Pacing and sending windows

A daily budget says how many emails may leave a mailbox. It says nothing about when. Two hundred emails sent one per minute over an afternoon and two hundred sent in ten seconds are the same number and completely different behaviour. This page is about the timing half.

Why not all at once

A human being with a mailbox does not send 200 emails in one second. Nothing that behaves like a person produces that shape. Bulk mail services do, and spam does, so a sudden burst out of a mailbox that normally does not burst is one of the easier things for a receiving mail server to notice.

So ColdMerge spreads the work out. It takes the time left in your sending window, divides it by the work still queued, and sends on that gap with a random wobble added on top. That wobble is called jitter, and it matters: perfectly even sends every 43 seconds also look like a machine. Human timing is uneven.

Two hard bounds sit around that gap.

A floor of ten seconds. No two sends from one mailbox ever go out closer together than ten seconds, and this floor outranks your daily budget. It is not a feel number. The largest provider ceiling is 2,000 a day and the default window is eight hours, which works out to a send every 14.4 seconds. A floor above that would quietly throttle a limit we publish. Ten seconds sits comfortably under it while making the absurd cases impossible.

Set a fifteen-minute window against a 500-a-day budget and you are asking for a send every 1.8 seconds. ColdMerge will not do it. The budget simply goes unspent and rolls into the next day. This inverts the intuitive priority and is worth stating plainly: the daily budget is a ceiling ColdMerge may leave unspent, not a quota it chases.

A ceiling of about fifteen minutes. The gap does not stretch indefinitely either. Spreading your last three sends across eight hours buys you nothing, because the anti-burst rule is a lower bound and there is no upper risk to protect against. It also means a nearly-finished campaign does not trickle for the rest of the day. The ceiling bounds the pace rather than each individual gap, so a jittered gap may run slightly over fifteen minutes. Nothing is unsafe about a gap being a bit long. Nothing may go under ten seconds.

Why a sending window exists

A sending window is a start time, an end time, a set of weekdays, and a timezone. The default is 09:00 to 17:00, Monday to Friday. You set it per campaign, and it starts from an organization-level default.

Two reasons it is there.

The first is that it looks right. Cold outreach arriving at 3am local time is obviously automated, and obviously not written by the person whose name is on it. Business email happens during business hours.

The second is that it works better. A message that lands at 9:40am sits near the top of an inbox someone is actually looking at. A message that lands at 2am is buried under everything that arrived before breakfast.

A few rules worth knowing:

A window never wraps past midnight. A window from 22:00 to 06:00 is refused rather than quietly interpreted. If you need that shape, you are probably reaching for a timezone instead.
Daylight saving is handled in your timezone. The next opening is worked out in the window’s own timezone rather than by adding 24 hours to the last one, so a clock change never pushes a send an hour outside the hours you set.
A scheduled start is a “not before” floor, never an override. Schedule a campaign for Saturday against a Monday-to-Friday window and the first send resolves to the following Monday morning. The launch screen shows you the resolved first-send time before you commit, so there is no surprise.

What happens when the window closes

Nothing dramatic. Sending stops, and picks up when the window next opens.

A campaign too large to finish today continues tomorrow. That is ordinary progress, not a failure, and the campaign reports it that way (“day 2 of 3”) rather than as an error. Whatever the budget allowed but the window had no room to pace simply goes unspent and overflows into the next sending day.

One consequence catches people out. A mailbox serving two campaigns paces against the later-closing of their two windows. If campaign A closes at 5pm and campaign B closes at 8pm, the mailbox derives its gap from the longer stretch. So campaign A may not finish inside its own window. That is ordinary overflow, not a fault, and the alternative would be worse: pacing against the earlier close would cram all the work into a fraction of the time and then sit idle.

Why the estimate is an estimate

A running campaign shows a pace estimate, something like “about N a day, roughly N days left”. It is labelled as an estimate on purpose, because the true answer depends on things that have not happened yet: how much budget each mailbox will actually have tomorrow, whether a mailbox hits Google’s quota, whether anyone replies, whether follow-ups come due and take priority over new sends.

Elapsed days are a fact. Remaining days are a forecast, and it counts sending days only, not weekend days your window excludes.

Similarly, “next send” is not a stored countdown. It is worked out fresh from the window, the mailbox’s availability, and the remaining budget, and when something is missing it names what is missing rather than guessing a time.

When one mailbox stops and the others carry on

If Google tells a mailbox it is over quota, that mailbox stops claiming new work until its campaign’s window opens again. Other mailboxes on the same campaign carry on as normal.

This is treated as a budget signal, not a failed message. It usually means ColdMerge’s picture of that mailbox was incomplete, most often because you also send by hand from it and those sends are invisible to us. It clears itself with time, and it is never recorded as a problem with the mailbox connection itself.

Ramp-up and send budgets covers how many emails a mailbox may send, as opposed to when.
Campaigns and sequences covers setting a campaign’s window and scheduling a start.