Most DMARC guidance stops at the p tag, the policy for your organizational domain. But spoofers do not limit themselves to your exact domain; they love subdomains, including ones that do not even exist, because a convincing name like billing.yourbrand.com inherits your reputation while slipping past a policy that was only ever thinking about the bare domain. DMARC has two tags for exactly this: sp for subdomains and np for non-existent subdomains. This guide explains what each one controls, how they interact with the main policy, and how to set them without accidentally breaking a legitimate subdomain sender.
Why a subdomain policy exists at all
DMARC is published at the organizational domain, for example in the TXT record at _dmarc.yourbrand.com. By default, that policy applies to the organizational domain and, through inheritance, to every subdomain that does not publish its own DMARC record. The problem is that inheritance is all or nothing: if you set p=reject but one subdomain legitimately sends mail you have not finished authenticating, you either bounce that mail or you weaken the whole policy. The sp tag breaks that deadlock by letting you set a different policy for subdomains than for the top-level name.
What sp controls
The sp tag sets the policy applied to mail that fails DMARC and is sent from a subdomain of your organizational domain, when that subdomain has no DMARC record of its own. It takes the same three values as p: none, quarantine, and reject. If you omit sp, subdomains simply inherit whatever p is. That inheritance is the default behavior defined in RFC 7489 (as of November 2026), so you only add sp when you want subdomains treated differently from the parent.
v=DMARC1; p=reject; sp=quarantine; rua=mailto:dmarc@yourbrand.com
In that record, forged mail from the exact organizational domain is rejected, while failing mail from an unaudited subdomain is quarantined instead, giving you a softer landing while you finish bringing subdomains up to standard.
What np adds
The np tag, introduced in RFC 9091, sets the policy specifically for non-existent subdomains, meaning subdomains that have no A, AAAA, or MX records at all. This matters because a huge share of subdomain spoofing uses invented names that were never provisioned, such as secure-login.yourbrand.com. Legitimate mail never comes from a name that does not exist, so you can safely set np=reject even while you are still cautious about real subdomains. It closes the invented-name attack path without touching your working subdomains.
How the three tags combine
The receiver picks exactly one policy for a given message based on where it originated and whether that name exists. The precedence is specific before general: np applies to a non-existent subdomain if present, otherwise sp applies to any subdomain, otherwise everything falls back to p.
| Mail originates from | Policy applied | Falls back to |
|---|---|---|
| Organizational domain | p | none implied if absent |
| Existing subdomain, no own DMARC | sp | p if sp absent |
| Non-existent subdomain | np | sp, then p |
| Subdomain with its own DMARC record | that record's p | not inherited |
One point deserves emphasis: if a subdomain publishes its own _dmarc record, that record wins outright and none of the parent's sp or np applies to it. Delegated subdomains are responsible for their own policy.
Setting the tags safely, step by step
The goal is to lock down the organizational domain and non-existent names aggressively while giving real subdomains room to catch up. Work in this order.
- Start the whole domain at
p=nonewith aruaaddress so you receive aggregate reports and can see which subdomains send mail. - Read the reports until you have a complete list of legitimate subdomain senders and whether each passes with alignment.
- Add
np=rejectearly. No real mail comes from non-existent subdomains, so this is low risk and closes the most common spoofing path. - Set
sp=nonewhile real subdomains are still being fixed, so tightening the parent does not bounce their mail. - Raise the parent to
p=quarantinethenp=rejectonce its own sources are clean. - Finally raise
sptoquarantineand thenrejectonce every legitimate subdomain authenticates correctly.
Common mistakes with sp and np
A few errors recur. The first is assuming subdomains are unprotected because you only set p; in fact they inherit p by default, so the risk is the opposite, that a strict p is already rejecting legitimate subdomain mail you forgot about. The second is setting sp=none and then forgetting to remove it, leaving every subdomain permanently open to spoofing long after they were ready for enforcement. The third is confusing the two tags: np is only for names that truly do not exist, so it will not help with a real but unaudited subdomain, which is sp's job. Finally, remember that a subdomain with its own DMARC record ignores the parent entirely, so a forgotten lax record on a subdomain can undo your careful parent policy.
Before trusting any of this, confirm the published record actually contains the tags you intend and has propagated. A quick way is to run the domain through the Valla DNS domain health report, which reads the live DMARC record and flags gaps across the whole domain rather than one resolver's cache.
If I only set p, are my subdomains protected?
Yes, by inheritance. Subdomains without their own DMARC record inherit the p policy, so a p=reject already covers them. The caution is the reverse: an inherited strict policy can reject legitimate subdomain mail you have not authenticated, which is exactly what sp lets you soften.
What counts as a non-existent subdomain for np?
A name with no A, AAAA, or MX records. Because real senders are never on a name that does not resolve, np=reject is safe to deploy early and specifically targets invented lookalike subdomains used in spoofing.
Does sp apply to a subdomain that has its own DMARC record?
No. A subdomain publishing its own _dmarc record uses that record's policy and ignores the parent's sp and np entirely. Delegated subdomains own their policy.
Can I use np without sp?
Yes. The tags are independent. You can publish np=reject to shut down invented subdomains while leaving sp absent so existing subdomains keep inheriting p. Receivers apply whichever tag matches the message.
Why would I ever set sp weaker than p?
To ramp safely. If the organizational domain is fully authenticated but some real subdomains are not yet, sp=none or sp=quarantine lets you enforce on the parent without bouncing subdomain mail, then you tighten sp once those senders are clean.
Subdomain policy is the half of DMARC that generic advice skips, yet it is where a great deal of real spoofing happens. Set np=reject early, use sp to ramp subdomains independently, and never leave a loosening tag in place longer than you need it. For the main policy ramp this builds on, see our guide to the DMARC policy ramp, and for why some mail fails even when the policy is right, our guide to DMARC alignment.