sending-spread-the-daylisted

# Spread the day, never send a burst A sender's daily volume is stretched over its whole working window. The wait before the next cold touch is the time left in the window divided by the touches left in today's plan, and it is recalculated on every tick. Replies inside a live conversation are the one exception: they go out at once. ## Why In our own sending, pacing used to be a random jitter of seconds between queue rows. On one day the full daily plan of five profiles (invites, messages and likes) went out between 07:00 and 08:00 UTC, the first hour of the window. That burst hurts three ways: - For the platform it is the handwriting of automation. - For the lead it is a pile of notifications in the same minute. - For us it is a blind spot: a profile that finished its day by 10am looks exactly like a stopped profile for the rest of the day. ## How to apply 1. Define each sender's working window (start and end, in the time zone that matters for the touch). 2. Define today's plan as the smaller of two numbers: the limit left for today and the length of the queue. 3. On every tick compute: `interval = time left in window / touches left in plan`. 4. Add jitter around that interval (we use plus or minus 30%) and a floor (we use 4 minutes between cold touches). 5. Send at most one cold touch per profile per tick, and never more than one cold touch per profile per minute. 6. Recompute on the next tick. A queue refilled at lunch then stretches over the rest of the afternoon by it