Publishing a DMARC record with p=none is a quick job. The real work is moving to a p=quarantine or p=reject policy without losing the month-end invoice, the CRM notification or the document scanned from the printer. This guide explains, with example DNS records, how to read aggregate reports, how to align sending sources, and how to roll out the policy domain by domain based on which measurements. As the current definition, it relies on RFC 9989, published in May 2026, which replaces RFC 7489.
How DMARC alignment works
DMARC (RFC 9989) doesn’t introduce a new authentication method. It ties the SPF and DKIM results to the domain in the From header that the user sees. For a message to pass DMARC, one of two conditions is enough:
- SPF alignment: SPF returns pass for the envelope sender (RFC5321.MailFrom, Return-Path in the headers), and that domain is aligned with the From domain.
- DKIM alignment: The
d=value of a valid DKIM signature is aligned with the From domain.
The aspf and adkim tags set the alignment mode. In the default relaxed (r) mode, a match of the organizational domain is enough: when the From address is [email protected], SPF verification with bounce.example.com counts as aligned. In strict (s) mode, an exact match is required. RFC 9989 determines the organizational domain through queries walking up the DNS (DNS Tree Walk) rather than the Public Suffix List; a receiver working under RFC 7489 may reach a different result for some domain structures. Staying in relaxed mode during the migration makes it easier to align services that use subdomains.
A commonly confused point: SPF being pass isn’t enough for DMARC. Most marketing services send messages with their own bounce domain; SPF passes, but because it isn’t aligned, it doesn’t count for DMARC. That’s the common reason for “SPF pass, DMARC fail” rows in reports.
Reading aggregate reports
Reports arrive at the address in the rua tag from receiving systems, usually once a day, as compressed XML. A starting record might look like this:
_dmarc.example.com. 3600 IN TXT "v=DMARC1; p=none; rua=mailto:[email protected]; adkim=r; aspf=r"
If you send reports to an address on another domain, for example to a report analysis service, external destination verification comes into play: on the receiving side there must be a TXT record named example.com._report._dmarc.report-service.example.net containing v=DMARC1. Without this record, reporters may not send the report. If the report mailbox stays empty, this is the first place to look. The format and sending of aggregate reports are defined in RFC 9990, and per-message failure reports in RFC 9991. If the volume is high, use a parser or analysis service instead of reading XML by hand; the team still needs to know the interpretation logic.
Each record block is a combination of a source IP and a result. When reading, look at these fields:
| XML field | What it tells you |
|---|---|
source_ip, count |
The sending IP and the number of messages in this combination |
policy_evaluated |
The aligned DKIM/SPF result from a DMARC standpoint and the action applied |
header_from |
The domain in the From header |
auth_results/dkim |
The signing d= domain, selector and raw result |
auth_results/spf |
The domain SPF was verified for and the raw result |
The key to diagnosis is the difference between policy_evaluated and auth_results. If DKIM shows pass in auth_results but the d= value is service-provider.example.net, the signature is valid but not aligned. Group source IPs by reverse DNS and ownership information. Don’t chase IPs one by one. The question to ask is “which service?”
An inventory of legitimate sending sources
Reports show who sends on behalf of your domain; they don’t tell you who owns that system. Build the inventory from two directions: backwards from the sources in the reports, forwards from conversations with departments. Accounting, HR, marketing, sales and production are asked a single question: “Do you have a system that automatically sends email to customers or suppliers with our domain?”
The sources most often missed in the inventory:
- Notification emails from the ERP or e-invoicing integrator
- CRM and marketing automation platforms
- Help desk and ticketing systems
- The “email my scanned document” feature of multifunction printers, and on-premises SMTP relays
- Website contact forms, e-commerce and HR applicant tracking systems
- Payroll and account statement sends that run at month end, quarter end or yearly
For each source, the owning department, sending addresses, domain used, SPF and DKIM alignment status and the action to take are written into a table. A source whose owner can’t be found isn’t a blocker that stops the migration. It’s an enterprise cybersecurity finding that needs separate investigation; it means a server nobody knows about is sending email with your domain.
SPF and DKIM alignment for third-party services
The order is the same for every service: first have DKIM signed with your own domain, then set up SPF alignment with a custom Return-Path if possible. DKIM comes first because it doesn’t break on forwarding and doesn’t add load to SPF’s 10 DNS lookup limit. Most services support signing on your behalf with CNAME or TXT records published on your domain:
s1._domainkey.example.com. CNAME s1.dkim.service-provider.example.net.
bounce.example.com. CNAME bounce.service-provider.example.net.
