There are two ways to get DMARC wrong, and most organizations pick one. The first is to publish p=none, feel responsible, and leave it there for six years while forgers continue to use the domain undisturbed. The second is to read an alarming article on a Friday, publish p=reject before lunch, and discover on Monday that payroll notifications, the customer newsletter, and every invoice since Friday afternoon were refused at the door.
The correct route runs between them. It has four stages, a checkpoint after each, and some housekeeping at the end. Followed in order, it gets a typical domain to full enforcement in two to three months without losing legitimate mail. The work is not difficult. It is mostly counting.
What p=reject buys you, and what it does not
At p=reject, a receiver that honors DMARC refuses mail that claims to be from your exact domain and cannot prove it. That shuts down the cheapest and most convincing forgery there is: a message whose From address is really, letter for letter, yours. It also satisfies the sender requirements of large mailbox providers and is a precondition for displaying a logo through BIMI, which you can test with the BIMI checker.
It does not stop lookalike domains, display-name tricks, or mail from a genuinely compromised account. Those need other defenses, described in Domain Impersonation Protection. And enforcement is carried out by the receiving server, not by you and not by any monitoring service. Your record is a request. Most large receivers honor it.
The three policies compared, briefly:
| Policy | What receivers do with failing mail | Use it for |
|---|---|---|
p=none | Deliver as usual; send reports | Discovery |
p=quarantine | Treat as suspicious, usually the spam folder | The rehearsal |
p=reject | Refuse during delivery | The destination |
For more on choosing among them, see None vs Quarantine vs Reject.
Stage 1: Publish p=none and turn on reports
Begin with a record that changes nothing and reports everything:
_dmarc.example.com. TXT "v=DMARC1; p=none; rua=mailto:dmarc@example.com"The rua tag is the reason this stage exists. It asks receivers to send daily aggregate reports listing every source that sent mail as your domain and how each fared. Without it, p=none is decoration. Build the record with the DMARC generator and confirm it with the DMARC checker.
Then wait. Observe for at least two to four weeks, and a full month if you can, so that monthly statements and occasional campaigns appear in the data. Reports arrive as zipped XML from each participating receiver. To have them collected and translated, follow the DMARC reporting setup guide.
Checkpoint: reports are arriving from several receivers, covering a complete business cycle.
Stage 2: Inventory every sender
Now make a list. Every source in the reports goes into one of four columns:
- Known and aligned. Your mailbox provider, properly configured. Nothing to do.
- Known and misaligned. Legitimate services whose mail fails DMARC. This is your work list.
- Forwarders and mailing lists. Your real mail, relayed by someone else’s server. More on these below.
- Unknown. Either a forgotten system or someone impersonating you. Investigate before deciding which.
The second column is always longer than expected. It contains the CRM, the help desk, the invoicing system, the survey tool, the applicant tracker, the web server’s contact form, and the copier that scans to email. Ask every department what they use. The reports will then tell you what they forgot.
SpoofKiller’s DMARC monitoring does this sorting for you, naming recognized services and separating likely forwarding from likely abuse. Remember that reports come from participating receivers only. They are a large sample, not a complete ledger.
Checkpoint: every source with meaningful volume has a name and an owner.
Stage 3: Fix alignment, one sender at a time
A message passes DMARC when SPF or DKIM passes for a domain that aligns with your From domain. For each service on the work list, arrange at least one:
- Turn on custom-domain DKIM. Nearly every reputable platform offers it. You publish a key or CNAME under your domain, and the service signs as you. Do this first; it survives forwarding.
- Set a custom Return-Path where offered, so SPF aligns too.
- Clean up SPF. One record, under ten lookups, no dead includes. The SPF checker counts them.
Provider-specific steps are in the guides for Google Workspace, Microsoft 365, Postmark, and SendGrid. The reasoning behind them is in DMARC Alignment Explained.
A service that cannot sign with your domain should not be sending as your domain. Move it to a dedicated subdomain or replace it.
Checkpoint: for two consecutive weeks, every legitimate sender shows an aligned pass in the reports.
Stage 4: Quarantine in measured doses, then reject
Now enforce, gently. The pct tag applies the policy to a percentage of failing messages, leaving the rest to the next-lower treatment. Start at a quarter:
v=DMARC1; p=quarantine; pct=25; rua=mailto:dmarc@example.com
v=DMARC1; p=quarantine; pct=50; rua=mailto:dmarc@example.com
v=DMARC1; p=quarantine; pct=100; rua=mailto:dmarc@example.comHold each step for a week or two. Quarantine is the forgiving stage: a legitimate message you missed lands in a spam folder, where it can be found and the sender fixed. After each increase, read the reports and listen for complaints. If something breaks, drop back a step, repair it, and resume.
Treat pct as a rough dial, not a precise instrument. It applies only to mail that fails, and receivers are not obliged to sample exactly.
Checkpoint: at pct=100, no legitimate source appears among the quarantined mail.
The last step: reject
The last change is one word. You may ramp it as before; under p=reject with a partial percentage, the unsampled failures are quarantined instead of rejected.
v=DMARC1; p=reject; pct=25; rua=mailto:dmarc@example.com
v=DMARC1; p=reject; rua=mailto:dmarc@example.comWhen pct is omitted it defaults to 100. Keep the rua tag. You will need it more after enforcement than before.
Reject is the only policy with consequences for forgers. It is also the only one with consequences for you. Hence the paperwork.
Subdomains and domains that send nothing
A policy on example.com covers its subdomains unless you say otherwise. The sp tag sets a separate subdomain policy. During rollout you can use it to hold busy subdomains at a gentler setting or, more commonly, to lock unused ones down first.
v=DMARC1; p=quarantine; sp=reject; rua=mailto:dmarc@example.comForgers are fond of subdomains that do not exist, such as billing.example.com. If you send no mail from subdomains, sp=reject costs nothing and can be set early.
Then the parked domains. Every company owns a drawer of them: the old brand, the typo, the .net. None sends mail, so all can go to maximum enforcement today, with no waiting period. Each one gets three records:
parked-example.com. TXT "v=spf1 -all"
_dmarc.parked-example.com. TXT "v=DMARC1; p=reject; rua=mailto:dmarc@example.com"
parked-example.com. MX 0 .The first says no server may send for this domain. The second tells receivers to refuse anything that tries. The third is a null MX, which declares that the domain accepts no mail either.
What will still fail, and how to stay at reject
At full enforcement your reports will not show zero failures. They never will. Two kinds of legitimate mail fail for reasons beyond your control:
- Forwarded mail. A recipient forwards to another mailbox. SPF fails because a new server is sending. If your DKIM signature is intact, DMARC still passes, which is why Stage 3 put DKIM first.
- Mailing lists. A list adds a footer or subject tag, breaking the DKIM signature, and resends from its own server, breaking SPF. Both fail. Many lists now rewrite the From address to their own domain to avoid this.
A small, steady residue of such failures is normal and is not evidence of an attack. What you are watching for is change: a new source, a sudden rise, a trusted sender that stops aligning. The same goes for a new vendor someone adds next quarter without telling you.
Staying at reject
Reaching p=reject is a project. Staying there is a habit. DNS records get edited, keys expire, and providers change their infrastructure. Under enforcement, each of those mistakes now bounces real mail.
SpoofKiller tracks this for you. Its enforcement readiness status shows green, yellow, or red for each domain, based on how much of your legitimate mail is aligned, so you know when each stage’s checkpoint has been met. After enforcement, scheduled DNS checks and report analysis flag the regressions. The free plan covers one domain; larger setups are on the pricing page.

