Warming a New Sending Domain Before a Cold Campaign

Plan for 20 to 50 sends per day in week one, all to contacts with a legitimate prior relationship: past customers, newsletter opt-ins, people who downloaded something, event sign-ups. Resend's advice is to use double opt-in and only send to people who opted in, because engagement is the signal that builds reputation per its walkthrough. This week is not a campaign. It is a proof of life.

Keep the daily count flat within a step. Mailbox providers reward predictability and penalize spikes, which is the same principle HighLevel's sending guide repeats when it warns that suspicious frequency, such as 100 messages on day one, disrupts warm-up in its email warm-up guide.

A worked schedule for the whole month, with the assumption of a 500-recipient-per-day launch target:

Two notes on the numbers. First, the doubling pattern follows the brief: grow only every three to four days, and only on green gates. Second, 500 sends per day at day 28 assumes one mailbox. Instantly's practical ceiling guidance is 30 to 50 emails per day from each mailbox, and it recommends capping at three to five mailboxes per domain in its strategy guide. If you need 500 per day out of the gate, you are running several mailboxes in parallel, each on its own ramp, and your gates must be evaluated per mailbox as well as in aggregate. An aggregate 1.8% bounce rate can hide one mailbox at 5%.

The daily ten minute routine

Do this at the same time each morning, before you start the day's send:

• Pull yesterday's hard bounces and complaints. Divide each by yesterday's delivered count.
• Compare both numbers against 2% and 0.1%. Write them down with the date.
• Open one raw header from a real send and confirm three passes.
• Check the seed test result if you ran one.
• Green means send the planned volume. Red on any gate means stop.

From week two you may begin mixing cold recipients into the volume, but keep a majority of engaged contacts in the mix while the domain is still young. The reason is arithmetic: you need positive engagement signals to offset the ignorance and occasional complaints that cold mail generates. A day of 120 sends to 120 strangers produces no opens and no replies; a day of 120 sends where 60 are warm produces engagement your reputation can use.

Copy discipline is now part of warm-up, because spam complaints are a copy problem before they are a reputation problem. Two data points worth keeping in mind. The Journal of Business and Entrepreneurial Studies analyzed 5,187 cold emails with an average response rate of 15% and found that messages under 150 words with the sender's introduction in the first three seconds lifted response from 17% to 44% in its published study. Short emails get replies, and replies are a strong positive signal. Meanwhile, Instantly's benchmark summary reports that personalized outreach, as opposed to generic template copy, reaches up to 18% reply rates against a platform average of 3.43% in its benchmark post.

The practical translation: three genuine custom snippets per message at this stage. Woodpecker's dataset of more than 20 million sales emails reports roughly 17 to 18% reply rates for advanced personalization and around 7 to 9% for basic or none, and notes that unique snippets also reduce the chance that identical messages trip spam filters in its statistics report. That second effect matters during warm-up more than at any other time.

Keep the legal footer present from day one, because it protects you independently of reputation. The FTC's CAN-SPAM guide requires accurate headers, a non-deceptive subject line, identification of the message as an ad, a valid physical postal address and a clear opt-out mechanism, and it makes no exception for business-to-business email per the FTC's compliance guide. One more number to keep you honest about why the footer is automated rather than manual: opt-out requests must be honored within 10 business days, and a failure to propagate an unsubscribe across systems is exactly where exposure sits per Transcend's compliance analysis. Build the unsubscribe into the template and sync it to every system that can send, before you send anything.

This is where most warm-up schedules fail, not because the gates are wrong but because the sequence gets ahead of the reputation. Going from 180 to 500 sends per day over nine days is a 2.8x increase. It is safe only if each step is green. If day 20 fails the bounce gate, the correct response is to hold at 250 and repeat the step, not to shave the target.

Two numbers that make the case for patience. Woodpecker's data puts the average bounce rate across its sending at 5.1%, with below 2% rated good and below 1.5% excellent per its statistics report. That gap between the average and the target is what you are trying to avoid, because as a new sender you start with no cushion. Separately, Martal's benchmark roundup reports that around 17% of cold emails never reach the inbox, often from poor domain authentication, high bounce rates or spam-triggering language in its statistics roundup. The dominant causes on that list are the two gates you check every morning.

Format testing belongs here as well. If you want to test whether subject-line length, plain text versus HTML, or a shorter body changes complaint rates, do it inside the current volume step and give it four full days before you read the results. A test that spans a volume increase confounds the copy change with the reputation change.

Full campaign volume starts in week four, and reputation management does not stop there. Lock in steady-state limits: a fixed per-mailbox daily send count, a fixed follow-up cadence, and a dashboard with the four gates on it.

On cadence: follow-ups generate replies, and replies are good for reputation, but follow-ups also raise complaint risk if they are aggressive. Follow-ups collectively generate 42% of campaign replies, yet nearly half of reps never send a second message, according to Martal's roundup of benchmark data in its statistics roundup. A schedule of one follow-up every third day with a hard stop on reply is enough to capture the majority of that upside without hammering the same recipients. Instantly's warm-up walkthrough states the same principle from the sender's side: reply mail that lands in the inbox and unproductive follow-up mail are both judgeable per its own guide.

A steady-state decision rule you can hand to an operator:

• Hold volume flat for any 24 hour window in which a gate fails, then resume at the previous step's volume.
• Pause all sends for 48 hours on any spam landing, then resume at 50% of the prior day's volume if the next four days are green.
• Re-run a seed test any time you change the template, the sending provider or the subdomain.
• Rebuild rather than repair if the same gate fails three resume attempts in a row. HighLevel's guidance is to set up a new subdomain instead of trying to fix a damaged one in its sending guide.

That last rule is the one nobody wants to hear, and it is the one that saves the average 30 day program. A domain at a 3% hard bounce rate after two rebuilds is not going to recover during your launch window.

What to do when a send lands in spam

Follow this in order and do not skip to the template step:

1. Pause the current step for 48 hours. Do not send a smaller volume; send nothing on the failing subdomain.
2. Read the headers of the delivered copy. Confirm SPF, DKIM and DMARC passes and check whether the message was routed to spam by the receiving filter or by your own seed setup.
3. Check the date. If the domain is under a month old, you are inside the window where filters treat it as suspicious by default per Woodpecker's guide, and the answer is simply time.
4. Check the content. Look for URL shorteners, a single image with no text, a lone link and no plain text alternative, and a subject line that does not match the body.
5. Check the complaint number for the last 24 hours. A complaint spike is a copy problem. A complaint count near zero with poor placement is an infrastructure or age problem.
6. Resume at 50% of the previous step's volume and require four green days before returning to the planned schedule.

If you want help deciding whether a stall is a copy problem or a placement problem, and you would rather pay for a diagnosis than guess, a consulting call with Quanta Intelligence can put the data and the decision in one place.

Warming a domain is a reputation accounting exercise, not a calendar. The dates in this schedule tell you when to check a gate; the gate values tell you whether to move. A program that doubles volume on schedule but ignores a 2.4% bounce rate is not two weeks ahead, it is two weeks closer to a burned domain. A program that holds at 180 sends for four extra days and reaches launch on day 34 at full volume has lost nothing.

Quanta Inference is likely one more helper rather than the central tool here: if you generate per-recipient snippets or classify inbound replies across several models, one API key is simpler than three vendor integrations, which you can see at Quanta Inference.