How to warm up a new sending domain

1. Know what you are warming, which is the domain

Two things get called warming and only one of them is usually yours. IP warming is about the addresses the mail physically leaves from, and at every provider that serves small senders those are shared: your few hundred messages a morning sit inside somebody else’s steady millions, and you neither build nor control that reputation. Domain warming is about the domain in the DKIM signature and the From line, which is yours alone and is the identity that carries your reputation now that authentication is a floor rather than an advantage.

So the decision that matters is not shared against dedicated IP addresses. It is which of your domains does the sending. Use a subdomain reserved for it — send.example.com, mail.example.com — and keep the root domain for the mail humans write to each other. A burned subdomain can be retired and replaced in an afternoon. A burned root domain takes your invoices, your password resets and your replies to customers with it and cannot be replaced at all.

Pick the subdomain before you publish anything, because every record in the next step hangs off it, and a warm-up run against the wrong host is a warm-up of a domain you are not going to send from.

2. Publish authentication before the first message, not after

SPF, DKIM and DMARC have to be in place and passing on message one. Since February 2024 Google and Yahoo have required all three of anyone sending in bulk to their users, and unauthenticated bulk mail is refused at the door rather than filed in a spam folder — a refusal, not a filtering decision, which means it never has a chance to build any reputation at all.

Check that the DKIM signature names your sending subdomain rather than your provider’s. In the headers of a delivered message the signature carries a d= value, and it should read d=send.example.com. If it reads the provider’s own domain, you are warming their reputation and not yours, and the entire exercise below is measuring somebody else’s weather.

Publish DMARC at p=none with a working address in rua= before the ramp starts, so the aggregate reports are arriving while the volume is still small enough to fix things. A policy tightened later, on top of reports nobody read, is how a sender blocks their own mail in week three.

3. Send the first messages to the people most likely to engage

Reputation is built out of what recipients do, so the order of your list is the most important decision in the ramp. Start with the addresses that did something recently and deliberately: bought, replied, clicked, logged in, signed up this week. Send to the least engaged part of the list last, or after a re-permission message, or not at all.

That ordering is also the honest test of whether the list is ready. If the most engaged decile produces complaints, the problem is not the ramp and no schedule will fix it. Stop and look at where the addresses came from.

Where the audience is spread across providers, keep the early batches mixed rather than sorted by domain. Reputation is per receiver: a week of clean sending that Gmail never saw tells Gmail nothing on the day you first send it two thousand messages.

4. Raise the volume on a ramp, not on a schedule

Neither Google nor Yahoo publishes a ramp. Both say the same thing in general terms — increase volume gradually and watch the signals as you go — so anyone quoting a daily table to three decimal places is quoting a habit rather than a rule. The shape below is that habit, written down honestly as a shape: a few hundred messages a day to the most engaged addresses, held for a couple of days, then roughly doubling every two to three days for as long as the numbers stay flat, reaching a steady state in something like two to four weeks.

What makes it a ramp rather than a schedule is that the next step is conditional on the last one. The volume does not double because three days have passed; it doubles because three days passed and nothing moved. That distinction is the entire technique, and it is the part that every published warm-up table leaves out.

5. Watch four numbers while the ramp runs

The spam complaint rate is the one that decides everything. Google asks bulk senders to keep it below 0.3% and treats 0.1% as the level to sit at, and the gap between those two figures is deliberate: a sender parked at the ceiling has no room for a single bad send. It is visible for free in Google Postmaster Tools to whoever can prove control of the domain, which makes it the cheapest instrument you will ever install.

Soft bounces and deferrals are the second. A 4xx response is a receiver saying not now rather than no, and a rise in them at one provider and nowhere else is that provider throttling you. Hard bounces are the third, and on a list that was cleaned before the ramp they should barely exist; a hard bounce rate that climbs as volume rises means the untouched part of the list is worse than the part you started with. Delivery per provider is the fourth, and it only means anything when you break it down — an aggregate that looks fine can hide one large provider quietly refusing everything.

None of these is an open rate, and that is deliberate. Open tracking stopped being a reliable signal in 2021, when Apple began preloading remote content on behalf of readers who turned Mail Privacy Protection on, so a warm-up judged by opens is judged by a number that moves for reasons that have nothing to do with you.

Two instruments worth having in place before the ramp starts, because both take days to become useful:

• Google Postmaster Tools, which reports your complaint rate, domain reputation and authentication pass rates for Gmail, free, for a verified domain.
• A seed list: a handful of accounts you control at the providers your readers actually use, mailed alongside a real send so you can open each one and see which folder the message landed in. Treat it as an indication rather than a measurement — seed accounts have no reading history and filtering is personalised.

6. Hold when a number moves, and find out what changed

The correct response to a rising complaint rate or a burst of deferrals is to stop increasing volume, not to push through and hope the next send averages it out. Averaging is exactly what does not happen: receivers weight recent behaviour, so a bad day at higher volume costs more than the good week before it earned.

Answering a deferral with more connections is the specific mistake worth naming, because it is the instinctive one. A sender that responds to being slowed down by retrying harder is confirming the pattern that caused the throttling. Send less, wait longer between attempts, and look at what went out immediately before the numbers moved — almost always a batch drawn from an older part of the list, or a subject line that surprised people, or a segment that had not heard from you in months.

Enrol in the feedback loops your provider supports, and act on every address they return the same day. A feedback loop reports the messages a provider’s users marked as spam so those addresses can be suppressed immediately, and the one thing worse than a complaint is a second message to somebody who has already made one.

7. Warm a domain when you have nobody to send to yet

The hardest version of this problem is a brand new domain and a list of zero, which is where most people launching a course actually stand, and it is the case every warm-up table ignores because the table assumes volume to ramp. There is nothing to double. What there is instead is a series of small, real, wanted messages: the confirmation mail to the first people who sign up, the first days of the first course, the receipts, the replies. That is a warm-up, and it is a better one than a synthetic ramp because every message in it was asked for.

What this means in practice is that the launch is the warm-up, so do not save the list up. A hundred people who sign up over three weeks and start receiving a course as they arrive is a gentler ramp than the same hundred imported and mailed on one afternoon. A course helps here by its shape: it fans out one message per subscriber per day counted from the day each person confirmed, so volume spreads across the days people happened to click rather than landing in a single hour.

Do not fabricate the engagement. Sending to accounts you control so that you can open them yourself produces a reputation built on behaviour no real reader will repeat, and the providers that matter have been reading through that particular trick for a long time.