Why a New Domain Delivers Worse: Reputation and Warm-Up

« Blog · By

A new domain, everything configured properly: SPF set, DKIM signing, DMARC aligned, content unremarkable. The first send goes out — and lands in spam folders by the row. The usual reflex is to check the configuration again. That is not where the fault is. Something is missing that cannot be configured.

A missing history is not a neutral state

A receiving system judges a sender by how its mail has behaved so far: was it read? Replied to? Reported as spam? Did it go to addresses that do not exist?

For a domain that is three days old there are no answers to any of that. The decisive point is that "no answers" is not scored as a draw. Throwaway domains are the standard tool for bulk sending and fraud: register, send for a few days, discard, next one. A domain with no history therefore looks like the start of exactly that pattern — and is treated with corresponding reserve.

On top of that comes the registration date itself. It is public in WHOIS, and in the first weeks it is a feature that works against the sender. This is not harassment but a statement of probability that has proved correct at scale.

What reputation actually measures

Reputation is not a value but a bundle of observations. The most important ones are known, because the large providers name them in their own postmaster tools:

SignalEffect
Complaint rate (“report as spam”)The sharpest negative signal. Around 0.3 percent it gets critical, at 0.5 percent it gets hard.
Delivery attempts to non-existent addressesIndicates bought or stale lists. A few percent is enough to cause a collapse.
Spam trap hitsAddresses that never consented. A single hit can cost weeks.
EngagementOpening, replying, moving out of the spam folder — the strongest positive signal.
Deleting without openingWeakly negative, but measurable in volume.
Consistency of volumeA jump from 200 to 50,000 messages in one day is suspicious in itself.

Two things stand out. First, most of the signals measure not the message but the recipient list — deliverability is predominantly a list problem, not a text problem. Second, the last row is the reason warm-up exists at all.

Warm-up means: start small and stay steady

The idea is simple. Instead of mailing the whole list on day one, you begin with a small quantity and increase it over weeks. That produces precisely what is missing: a history, and an unremarkable one.

100 1,000 10,000 day 1 3 6 9 13 17 23 30 messages per day, logarithmic axis
A typical curve: from 50 to 50,000 in thirty days. The axis is logarithmic — drawn linearly, the first two weeks would be invisible, and those are the ones that decide.

Recipient order matters more than volume

The commonest warm-up mistake is to follow the ramp and leave the selection to chance. That gives away half the benefit.

Reputation is built from engagement. In the first days, then, the recipients to write to are exactly those most likely to react: whoever bought most recently, whoever opened in the past few weeks, whoever just signed up. Addresses that have done nothing for two years belong at the end — and those that have done nothing for five years do not belong on the list at all.

The same logic applies to providers. Split the list by recipient domain and ramp evenly per provider, and you build trust everywhere in parallel. Reputation is kept per receiving system; a good standing with one provider does not carry over to the next.

Dedicated IP or shared IP

Choosing the sending path raises the same question in another form. A dedicated IP address has its own reputation — yours. A shared IP shares its standing with every other customer of the same provider.

Dedicated IPShared IP
Worth it fromaround 100,000 messages a month, evenly spreadbelow that
Advantagenobody else can damage the standingan existing history is put to use
Drawbacktoo little volume means no reputation, permanentlyanother customer can drag everything down

The drawback in the first column is regularly underestimated. Sending 2,000 messages twice a month over a dedicated IP is too infrequent for a standing to form. Every send effectively starts from scratch. Below a certain regularity, a shared IP with a maintained history is clearly the better choice.

Separating by purpose

One point often overlooked during warm-up: confirmation mail, invoices and password resets must not take the same path as the newsletter. Otherwise a promotional send with an elevated complaint rate drags the deliverability of the invoices down with it — and nobody notices until a customer calls because their invoice never arrived.

When it still stalls

A few observations from practice that save unnecessary debugging:

  • Deferred acceptance is normal. New senders are provisionally rejected more often and have to try again. Why that happens and how long it takes is a subject of its own — it is not a delivery failure.
  • Reputation falls faster than it rises. A single send to a bought list costs weeks. Building it up is laborious; tearing it down takes an afternoon.
  • A setback is not a reason to send more. The reflex to increase volume when open rates fall makes worse exactly the signal that is already looking bad.
  • Authentication has to be right first. Warming up with unaligned DMARC builds no reputation, because the messages cannot be reliably attributed to the sender at all — see the alignment problem.

The core of it: deliverability is not a setting you apply but a standing you earn. The configuration is the ticket in. What happens afterwards is decided by the list.