Mobile attribution, without the sales pitch

A short, honest account of what can actually be measured on iOS today, and what vendors imply is measurable but is not.

6 August 2026 by PAD team

The problem being solved

Someone taps an ad, goes to the store, installs, and opens the app. Nothing carries that person's identity across those steps by default, so attribution is the business of reconnecting the two ends.

Deterministic matching

On Android, the Play install referrer travels through the install itself, so the connection is exact. On iOS there is no equivalent, which is the root of every difficulty that follows.

Probabilistic matching

Without an identifier, vendors match on signals: IP address, device model, OS version and timing. It works statistically at volume and is unreliable for any individual install. Apple discourages it, and it is prohibited outright when the user has denied tracking.

SKAdNetwork and AdAttributionKit

Apple's own answer: the store reports a conversion to the ad network, with deliberate delays and coarse values so individuals cannot be identified. You get campaign-level truth and lose everything below it.

• Delayed by design, so same-day optimisation is impossible
• Limited campaign identifiers, so granular structures do not fit
• Privacy thresholds suppress small numbers entirely
• No user-level data, ever

What is left that is simple and true

Campaign tokens on the store URL. Apple reports installs against ct in App Store Connect without any SDK, at campaign granularity, for free. It is coarse, but it is deterministic and nobody can take it away.

The measurement nobody sells you

Before any of this: did the tap even reach the store? Every ad platform counts a click that ends on a blank page as a click. Measuring the share of taps that actually arrive is unglamorous and often reveals more than a sophisticated attribution model applied to traffic that never landed.