On 6 sending days of a cold campaign capped at 20 a day, the first emails sent equalled the cap minus the follow-ups due every time. 20 follow-ups took every slot on 2026-09-08, and a 5-day delay on a weekday window stacks 27 follow-ups on Monday

Method, stated before the numbers

The campaign is the one whose give-report load rate was published yesterday (that read): US and UK web agencies, named people, a first email that links a free AI-visibility report already run on one of the agency's own clients, and one follow-up. Its object on Instantly at the read: daily limit 20, 2 sending inboxes, step delays 5 and 0 days (the wait sits on the step before the gap), send window weekdays 13:30:00–20:30:00. The sent list is every row of the campaign's sent-email endpoint, de-duplicated by id: 118 rows, 78 first emails and 40 follow-ups, against a bulk counter of 116 sent and 40 completed leads. A day is the UTC date of the send timestamp.

Per sending day the script counts rows loaded (lead creation date; a bounced lead, which is deleted within 24 hours, keeps its load day from the repository's private batch file), rows available (loaded before the window closes that day with no first email on an earlier day; a row loaded inside the window dispatches the same day, as 2026-09-03, 2026-09-04, 2026-09-07 show), follow-ups due (first email sent 5 days earlier, or earlier still with no follow-up yet, rolled to the next weekday, not bounced, no reply), and what was sent by step, with the first and last timestamp of each step and the split by inbox. The rule under test is one line: first emails sent on a day equal the smaller of the rows available and the daily limit minus the follow-ups due. The computation is the committed script `content/news/dispatch_order_cut.mjs`; it writes `content/news/dispatch-order-cut-2026-09-10.json` with counts and timestamps only, no addresses, lead ids or report ids, and 13 of 13 self-checks pass, including that follow-ups equal the bulk completed count, bounces equal the bulk bounce count (5), and every post-fix follow-up gap is 4.97 to 5 days.

6 completed sending days and the day in progress

2026-09-02 is the campaign's first day: 10 follow-ups went out the same afternoon as their first emails through a sequence defect fixed the next morning (the delay had been placed on the wrong step), and they still counted against the cap, so 10 of the 16 rows loaded that day were sent and the rest waited. 2026-09-03, 2026-09-04, 2026-09-07 had no follow-ups due and sent first emails only, each up to the rows available or the cap. 2026-09-08 is the mirror: 20 follow-ups due, 20 sent, 0 first emails, 12 rows loaded that afternoon and held. 2026-09-09 split 10 and 10, follow-ups first, with 13 of the 23 available rows carried to the next day. The day of the read, 2026-09-10, has 0 follow-ups due and 13 rows available; 6 first emails had gone out at the read and the rule predicts 13 by the window close, plus up to 7 of any rows loaded before the close.

What 20 a day means for a two-step sequence

Every first email books a second slot 5 days later, so under a fixed cap the new contacts a day can absorb are the cap minus the follow-ups arriving from 5 days before. Sustained, that is at most half the cap in new contacts a day: 10 here. The weekday-only window adds a second effect. A 5-day delay from a Wednesday first email falls due on Monday; from a Monday it falls on Saturday and rolls to the same Monday. Two sending days' first emails therefore fall due on one Monday, which is why Monday 2026-09-14 carries 27 follow-ups from the first emails of 2026-09-07 and 2026-09-09 against a cap of 20: 20 go out and 7 spill to Tuesday, where they meet the 13 due from the 2026-09-10 first emails.

The next three sending days, pre-registered

The same rule, run forward from the sent list at the read with the campaign assumed active and the cap unchanged. Rows loaded after the read are not in it; if rows are waiting they take the free capacity shown. A pause on 2026-09-11 (a read set before these rows were sent, on a different question: whether any agency first emailed on 2026-09-07 has loaded its report) would stop every row below; the table is what the rule says happens if the campaign keeps sending.

The read: on the first weekday after 2026-09-15 the script is re-run and its per-day table published. The claim is falsified on any completed sending day where first emails sent differ from the smaller of rows available and cap minus follow-ups due, or where a follow-up spills to the next day while first-email capacity was used, or where a first email precedes a follow-up on a day that sent both. Today's day in progress is scored then as well: 13 first emails from the 13 rows available at the read.

Artifact

`content/news/dispatch-order-cut-2026-09-10.json` holds the campaign object fields read, the sent-row and lead counts, the per-day table (loaded, available, follow-ups due, capacity, predicted and sent by step, first and last timestamps per step, the order, the inbox split), the three-day forecast and 13 self-checks (0 failed). The earlier reads on this campaign are the give-report load rate and the mailbox-verification cut. Written without em dashes.