Moving DMARC from p=none to p=reject safely
Updated
p=none only collects reports. p=quarantine and p=reject are what stop others from sending as your domain. The risk in moving up is blocking your own mail from a service you forgot. The way to avoid that is to let the aggregate reports tell you when each step is safe.
Before you start
You need a DMARC record with a reporting address, and a way to read the reports (a report service, or the XML itself; see How to read DMARC aggregate reports):
_dmarc.example.com TXT "v=DMARC1; p=none; rua=mailto:[email protected]"
Keep the default relaxed alignment (adkim=r, aspf=r). Strict alignment breaks mail that is signed by mail.example.com but sent as example.com, and gives little in return.
Step 1: p=none, collect and fix (2 to 6 weeks)
Let reports arrive for at least two weeks, better a full month, so monthly mail (invoices, statements, newsletters) shows up. Then list every source that sends with your domain in From: and sort it into three groups:
- Yours, passing: DMARC passes on DKIM or SPF. Nothing to do.
- Yours, failing: a real service (CRM, helpdesk, billing, website form) that fails. Set up DKIM with your domain at that service. SPF alone is weaker: it breaks on forwarding, and every
include:counts towards SPF's limit of 10 DNS lookups. - Not yours: unknown IPs that fail both, forwarders and mailing lists. Leave them. Spoofing is what the policy is for, and forwarded mail usually still passes on DKIM.
Ready for the next step when: every source you recognise passes, over at least two weeks of reports, and what still fails is spoofing, forwarding or lists. Check that the reports come from more than one receiver (Google, Yahoo, Microsoft and others), since each sees different mail.
Step 2: p=quarantine (2 to 4 weeks)
v=DMARC1; p=quarantine; rua=mailto:[email protected]
Failing mail now goes to spam instead of the inbox. A mistake costs a trip to the spam folder, not a lost message, which is why this step comes first.
You may have used pct= to quarantine only part of the failing mail. RFC 9989 retired pct, and receivers that follow it ignore the tag. Instead it adds t=y, which marks the policy as being tested. Do not build the rollout on either: go to p=quarantine for everything when step 1 is done.
What to watch: in the reports, rows with disposition quarantine. Each should be a source from the "not yours" group. A source you recognise there is a service you missed: fix it, or go back to p=none while you do. Also ask colleagues whether replies from customers mention mail in spam.
Ready for the next step when: two weeks or more with no quarantined mail from your own services.
Step 3: p=reject
v=DMARC1; p=reject; rua=mailto:[email protected]
Receivers now refuse failing mail. The sender gets a bounce (Gmail's is 550 5.7.26, see why Gmail and Outlook reject your mail).
What to watch: rows with disposition reject from your own sources, and bounce reports from your email services. A disposition of none with a reason such as forwarded or local_policy is the receiver overriding the policy, which is normal.
Subdomains
Without sp=, subdomains get the same policy as the domain. That is what you want unless a subdomain has its own senders that are not ready. RFC 9989 also has np= for subdomains that do not exist, which you can set to reject at once: no legitimate mail comes from a name that is not in DNS.
v=DMARC1; p=reject; sp=reject; np=reject; rua=mailto:[email protected]
If you need a weaker sp= for a while, write down why and when it goes back up. A temporary sp=none tends to stay.
After p=reject
Keep reading the reports. Every new tool that sends mail as you (a new CRM, a survey tool, an agency) has to be set up with DKIM before it goes live, or its mail is rejected from day one. The other risk is someone setting the policy back to p=none to debug a problem and leaving it there.
Senderward checks your domain's DMARC, SPF, DKIM, MX, MTA-STS and TLS-RPT every day and emails you, with the fix, when something gets worse. $7 per domain per month, 14 days free. See pricing