Cold email sender reputation

The five signals that matter

Receiving providers do not publish their models, but the inputs are visible in your own data. Authentication decides whether the message is trusted at all. Complaints decide whether it is wanted. Bounces decide whether your list is clean. Volume consistency decides whether you look automated. Engagement, where a provider can observe it, decides whether recipients read you.

A reputation dashboard is a summary of those five, often delayed and rarely specific to your domain. Treat it as a hint, not a verdict. The primary evidence is in your own sending logs: what bounced, what complained, what replied.

• Authentication: SPF, DKIM and DMARC aligned on every message.
• Complaints: the fastest way to lose a domain; one bad batch can outweigh weeks of clean sending.
• Hard bounces: a list problem that looks like a sender problem.
• Volume pattern: steady beats spiky at identical totals.
• Engagement: replies and deliberate clicks, not opens.

If you cannot answer the authentication question, start with SPF, DKIM and DMARC and fix that first.

Why opens are the wrong metric to optimise

Open tracking works by embedding a remote image. Mail clients increasingly block that image, and privacy features pre-fetch it, so an open can be recorded without a person seeing anything, or missed when they read every word. Optimising for opens pushes you toward larger images and more tracking, which are the opposite of what filters reward.

Measure replies instead. A reply is unambiguous, and it is the outcome a cold email is actually for. If a subject-line change raises opens and lowers replies, the change is a loss.

• Track replies per hundred delivered messages, not opens.
• Compare subject lines by reply rate over at least a few hundred sends.
• Disable open tracking on the first contact if your provider allows it.

For a method that survives tracking changes, see measuring reply rate without guessing.

Reputation is domain-level, so containment matters

Your main domain carries support, billing and transactional mail. Cold outreach has different complaint dynamics. Sending both from the same domain means one bad campaign can affect password resets. Use a dedicated outreach domain or subdomain so the blast radius is limited to campaigns.

Containment also means one mailbox, one campaign. Running two campaigns from the same mailbox doubles the apparent volume and mixes two audiences, so a complaint about one offer affects the other. Parallel mailboxes keep the volume plausible and the reputation separable.

Separate mailboxes are the practical way to do this; SMTP accounts for outreach supplies inbox-ready accounts on your own domain.

Recovering a reputation that has slipped

The first move is always subtraction: stop the campaign, remove the segment that bounced or complained, and confirm authentication still passes. Sending more mail to prove the domain is fine has never worked, because every additional message during a bad period reinforces the pattern.

Then resume slowly on the cleanest segment, at a lower volume, and watch replies and bounces before widening. Worked example: a sender at 600 messages a day dropped to 80 after a complaint spike. Two days at 80 produced no further complaints, so the pace was raised by 50 each day. The domain recovered without a new domain or a new IP.

• Stop the campaign immediately; do not finish the batch.
• Remove the segment that produced complaints or bounces.
• Resume at a fraction of the previous volume with the best-performing message.
• Raise volume only after a full day of clean signals.

Signals you can see before a provider acts

You do not need a reputation score to notice a change. The first evidence usually appears in three places you already own: the bounce log, the deferral pattern and the tone of replies. A domain that starts deferring a larger share of messages at the same volume is telling you something, even if no dashboard has updated.

Keep a daily note of accepted, deferred and rejected counts per mailbox. Three numbers a day take a minute and give you a trend line. When rejection rises while the content is unchanged, the problem is list or sender history rather than copy.

• Accepted, deferred and rejected per mailbox, logged daily.
• Reply tone: irritation is an early complaint signal.
• Bounce clustering by recipient domain, not just a total.

Signals, where to read them, and the response

A single 5xx aimed at a recipient address is a list problem, not a sender problem.

Frequently asked questions