Reading Thailand retention curves without importing someone else’s calendar
Songkran, exam weeks, and prepaid top-up days rearrange week-4 points. A US “always-on” template will call that noise. It is the signal.
Journal
Field notes, not recap threads.
Short public essays from the Bangkok desk. The benchmark tables stay behind a Desk Pass; the arguments about how to read them do not.
When DAU misleads product leads
Daily active users count a crowd. Engagement asks whether anyone returned with intent. A Bangkok grocery example where the two diverged in public.
Benchmarking push opt-in without vanity
Opt-in rate after a celebrity campaign is not a durable consent baseline. How we treat decay, silent hours, and the tap that is not a contract.
Session length is not engagement
A longer session can be a lost user. Why Session Quality Lab scores task completion and ignores time-on-app as a trophy.
Building a cohort calendar for festival season
How to schedule first-open windows so Q4 and Songkran do not contaminate each other, with a worked example from a utility app.