The 28-day warm-up schedule
Twenty-eight dated rows with the volume, the audience, and the gate for each one, plus the honest note that most senders never need row seven.
A domain that has never sent email starts at roughly 150 messages on its first day and multiplies by about 1.4 each day through day 7. From day 8 it grows 10 to 20 percent a day, and only while complaints stay under 0.1 percent and bounces under 2 percent. Below is that curve as 28 dated rows, plus a column for what actually sent and what it cost.
It is for a domain with no sending history: a new business domain, a marketing subdomain split off from the root, or a domain restarting after months of silence, which mailbox providers score much like a new one. The reasoning behind each number is in our warm-up article; this page is the planner.
One prompt, the exact curve above built into it: tell it your start date and target volume, and it builds the dated schedule, then holds you to the gate every day you report back. Paste it into ChatGPT, Claude, or your own agent.
Or skip the copy-paste: connect your agent over MCP and it can track the real numbers against a real domain as you send.
The schedule
| Day | Phase | Target volume | Send to | Gate before you increase |
|---|---|---|---|---|
| 1 | Warm | ~150 | The most engaged people you have, active in the last 90 days | None. This is the floor, not an achievement |
| 2 | Warm | ~210 | Same segment | No bounces from yesterday's send |
| 3 | Warm | ~290 | Same segment | Bounces under 2%, no complaints |
| 4 | Warm | ~410 | Same segment | Bounces under 2%, no complaints |
| 5 | Warm | ~575 | Same segment, widened slightly | Complaints under 0.1%, bounces under 2% |
| 6 | Warm | ~805 | Same segment | Complaints under 0.1%, bounces under 2% |
| 7 | Warm | ~1,130 | Same segment | Complaints under 0.1%, bounces under 2% |
| 8 to 14 | Ramp | +10 to 20% daily, reaching ~2,200 to ~4,000 | Engaged first, widening one step per day | Both rates checked before every single increase |
| 15 to 21 | Ramp | +10 to 20% daily, reaching ~4,300 to ~14,500 | Widening toward the full list | Both rates, plus no gap days |
| 22 to 28 | Ramp | +10 to 20% daily, reaching ~8,400 to ~52,000 | Full list minus never-engaged | Both rates, plus no gap days |
| 29+ | Steady | Your relay's cap | Full list minus sunset inactives | Hold. What scores you now is behavior, not volume |
Two notes on the arithmetic, because published ramps rarely show their working. Compounding 1.4 from 150 lands near 1,130 on day 7, while Resend's warm-up documentation (checked June 2026) describes the same ramp as reaching about 1,100 a day by the end of week one. Take the lower figure: a cap is a ceiling, not a quota, and no domain was ever penalized for sending under its maximum. The week-two-to-four ranges are just that compounding continued at each end of the 10 to 20 percent band, which is why the day-28 spread is so wide. Pick one growth rate and stay on it.
The gate, in four lines
- Complaints under 0.1 percent, the target from Google's bulk-sender rules (February 2024, updated 2025).
- Bounces under 2 percent, measured on the send that just went out, not the campaign before it.
- Hard stop at 0.3 percent complaints. Above that line Google withholds delivery mitigation until you have been back under it for seven consecutive days, so one bad day costs a week.
- No gaps. A day you skip is worse than a day you send small, because consistency is itself the signal Outlook rewards.
Our send path enforces the same two thresholds automatically and pauses itself when either breaks. Automatic matters here: the moment you most want to keep sending is the moment the numbers say stop.
What to fill in each day
| Column | What goes in it |
|---|---|
| Date | The real calendar date, written before you send |
| Planned | The target from the table, or lower |
| Sent | What actually left |
| Bounces | Count, then percent |
| Complaints | Count, then percent |
| Gate held | Yes or no, decided before tomorrow's number |
| Tomorrow's cap | Today's sent, times 1.1 to 1.2, or a repeat of today if the gate failed |
Fill that last column in the evening, not the next morning. It turns tomorrow's send into arithmetic rather than a decision made under pressure, which is the entire discipline.
The honest caveat
Most senders never reach row 7, and that is fine. A list of 900 people is a 900-message send, once a week, and after the first two sends the ramp stops being what limits you. What limits you then is whether the people on the list asked to be there.
If your list is smaller than day 1's number, the schedule collapses into one instruction: send to everyone who genuinely opted in, on a predictable cadence, and watch the two rates. We say this from a small workspace rather than a large one. LetterDuck has run live on a real site's mail since June 2026 at modest volumes, and at that scale the free tier's own daily cap enforces the early ramp before our throttle gets a vote.
Gmail's bulk-sender line starts at 5,000 messages a day, which almost nobody reading this will cross. The authentication and complaint rules underneath it apply at every volume, including dozens a day, and we follow them at dozens a day.
How to use it
Mark a date next to day 1 and fill the row in the evening from your platform's reporting or Google Postmaster Tools, whether that's on paper, in the AI prompt above, or in the browser tool below. Miss a day and do not resume where you stopped: drop back one row and climb again.
Prefer a plain reference to read, not a prompt? Download the PDF.
