstream of the report. A purchase event may be duplicated, lack revenue, or disappear after a subscription restore. Start the audit with test transactions.
A minimum viable measurement routine
Document event definitions and windows, test the integration on real devices, and keep one dashboard version throughout an experiment. Each week, reconcile spend, product events, App Store Connect data, and postbacks.
Measurement does not need to explain every install. It needs to separate a promising hypothesis from a weak one quickly enough to prevent the team from scaling activity that creates no revenue.