The deliverability checklist, in the order that matters

This is the highest-impact decision on the list and it is made once. A bad month of cold outbound on your root domain damages the reputation your invoices, password resets and customer replies depend on.

Attach a subdomain. If the whole programme goes wrong, you retire a subdomain rather than your company.

SPF, DKIM and DMARC. All three, published, and checked from outside rather than assumed from the fact that you pasted them in.

None of them makes your mail wanted. What they buy is that your legitimate mail is not discarded for looking forged.

Ten a day in week one, twenty-five to day thirty, fifty to day sixty, and only then your real target. Every deliverability disaster begins with somebody deciding that this is too slow.

The ramp is per sending address, not per domain. Reputation attaches to the sending identity.

Hard bounces damage reputation disproportionately. A five percent bounce rate is not five percent less delivery. It is a reputation event that affects the other ninety-five.

Verify before the first send, suppress on the first hard bounce, and never retry a hard bounce.

A three percent bounce rate that has been three percent for a month is fine. A one percent rate that was point-two last week is a problem. The same is true of complaints.

Spikes should re-anchor your ramp automatically. Continuing to send at target volume through a reputation event is how a recoverable problem becomes a dead domain.

Fewer links, no tracking pixel, plain text over HTML, and no spam-trigger-word lists. Those have been folklore for a decade. Content matters at the margin. The five items above matter more, in order.