Why Thailand UA often needs a different window

Model Routehub teaches from Thonburi because the local path from ad to install is not a footnote. LINE conversations sit between click and store. Checkout is often delayed by cash, transfer slips, or a family decision. Dual-SIM Android devices hop networks mid-funnel. Prepaid users delete and reinstall without the emotional weight a US retention model assumes.

LINE is not a side channel

When a prospect clicks, chats, and only later follows a deep link, last-click credit can land on the message rather than the campaign that started the chat — or the reverse, depending on how you instrumented the OA. Neither outcome is “wrong” until you write the rule. Teams that never write the rule spend stand-ups arguing about which dashboard likes LINE more.

Our practical advice: keep a short window for media optimisation so bidders do not feast on delayed chats, and a longer observed window for reporting so finance sees applications that complete after the conversation. Present them as two numbers. Do not average them into a fake ROAS.

Delayed revenue is not fraud

Card products and marketplaces in Thailand routinely see conversion days after the click. If you shrink the window to fight suspected farms, you will also amputate real late buyers. Separate the fraud screen (CTIT, ASN, install velocity) from the revenue window. They answer different questions.

Prepaid SIMs and reinstalls

Reattribution windows copied from always-on US subscribers will recapture too much. A user who burns through a SIM and returns in three weeks is not the same story as a subscriber who churned. We ask students to plot reinstall lag for Thailand Android separately before they copy a global MMP default.

None of this means “longer windows are more true.” Longer windows are more inclusive and more contaminated. The craft is naming the contamination. That sentence belongs in the lab capstone and, frankly, in your next finance memo.

If you want the module list that includes this as a full week, see the Attribution Measurement Lab. For the conceptual map, the attribution desk page is shorter.