SKAN windows when Thai iOS volume is thin
In several Thai categories, a week of iOS user acquisition produces fewer SKAN postbacks than a slide can honestly split by channel. That is not a broken MMP. It is a sample size. Mobile attribution reporting has to say so out loud.
Teams arrive at Apiauditgrid hoping SKAN will replace the deterministic IDFA story they lost. Sometimes it does, on large catalogues with conversion values that were designed by someone who still works at the company. More often in Bangkok, the postback file for the week is a short list with many nulls. The temptation is to treat null as zero, or to push the same percentage split you saw on Android onto iOS. Both moves make a board pack look complete. Both fail a finance question two meetings later.
Lock windows first
Before any total, write the lock window on the page. A number published on Monday that can still move on Wednesday is not a closing figure. In class we put the lock state in the same cell as the install count. If the window is open, the cell is labelled provisional. If you need a number for media buying the same day, use a separate line called “operational, unlocked,” and never mix it with the signed total.
Null is not zero
A null postback means Apple did not return a conversion value you can read, not that the user did nothing. Folding nulls into zero conversion quietly understates value and overstates cheap channels. The memo sentence we teach is plain: “N postbacks received, M null, conversion-value map covers events A–C, no channel split claimed below thirty observed postbacks.”
Thailand-specific sparsity
Public holidays, carrier billing, and smaller iOS shares in some verticals make weekly SKAN even thinner. We ask students to keep a rolling four-week SKAN appendix rather than forcing a weekly channel chart. Four weeks is still a sample. It is simply less embarrassing than seven postbacks drawn as a pie.
If you want to drill this with files, it lives in module 02 of The Attribution Ledger and in the SKAdNetwork Postback Studio.