Warmup & Migration
New infrastructure starts with zero trust, and mailbox providers extend credit slowly. Warmup is how you earn it, migration is how you keep it, and moving too fast is how you start over.
Every ESP hands you a warmup schedule, and most teams follow it like a compliance document. Here is what the schedule actually is: a guess. The providers publish the principle and refuse to publish the numbers. Start low. Send to people who want your mail. Increase gradually. Slow down when the servers tell you to. Everything more specific than that is practitioner convention, which is why published schedules disagree with each other and why following one on faith fails. The schedule is not the mechanism. The audience is the mechanism, because new infrastructure is graded almost entirely on how recipients react to its first weeks of mail.
Know what transfers and what does not. Domain reputation travels with you across ESPs and IPs as long as authentication stays intact. IP reputation does not exist until you build it. That asymmetry sets the timeline: a new IP under an established, trusted domain warms in roughly two to four weeks, and a new IP under a new domain is a four-to-eight-week project, because you are asking providers to trust two strangers at once.
I learned the cost of rushing this the expensive way. Warming dedicated IPs for a national retail brand, we went too large, too fast. Reputation dropped, the ramp collapsed, and the only way out was backward: cut volume, restart the schedule, and spend weeks re-earning what a patient ramp would have kept. We got to trusted-tier reputation in the end, and the setback became the method I have used since. Watch reputation daily, not weekly. Start with your most highly engaged recipients, the people who reliably open and click, and let their behavior introduce you to the algorithms. Expand to broader audiences only after the trust is visible on the dashboard. The ramp is not a volume schedule with dates. It is an evidence schedule with gates.
The daily watch is Google Postmaster Tools: user-reported spam rate, domain reputation, authentication pass rates, IP reputation. Those four, every day of the ramp, before you look at a single campaign metric. Opens tell you how the campaign did. Postmaster tells you whether there will be a next campaign.
One hard line before the steps: artificial warmup services, the tools that simulate opens, replies, and spam-folder rescues across bot networks, are filter evasion. Blocklist operators classify them as exactly that, and providers penalize what they detect. Reputation you did not earn from real recipients is a debt that comes due the first time you mail a real audience.
Two artifacts come out of this playbook: a volume-tiered warmup schedule with pause gates attached, and a cutover runbook for moving ESPs or IPs without dropping the trust you already hold.
| Phase | Typical window | Daily volume shape | Audience | | --- | --- | --- | --- | | Establish | Week 1 | Hundreds to low thousands | 30-day actives only | | Build | Week 2 | To roughly 5,000 | 30-day actives | | Stabilize | Week 3 | To 15,000-20,000 | Add 60-day actives | | Scale | Week 4 and on | Toward target | Add 90-day actives last | Ranges, not rules: high-volume programs run the same shape at a larger multiple, and the gates apply identically. Above roughly 100,000 sends a day you should be on dedicated IPs and warming them deliberately; below it, a reputable shared pool is the better start, and the ramp discipline still applies to the domain.