12 Mar 2026 · Measurement

Why SKAN and GAID refuse to tell one story

Leadership still asks for a single acquisition number. iOS and Android still refuse to share a clock. Alignment is the work of naming that refusal, not papering over it.

SKAN postbacks arrive late, coarse, and sometimes not at all. GAID-based Android paths, where they still exist, arrive faster and with more identity. A Thai catalogue that sells more on iOS will always look ‘worse’ on a blended ROAS slide if you pretend those two streams are the same instrument.

What actually differs

Conversion values on iOS are buckets, not receipts. A wallet top-up of THB 49 and THB 490 can land in the same value if your mapping is lazy. Android, even with privacy tightening, still often carries a device-shaped identifier into the MMP. When finance compares ‘iOS revenue’ from SKAN to ‘Android revenue’ from the warehouse, they are comparing a sketch to a ledger.

Add ICT. Campaigns that end at midnight Bangkok time do not end at midnight in the SKAN window. Teams in our flagship studio now print both clocks on the reconciliation sheet. The delta stops looking like fraud once it looks like weather.

A known delta, not a blended average

The Alignment Atlas treats SKAN and GAID as neighbouring territories, not as two sensors on one machine. You still report both. You do not average them into a number nobody can defend in a boardroom in Sathorn.

If your iOS share is high — common for premium Bangkok apps — expect SKAN to under-count relative to first-party purchase events. That under-count is often the honest number. Inventing users to close it is how stacks lose finance later.

What we ask cohorts to bring

Two exports from the same week: SKAN postbacks and Android attributed installs or purchases. Plus the first-party purchase table. If you cannot produce that trio, you are not ready to argue about ROAS. You are still arguing about instrumentation.