Cold Email Deliverability: Set It Up Before You Send Anything

Most cold email fails in the plumbing rather than the copy. Here is what to put in place before the first send, and what quietly burns a domain afterwards.

Quick answer: Cold email deliverability is decided before you write a word: separate sending domains so your main one is never at risk, full authentication with SPF, DKIM and DMARC, a warm-up period on every new mailbox, verified addresses so you are not bouncing, and volumes low enough per mailbox that no single one looks industrial. Copy matters after all of that is true, not before.

Never send from the domain you get paid on

The first rule is the one people skip because it costs money and looks like paranoia. Do not run cold outreach from your primary domain.

If a campaign goes wrong, and eventually one will, the damage lands on the domain's reputation. When that domain is the one your invoices, contracts and client threads travel on, a bad month of outbound turns into deliverability problems with people who are already paying you.

Buy lookalike domains, point them at the main site, and send from those. They are cheap and they are the firewall.

Authenticate properly, then check that you did

SPF, DKIM and DMARC are not optional and have not been for years. The mistake is rarely skipping them. It is setting them once, assuming they took, and never verifying.

Send a message to a checking tool and read what actually arrives. A DKIM signature that does not validate, an SPF record with too many lookups, a DMARC policy left at none forever: all of these are invisible from your side and obvious from the receiving end.

Warm up every mailbox, including the fifth one

A brand new mailbox that starts sending dozens of messages to strangers looks exactly like what it is. Warm-up is the boring period where the mailbox builds a history of normal-looking traffic before it does anything else.

The failure here is impatience. Someone adds new mailboxes mid-campaign to raise volume, skips the warm-up on those because the campaign is already running, and cannot understand why reply rates fell across the board.

Verify the list, every time

Bounces are the fastest way to be classified as a sender nobody wants. A bought list is full of addresses that stopped existing when people changed jobs, and a scraped one is full of guesses.

Verification is cheap relative to what a burnt domain costs. Do it immediately before sending, not when the list was built, because the two are usually weeks apart.

Keep the volume per mailbox unremarkable

Total volume is a function of how many mailboxes you have, not how hard you push each one. A mailbox sending a modest number of messages a day, spread out, with pauses, looks like a person. The same total split across too few mailboxes looks like software.

This is also why the answer to wanting more pipeline is rarely to send more from what you have. It is either more sending capacity, properly warmed, or a better list.

The things that get you filtered after a good start

Deliverability is not a setup you finish. It degrades, usually for reasons nobody logged.

Links and images in the first message make it look like marketing rather than a personal note. Long threads of follow-ups to someone who never replied accumulate negative signals. Sending to a role address that forwards to five people multiplies your complaint risk. And the single biggest one: continuing to send to a segment that is clearly not responding, because the campaign was set up once and nobody stopped it.

Where the copy comes in

Once all of the above is true, copy does the work that everyone assumes it was doing all along.

But it is worth being clear about the order. A perfect message from an unauthenticated, unwarmed mailbox to an unverified list will not be read, because it will not arrive. A plain, honest message from clean infrastructure to a list built on a real reason will. The plumbing is not the unglamorous part of outbound. It is the part that decides whether the rest of it exists.