2 April 2026 · Journal
Quiet hours and time-to-open in ICT
Thailand is one timezone on paper and several daily rhythms in practice. Push notification performance that ignores local clocks will misread both fatigue and “engagement”.
A tap at 02:10 ICT is not the same signal as a tap on a BTS commute. System quiet hours, user-set do-not-disturb, and vendor throttling all swallow messages without always returning a clean failure. Time-to-open averages hide this. Distributions do not.
Draw the day before you draw the chart
In Quiet Hours, Loud Data we ask operators to sketch three audiences: inner Bangkok, suburban ICT, and a provincial cluster they actually serve. Morning peaks differ. Evening meals differ. A single “best hour” taken from a national average usually fits none of them.
Festival weeks deserve their own row. Songkran and the Western new year are not merely “lower volume”. They are weeks when lock-screen real estate is contested by family chat, travel alerts, and carrier messages. Your promotional card is not competing with last Tuesday’s control; it is competing with a different world.
When the ESP cannot honour a window
Several alumni arrive hoping we will reveal a hidden toggle. Often there is none. The studio then becomes a briefing: which user segments must be suppressed, which can be delayed in your own job queue, and which leakage you will simply report. That outcome disappoints people. It is still better than a dashboard that pretends the window exists.
A metric you can defend
We prefer median time-to-open within the intended local window, plus a leakage rate for messages first opened outside it. Influence is still measured separately. A fast tap outside quiet hours is not automatically a win if the next action would have happened from a receipt anyway.
If you only change one thing after reading this, stop reporting a single national best-send hour. Split the clock. The worksheet in class is deliberately ugly — a spreadsheet with three time bands — because ugly and true beats polished and averaged.