September 15, 2026 · Admin
What DMARC Policy for Cold Email? p=none on the Sending Domains, p=reject on the Domain You Run On
Three DMARC policies exist, and on a cold email sending domain the right one is p=none. Gmail and Yahoo both require a DMARC record from senders above 5,000 messages a day, and both state plainly that p=none satisfies the requirement. Enforcement belongs on the domain your company actually runs on, where somebody is spoofing you, not on the throwaway domains carrying your outbound.
This is the third article in our deliverability series. If the records themselves are not in place yet, start with the 40-minute SPF, DKIM and DMARC setup, and if mail has already stopped landing, read the eleven causes of cold email going to spam first. Everything below is from published mailbox provider documentation and the IETF specification, checked on 15 September 2026. One thing changed this year that most of the advice online has not caught up with: the pct tag no longer exists in the standard. Every page ranking for this query today was published by a DMARC or cold email tool vendor, and several of them still tell you to start at pct=10.
What DMARC policy do Gmail, Yahoo and Outlook actually require?
A record, at minimum p=none, plus alignment. Gmail’s sender guidelines state that for senders of 5,000 or more messages a day to personal Gmail accounts, “your DMARC enforcement policy can be set to none,” and that the From: header domain must align with either the SPF domain or the DKIM domain. Yahoo’s requirement is worded almost identically: publish a valid DMARC policy with at least p=none, and align the From: domain with SPF or DKIM. Neither one asks for enforcement.
What they do enforce is the complaint rate. Gmail publishes a 0.30% ceiling on spam rates reported in Postmaster Tools and separately advises staying under 0.10%. Yahoo publishes the same 0.30% figure. Your DMARC policy is a one-time DNS decision; the spam rate is the number that actually decides whether your cold email lands next month.
| Provider | DMARC required? | Minimum policy | Spam rate ceiling |
|---|---|---|---|
| Gmail (5,000+/day) | Yes | p=none accepted |
0.30%, advises under 0.10% |
| Yahoo (bulk senders) | Yes | p=none accepted |
0.30% (inbox-delivered mail) |
| Microsoft 365 / Outlook | Yes, for high-volume senders | Not lowered below p=none in published docs |
Not published as a number |
Does p=reject improve cold email deliverability?
No mailbox provider publishes a rule that gives p=reject better inbox placement, and Gmail’s own requirement document says the opposite by accepting p=none. The p tag is not a trust score. It is an instruction to receiving servers about what to do with mail that fails authentication while claiming to be from your domain. If your own mail passes SPF, DKIM and alignment, your policy value never gets consulted for it.
RFC 9989, the current DMARC specification, is explicit about what the three values mean. none means “the Domain Owner offers no expression of preference.” quarantine means the owner “considers such mail to be suspicious.” reject means the owner “considers all such failures to be a clear indication that the use of the domain name is not valid.” All three describe failing mail. None of them describes your mail.
This matters because the cold email industry has spent two years selling p=reject as a deliverability upgrade. It is a spoofing control. The four things that move cold email inbox placement are authentication passing at all, sending volume per mailbox, list quality, and complaint rate, and the last one is the only one with a published threshold attached.
Then why does anyone publish p=reject?
Two real reasons, neither of which applies to a sending domain you bought last month. The first is spoofing: p=reject on your brand domain stops phishing mail that forges your From: address from reaching your customers and your own staff. That is worth doing, and it is why the domain on your business cards should be at enforcement.
The second is BIMI. The BIMI Group’s implementation guide states that the organizational domain and its subdomains must be at p=quarantine or p=reject, and notes that “‘None’ policies or ‘pct’ less than 100 percent are not accepted.” If you want your logo in the Gmail avatar slot, enforcement is the entry fee. That is a brand domain project, not a cold outbound one, since nobody is putting a verified mark on a domain named after a hyphenated variant of their company.
There is also a cost to enforcement that the vendor pages skip. RFC 9989 warns domain owners publishing an enforcement policy to first understand “the interoperability issues involved and problems that can result for mailing lists and for delivery of legitimate mail.” Forwarding breaks SPF. If a recipient auto-forwards your message and the DKIM signature does not survive the hop, a p=reject domain gets the message refused outright instead of filtered. On a cold campaign, that failure is invisible to you.
Why has pct=10 stopped being valid advice?
Because the tag was deleted from the specification. RFC 9989, published in May 2026, obsoletes both RFC 7489 and RFC 9091, and Appendix A.6 is titled “Removal of the ‘pct’ Tag.” Percentage-based sampling is gone. In its place is the t tag, which takes y or n, defaults to n, and signals whether the owner wants the declared policy actually applied.
The behaviour of t=y is a one-step downgrade rather than a percentage. Under RFC 9989, a record of p=quarantine; t=y asks validators to apply none instead, and p=reject; t=y asks them to apply quarantine. So the modern test-mode record is a policy plus t=y, not a policy plus a percentage.
Worth knowing before you go rewrite your DNS: vendor documentation has not caught up. Google’s recommended rollout still shows p=quarantine; pct=5, and Microsoft’s DMARC configuration page still walks through pct=10, 25, 50, 75, 100, both checked 15 September 2026. A leftover pct tag is not going to break your mail, because receivers that implement the old behaviour keep honouring it and receivers that implement RFC 9989 ignore an unknown tag. It just no longer does anything you can rely on.
