Late postback / delayed purchase handling in F2P daily pacing

Daily pacing that assumes SKAN revenue is complete by midnight will under-spend every iOS learning cell.

Checklist

1. No same-day SKAN kill — Apple documents a 24–48 hour random delay after the first conversion window ends or after a lock (multiple conversion windows).
2. Budget for later windows — Second and third postbacks send after a 24–144 hour random delay following their windows (days 3–7 and 8–35).
3. MMP continuous lag — Data freshness cites roughly 15–60 minutes for continuous attribution updates; do not treat a 10-minute dashboard hole as under-delivery.
4. SKAN reporting day — Same freshness article: postbacks received on day D are available by 10:00 UTC on D+1.
5. Lock ≠ instant — Locking a conversion value finalises it early but the random delay still applies.

Facts vs limits

CSV.

Draft prepared 20 September 2026. Local expansion-500 file; human editorial review pending. No CMS write.