Thailand · January 2026
Experiment cadence for Bangkok product teams
Imported experimentation playbooks assume a quiet American quarter. Thai mobile teams live on a different clock: store review queues, Songkran, and OEM update waves that arrive without a calendar invite.
We keep a paper calendar on the Pranakhon wall. It is more useful than most dashboards in April.
Holidays are not outliers; they are the design
Songkran, Chinese New Year shopping spikes in Yaowarat, and long royal holidays change mix more violently than a typical UI variant. If your metric window includes those days without a pre-registered note, you are measuring Thailand, not your flag.
A Chiang Mai grocery pair in Bench Cohort used to launch “small tests” ten days before Songkran because the office would be empty. Their results were noisy for reasons that had nothing to do with copy. They now freeze new variants across that window. Conversion still moves. They no longer attribute it to a button colour.
Store time is not warehouse time
iOS review can strand a treatment binary while Android is already at 40% rollout. A single “experiment start” timestamp in the warehouse then mixes two different populations. Release Window Analytics exists because this mismatch is ordinary in Bangkok teams shipping both stores.
OEM skins add another delay. A device cohort on a popular Thai Android build may receive your variant days later, then retry events in a burst. If you close the window on Friday night, you flatter iOS.
What we ask teams to put on the wall
Ship dates, review buffers, Thai holiday blocks, and a rule: no new primary KPI claims that include an unmarked holiday. It is unfashionable. It prevents a certain kind of board slide.
The studio is in Pranakhon because this calendar is not theoretical for us. If you want help fitting your next flag to it, write the desk.