Why this matters more than it used to
Receiving mail servers used to be forgiving. They are not any more. Google and Microsoft both tightened their requirements for bulk senders in 2024, and the practical effect for a small organisation is that mail from a domain with no authentication now gets treated with suspicion by default. It does not bounce, which is the problem: it disappears quietly into junk, and you find out weeks later when someone mentions they never got the quote.
The three records below are the whole of the fix in most cases. They are free, they take an afternoon, and once they are right they stay right.
What each record does
SPF: who is allowed to send as you
SPF is a list, published in DNS, of the servers permitted to send mail using your domain. A receiving server looks it up and checks whether the message came from one of them.
Two rules people get wrong. There must be exactly one SPF record for a domain: two records is a failure, not a merge, so every sender goes in the one line. And SPF is limited to ten DNS lookups, which sounds generous until a mail provider, a newsletter tool, a CRM and an invoicing system are all included, at which point the record silently stops evaluating.
The ending matters too. -all means anything not on the list is unauthorised. ~all means treat it as suspicious but do not reject. Start at ~all if you are not certain you have found every sender.
DKIM: a signature that proves the message is intact
DKIM puts a cryptographic signature in the header of every message you send. The matching public key sits in your DNS, so a receiving server can verify both that the message really came from your mail system and that nothing was altered along the way.
In Microsoft 365 and Google Workspace, DKIM is a switch you turn on plus a record you publish. It is off by default on a lot of tenants that were set up quickly, and turning it on is usually the single highest value change available.
DMARC: the policy, and the reports
DMARC does two jobs. It tells receiving servers what to do with mail that fails SPF and DKIM, and it asks them to send you reports about what is being sent in your name.
The policy has three settings. p=none means take no action but report. p=quarantine means send failures to junk. p=reject means refuse them outright. Everyone wants to end at reject. Almost nobody should start there.
The reports are the useful part, and they are the reason to publish DMARC even before you are ready to enforce anything. They tell you exactly which systems are sending as your domain, including the ones nobody remembers signing up for.
The three failures we see most
Some registrars and hosting panels create a DMARC record automatically, occasionally at quarantine. If your senders were never aligned, that record is instructing the world to bin mail you actually sent. We have found this at organisations who had no idea DMARC existed, and it is the fastest serious problem to fix.
Usually the result of two suppliers each adding their own. Sometimes it is one record that has accumulated so many includes that it has passed the ten lookup limit. Both fail in the same undramatic way: mail is authenticated for some recipients and not others, so the problem looks random.
The mail platform is the obvious one. Then there is the newsletter tool, the booking system, the accounting software sending invoices, the recruitment portal, and in community organisations, volunteers sending on the group’s behalf from their own accounts. Every one of them has to be either authorised or moved.
Checking yours in ten minutes
You do not need us for this part. On a Mac or Linux machine, in a terminal:
On Windows, in PowerShell:
What you are looking for: exactly one line starting v=spf1, and one starting v=DMARC1. If the DMARC line says quarantine or reject and you did not put it there, treat that as urgent.
For a score rather than a reading, send a message to a checking service such as mail-tester and it will report SPF, DKIM, DMARC and a handful of other issues in one page. That is the same tool we use to produce the before and after figures in the email deliverability check.
The order the fixes have to happen in
This sequence matters. Doing it out of order is how organisations break their own mail while trying to secure it.
Publish DMARC at p=none first and read the reports for a fortnight. They will find senders no list ever would.
Every legitimate sender in a single line, ending in ~all while you are still confirming the list.
On the mail platform and on every third party tool that offers domain authentication, not just the main one.
Volunteers and staff sending as the organisation from personal accounts need a mailbox or an alias, not an exception.
None to quarantine, then quarantine to reject, watching the reports between each move. Weeks apart, not minutes. If the reports look wrong, you go back a step, which is a decision you can only make if you were reading them.
What this does not fix
Authentication proves a message is genuinely from you. It does not make the message wanted. If you are sending to a list you bought, or to addresses that have not opted in, or with content that reads like a scam, correct DNS records will not save you. Reputation is built by people not marking your mail as junk, and no record can substitute for that.
It also does not fix a shared mailbox being used as a CRM, or a domain that is registered in a former employee’s name. Those come up on nearly every job we do, and they are separate problems with separate answers.
The email deliverability check is free, takes thirty minutes, and you leave with the findings in writing.
A before score, every sender we can find, what each of your three records is doing, and an ordered list of what to fix. Yours to keep.
Run my email check