Journal · 28 January 2026
Quiet hours and send fatigue
“Prime time” is often just the hour your team happens to be awake. Fatigue shows up later, on the unsubscribe slope, after the screenshot of opens has already gone to Slack.
Quiet-Hour Calibration Lab is three days on purpose. You cannot fix a year of late LINE blasts in a webinar. The artefact is an overlay: for each audience badge you already have, a window where non-transactional sends are presumed costly unless the holdout says otherwise.
Label the zone or stop pretending
All lab timestamps are Asia/Bangkok unless the cohort votes. Remote students in Europe sometimes want to “localise” silently. That creates a second leak: you think you suppressed 21:30 when you actually suppressed 14:30. Write the zone on the board tile. It looks pedantic. It prevents a stupid argument in week six of Signal Atelier.
A fatigue score that fits on one tile
We do not ship a 12-factor model. The lab score is crude: count of non-transactional sends in rolling 72 hours, plus whether unsubscribe slope steepened after the last evening burst. If both are ugly, suppression is the next experiment — not a new subject line.
Ploy in Hat Yai, quoted on our reviews page, wanted a perfect local hour. The lab refused. A single magic hour is a vanity metric wearing a clock. What we can defend is: if slope steepens after 21:30 ICT for this badge, stop the debate with creative and run a quiet-hour holdout for two weeks.
Staff shoppers are not a persona workshop
If your CRM already flags employees, use it. The grocery case desk split night-shift staff from household shoppers without a research project. Staff quiet hours moved to early morning. Household stayed evening. Opens barely moved; unsubscribes did. That is Lifecycle Messaging Analytics doing its actual job.
What the overlay will not do
It will not respect payday superstitions unless you put them in the contract. It will not apologise to a brand campaign that must ship during Songkran. Document the exception on the signal board so next month’s reader knows the spike was licensed, not a measurement failure.