What should each domain in a cold email stack publish?
Different records for different jobs. A cold email operation typically runs one brand domain that carries the website and the real mailboxes, plus somewhere between three and twelve lookalike sending domains that carry the outbound. Those two categories want opposite policies.
| Domain | Record to publish | Why |
|---|---|---|
| Brand domain, root | v=DMARC1; p=reject; rua=mailto:dmarc@yourdomain.com |
Stops forged mail using your real name. Required for BIMI. |
| Brand domain, non-existent subdomains | Add np=reject |
RFC 9989 adds np for subdomains that do not exist, a favourite of spoofers |
| Brand subdomain used by an ESP | Its own record, p=none at first |
Subdomains inherit the parent policy unless they publish their own |
| Cold email sending domains | v=DMARC1; p=none; rua=mailto:dmarc@yourdomain.com |
Meets Gmail and Yahoo minimums with zero risk of self-inflicted rejection |
| Parked domains that send nothing | v=DMARC1; p=reject; |
Nothing legitimate sends from them, so there is nothing to break |
sp is absent, the p policy applies to subdomains.The rua address is the part people skip, and it is the only part that produces information. Without it you get a compliant record and no reports, which means no way to discover that a tool you forgot about has been sending unaligned mail as you for six months. RFC 9989 keeps the daily default reporting interval, so reports arrive within about 24 hours of publishing the address.
How do you move from p=none to enforcement without losing mail?
Slowly, and on the brand domain only. Google’s published rollout is one week at p=none reading reports daily, then p=quarantine at a small sample, then increasing coverage over time. Google suggests a smaller organisation might begin at 10% and a larger one at 1%. Microsoft documents the same three phases and advises starting with a low-volume subdomain, repeating the phases, and saving the parent domain for last.
| Stage | Record (RFC 9989 syntax) | Effect on failing mail | Time at this stage |
|---|---|---|---|
| 1. Monitor | v=DMARC1; p=none; rua=mailto:dmarc@yourdomain.com |
Nothing blocked, daily reports arrive | At least 1 week (Google), 14 days is our practice |
| 2. Test enforcement | v=DMARC1; p=quarantine; t=y; rua=... |
Validators apply none, so still nothing blocked |
14 days |
| 3. Quarantine | v=DMARC1; p=quarantine; rua=... |
Failing mail goes to Junk. Microsoft logs dmarc=fail action=quarantine |
14 to 30 days |
| 4. Reject | v=DMARC1; p=reject; np=reject; rua=... |
Failing mail refused at SMTP with 550 5.7.1 |
Ongoing |
Microsoft’s documentation is worth quoting on the reason for the pace: “Blocking legitimate email in any significant volume is unacceptable to users, but it’s almost inevitable that you’re going to get some false positives. Go slowly and methodically deal with issues that are revealed in DMARC reporting.” The reports are the work. The DNS edits take four minutes.
What breaks when you get this wrong?
Inheritance, almost every time. A DMARC record covers every subdomain that does not publish its own, which is Microsoft’s stated behaviour and RFC 9989’s default when sp is absent. So a marketing team moving the root domain to p=reject can silently kill mail from mail.company.com that an ESP has been sending unaligned for two years. The fix is either sp=none while you sort the subdomains out, or a dedicated record on each sending subdomain.
The second failure is the mirror image of it: a sending domain that never got its own record at all and inherits nothing, because a bare second-level domain has no parent to inherit from. That domain has no DMARC record, which means it fails the Gmail and Yahoo bulk sender requirement outright. If you run nine sending domains, that is nine separate _dmarc records, and the ninth one is the one nobody publishes.
The third is the record that was never read. A p=none record with a rua address nobody monitors is functionally a compliance checkbox. It passes Gmail’s requirement and it teaches you nothing, which is fine on a cold email domain you will retire in a year, and a wasted control on the domain your company runs on.
How LeadButton handles this
Every sending domain we run publishes v=DMARC1; p=none with a rua address we monitor, and it stays there. We do not put client sending domains at enforcement, because the requirement does not ask for it and the downside is refused mail we would not see. On the client’s brand domain we do the opposite: read the aggregate reports, get every legitimate sender aligned, then walk the four stages above to p=reject with np=reject. Reports are reviewed monthly against the four failure modes, and new domains get their records before the first prospect is contacted. Launch is $1,500 a month for email, Growth is $3,500 a month for email plus LinkedIn, and the DNS work is included in both rather than billed as setup. Full pricing is on the LeadButton pricing page.
Sources
- RFC 9989, Domain-Based Message Authentication, Reporting, and Conformance (DMARC), published May 2026, obsoletes RFC 7489 and RFC 9091
- Google, Email sender guidelines, checked 15 September 2026
- Google Workspace, Recommended DMARC rollout, checked 15 September 2026
- Yahoo, Sender best practices, checked 15 September 2026
- Microsoft Learn, Set up DMARC to validate the From address domain, checked 15 September 2026
- Microsoft, Outlook’s new requirements for high-volume senders
- BIMI Group, BIMI implementation guide, checked 15 September 2026
- Google, New Gmail protections for a safer, less spammy inbox, requirements effective February 2024
Leave a Reply