Skip to content
Part 1, chapter 3 · Outbound Fundamentals · 2026-08-03

Why Cold Email Goes to Spam (and How to Fix It)

You can write the best cold email of your life and have it filtered before a human sees it. Deliverability is the layer underneath it all.

Three different things get called deliverability and conflating them is why most diagnoses go wrong.

  • Delivered, the receiving server accepted the message. It did not bounce. This tells you almost nothing.
  • Inbox placement, it landed in the primary inbox rather than spam or promotions. This is what you actually care about.
  • Opened, a human looked at it. Increasingly unmeasurable, because privacy proxies fire open pixels automatically.

A campaign can show 98% delivered and near-zero inbox placement. Every sending tool reports the first number prominently and the second not at all, because the first is flattering and the second is hard to measure.

The three records, and what they actually do

SPF, DKIM and DMARC are authentication, not reputation. They prove the mail came from where it claims. Setting them up correctly does not make you trusted, it makes you identifiable, which is a prerequisite for being trusted and also for being reliably blocked.

RecordWhat it provesWhat it does not do
SPFThis server is allowed to send for this domainSay anything about content or reputation
DKIMThe message was not altered in transitVouch for the sender's behaviour
DMARCWhat to do when SPF or DKIM failsImprove placement on its own

All three are table stakes. Configure them and you have removed a reason to be blocked; you have not given anyone a reason to trust you.

What actually drives placement

Engagement, mostly. Filters watch what recipients do: opens, replies, deletions without reading, and above all spam complaints. A domain that consistently produces replies keeps landing in inboxes. A domain that produces silence and the occasional complaint stops.

This is the part vendors will not tell you. If your offer is irrelevant to the list you bought, engagement will be low, and low engagement is the signal that gets you filtered. No amount of warmup, domain rotation or infrastructure fixes an offer nobody wants, it just spreads the damage across more domains before it catches up with you.

What Spamhaus thinks you are doing

Worth knowing before you spend money on this: Spamhaus, whose blocklists a large share of the internet consults before accepting mail, publishes its position on cold email, and the position is that unsolicited commercial bulk mail is spam. Not spam-adjacent. Spam.

Every vendor page in this category frames deliverability as a technical problem with a technical solution. The organisation that operates the filters frames it as a consent problem. Both descriptions are accurate about different things, and the gap between them is the actual risk you are managing.

The practical consequence. You are not optimising toward a system that wants to let you through. You are operating inside tolerances set by people who consider the activity itself illegitimate, and who adjust those tolerances when volume rises. That is why deliverability tactics decay, not because the tactic stopped working, but because working at scale is the thing being filtered.

Warmup, and what it is really for

Warmup services send mail between networks of accounts and mark it as read and replied. The stated purpose is establishing a positive sending history on a new domain. That is real, and it works for the specific case of a domain with no history at all.

What it cannot do is repair reputation while you are actively damaging it. Running warmup alongside a campaign generating complaints is applying and removing the same signal simultaneously, and the campaign is louder.

The setup that survives

  • Send from a separate domain to your main one. A burned outbound domain must not take your company email with it.
  • Authenticate all three records before the first send.
  • Start low — tens of mails a day per mailbox, not hundreds, and increase over weeks.
  • Verify the list. Bounces are the fastest way to signal you do not know who you are mailing.
  • Make replying easy and one-click unsubscribing easier. An unsubscribe is vastly cheaper than a complaint.
  • Watch reply rate, not open rate. It is the metric filters weight and the one you can still measure honestly.

If reply rate sits under about 2% on a clean, well-targeted list, the problem is upstream of sending. Stop optimising infrastructure and go back to the offer.

Questions

Why are my cold emails going to spam?
Usually low engagement rather than technical misconfiguration. Check authentication first because it is cheap to fix, but if SPF, DKIM and DMARC are correct and mail still lands in spam, the list or the offer is the cause.
How do I stop cold emails going to spam?
Authenticate properly, send from a separate domain, ramp volume over weeks rather than days, verify the list, and make unsubscribing trivially easy. If reply rate stays under 2% after all that, the problem is the offer and no sending change will fix it.
Is cold email still effective in 2026?
For well-targeted lists into segments that genuinely have the problem, yes. For volume sprayed at a bought list, decreasingly — filters have got better at exactly that pattern, and Spamhaus classifies it as spam outright.
Do I need a separate domain for cold email?
Yes. Outbound sending can damage domain reputation, and you do not want that spreading to the domain your invoices and support mail come from. Use a lookalike domain dedicated to outbound.
Does email warmup actually work?
For establishing history on a genuinely new domain, yes. For repairing reputation while a campaign is generating complaints, no, you are adding and removing the same signal at once.
How many emails a day can I send per mailbox?
Start in the tens and increase over weeks. The exact ceiling matters less than the ramp: sudden volume changes are themselves a spam signal regardless of the absolute number.
Is open rate still a useful metric?
Barely. Privacy proxies fire tracking pixels automatically, which inflates opens without a human involved. Reply rate is the metric worth optimising and the one filters actually weight.
What reply rate should I expect?
Under 2% on a clean, well-targeted list means the offer or targeting is wrong, not the infrastructure. That threshold is more diagnostic than any deliverability dashboard.

Tools mentioned

Sources

Source interests are labelled. Almost everything published about this subject is written by someone selling into it.

← All chapters