Email Warm-up: Build Sender Reputation Before You Scale
Warm-up is how a new domain earns trust: which signals receivers measure, a six-week ramp schedule, how to pick the audience, what to send, and the metrics to watch daily.

Email Warm-up: Build Sender Reputation Before You Scale
A domain that has never sent mail has no reputation. That is not a neutral starting point — it is a suspicious one. Receiving servers have watched domains appear overnight, blast a million messages and vanish, so a brand-new sender pushing volume is treated as a risk until it proves otherwise.
Warm-up is the process of earning that trust deliberately: start with a small volume, send to people who will engage, ramp slowly, and let the engagement and low complaint rate build a history. Skip it and the same list that would have landed in the inbox arrives in spam, or does not arrive at all.
What Receivers Actually Measure
Reputation is not one number. Mailbox providers weigh several signals, and most of them are about behaviour over time rather than content:
| Signal | What it looks like when healthy | What damages it |
|---|---|---|
| Volume pattern | Steady, gradual growth | Sudden spikes from zero |
| Complaint rate | Under 0.1% | Users pressing "report spam" |
| Hard bounce rate | Under 2% | Stale or purchased lists |
| Engagement | Opens and clicks, especially early | Long stretches of unread mail |
| Authentication | SPF, DKIM and DMARC all pass | Missing or misconfigured DNS |
| Spam trap hits | None | Mailing addresses that never opted in |
Reputation is tracked per sending IP, per domain, and increasingly per subdomain. If you send from a shared IP pool, you inherit some of your neighbours' behaviour; if you have a dedicated IP, you also own its entire history, including the cold start. A new dedicated IP with no history is worse than a shared pool for the first few weeks, which is why providers ramp it gradually.
The Ramp Schedule
There is no universal schedule, but the shape is consistent: start small, roughly double while things look clean, and back off at the first sign of trouble. A six-week ramp for a list in the tens of thousands:
| Week | Messages per day | Audience |
|---|---|---|
| 1 | 50–200 | Most engaged: opened in the last 30 days |
| 2 | 200–1,000 | Opened in the last 60 days |
| 3 | 1,000–5,000 | Opened in the last 90 days |
| 4 | 5,000–20,000 | Active subscribers, still excluding cold segments |
| 5 | 20,000–50,000 | Most of the list |
| 6+ | Full volume | Everything, including re-engagement |
Two rules override the table. First, never send more than you can handle if a batch bounces: if deliverability degrades, drop back to the previous week's volume for three days. Second, volume is per day and should be spread across the day rather than dumped in one burst — a steady trickle looks like a normal business, a single spike looks like a campaign from an unknown sender.
Audience Selection Matters More Than Volume
The fastest way to ruin a warm-up is to send to your least engaged addresses first. Cold contacts do not open, do not click, and are the most likely to hit "report spam" — exactly the signals you cannot afford while reputation is being established.
Start with the segment that has opened something in the last 30 days. Exclude anyone who has not opened in six months, anyone who has hard bounced, and anyone on a suppression list. If you are starting from zero and have no engagement history, use a double opt-in confirmation flow: the people who confirm are, by definition, engaged, and the confirmations themselves are a clean signal.
Get Authentication Right First
Warm-up cannot compensate for a broken DNS setup. Before the first message:
- Publish an SPF record that authorises every sending source. See SPF record setup.
- Enable DKIM signing on the sending domain and verify a test message passes.
- Publish a DMARC record. Start at
p=nonewith a reporting address, read the aggregate reports for two weeks, then move top=quarantine. The DMARC guide walks through the record itself. - Use a dedicated subdomain for bulk or marketing mail, so a campaign mistake does not damage the reputation of password resets and invoices.
That last point is worth emphasising. Transactional mail — password resets, receipts, security alerts — is the mail your users actually want, and it should not share a reputation with a promotional blast. Sending from mail.example.com for transactional and news.example.com for marketing lets you throttle or pause one without touching the other.
What to Send During Warm-up
Content does not have to be promotional, but it should be wanted. The best warm-up material is mail people expect and act on: account notifications, receipts, digests they signed up for, a welcome series. Avoid, during the ramp:
- Purchased or scraped lists, in any amount
- Re-engagement campaigns aimed at people who have ignored you for a year
- Image-only emails with little text
- Link-heavy messages to domains with no reputation
- Seasonal blasts timed to the ramp — a new sender at Black Friday volume is the classic failure
Keep the HTML clean, include a plain-text alternative, and make unsubscribe obvious. A visible unsubscribe link reduces complaints, and complaints are the signal that actually gets you filtered.
Monitor These Numbers Daily
| Metric | Healthy during warm-up | If it degrades |
|---|---|---|
| Delivered rate | Above 95% | Check authentication, pause the ramp |
| Hard bounce rate | Under 2% | Clean the list before sending more |
Deferral rate (4xx) | Low and falling | You are sending faster than the receiver accepts — slow down |
| Complaint rate | Under 0.1% | Stop, find the segment causing it, remove it |
| Open rate | Stable or rising | Falling opens mean spam placement, not disinterest |
Deferrals deserve a note. A 4xx temporary rejection is the receiving server telling you to slow down, and it is the earliest warning you will get. Providers often surface deferrals separately from bounces; if you only look at "delivered" and "bounced", you will miss them. The bounce handling guide covers the classification.
Common Mistakes
Ramping on a schedule while ignoring the signals. The table is a starting point, not a contract. If complaints rise or deferrals spike, the correct move is to slow down, not to keep to the calendar.
Warming up with the full list "just to test". One large send to an old list can undo weeks of work, and some providers never fully recover a reputation damaged this way.
Warming up a domain but not the subdomain you actually send from. Reputation follows the domain in the From header and the DKIM signing domain. Make sure the ramp happens on the identity you will use in production.
Treating transactional mail as exempt. It is not. A password-reset email from a domain with no history lands in spam just as readily, and users blame the product. Warm the transactional subdomain too, at lower volume — it only needs a few hundred messages a day to establish a pattern.
If Reputation Is Already Damaged
Pause high-volume sends. Segment down to the most engaged recipients and send only to them for a week or two, at low volume. Fix any authentication failure first — a broken DKIM signature will keep you filtered no matter how good the engagement is. Then resume the ramp from a lower rung. Recovery is slower than the original warm-up, which is the argument for doing it properly the first time.
A warm-up is not a formality before "real" sending; for the first month it is the sending. Plan it the same way you plan the campaign that follows.