If the service only signs with its own domain and doesn’t support a custom Return-Path, two routes remain: send from that service on a separate subdomain (for example notifications.example.com), or change the service. Before adding a new service to SPF, check the lookup budget; details are in the SPF 10 DNS lookup limit and fixing permerror guide.
For on-premises applications and printers, the cleanest route is to pass devices through an authenticated relay instead of sending directly to the internet; if you use Microsoft 365, to Exchange Online with a connector. The DKIM signature is then added centrally. You don’t have to deal with creating DNS records printer by printer. If delivery problems continue for messages leaving through Microsoft 365, compare the message trace and the rejection code using the Exchange Online mail flow diagnosis steps.
A none, quarantine, reject migration plan
Stages progress by measurement, not by calendar. Before moving to the next stage, there must be no remaining volume in the reports that fails DMARC and is known to be legitimate. Phase by domain, not by percentage: first domains that don’t send email, then subdomains with few sending sources, and last the main domain.
| Stage | Example record | Criterion to proceed |
|---|---|---|
| 1. Monitoring | p=none |
All legitimate sources are in the inventory and aligned |
| 2. Quarantine in test mode | p=quarantine; t=y |
Failing traffic in reports is spoofing only; receivers that support t=y don’t apply the policy |
| 3. Quarantine | p=quarantine |
Infrequent sources have also been observed for at least one cycle |
| 4. Reject in test mode | p=reject; t=y |
No reports of legitimate messages landing in quarantine; receivers that support t=y apply a lower policy (quarantine) |
| 5. Reject | p=reject |
Moved to a continuous monitoring routine |
RFC 9989 removed the old pct tag; the t (test mode) tag replaced it. t=y asks the receiver not to apply the published policy but a lower one: none for p=quarantine; t=y, quarantine for p=reject; t=y. The default value is t=n. In RFC 7489, pct was a tag asking for the policy to be applied only to a certain share of failing messages; because receivers didn’t apply it consistently, it never became a reliable phasing method, and today it’s a historical detail. Don’t use pct in new records. t=y shouldn’t be treated as a safety net either: not every receiver may implement RFC 9989 yet, and some may ignore the tag. In every case, the receiving server decides how to apply the policy with its own local policy; publishing p=reject doesn’t mean the message will be rejected by every receiver.
The final record might look like this:
_dmarc.example.com. 3600 IN TXT "v=DMARC1; p=reject; sp=reject; rua=mailto:[email protected]; adkim=r; aspf=r"
Verify after publishing:
dig +short TXT _dmarc.example.com
Resolve-DnsName -Name _dmarc.example.com -Type TXT
If the command returns two lines, there’s a problem. When there’s more than one DMARC record with the same name, receivers ignore the policy; this usually happens when a new record is added in the DNS panel and the old one isn’t deleted.
Subdomains and unused domains
If there’s no sp tag, subdomains inherit the main policy. A subdomain with its own _dmarc record uses that record. RFC 9989 also defines the np tag for non-existent subdomains; if there’s no np, sp applies, and if that’s missing too, p. Before the migration, inventory the _dmarc.* records. If you need to move a subdomain reserved for marketing more slowly than the main domain, write a separate record for that subdomain; don’t leave sp=none on the main domain. sp=none weakens protection against spoofing with random subdomains.
Domains that don’t send email (old brand names, spelling variations, defensively registered domains) can be locked down straight away with the strictest setting. There’s no legitimate traffic to lose:
example.org. IN TXT "v=spf1 -all"
_dmarc.example.org. IN TXT "v=DMARC1; p=reject; sp=reject; rua=mailto:[email protected]"
example.org. IN MX 0 .
The last line is the “null MX” record from RFC 7505 and states that the domain doesn’t accept email. Reports from these domains cleanly show spoofing attempts made in your name; every send that appears there is forged.
Monitoring after the migration
The work continues after p=reject is published; the record should be tied to change management. If an email authentication check isn’t added to the purchasing process when a new SaaS service is acquired, the problem may surface as invoices that never reach customers. For a structure that runs smoothly:
- A weekly report summary: new source IPs, unaligned legitimate volume, spoofing attempts
- The question “will it send email with our domain?” in the purchasing and new application approval process
- Periodic rotation of DKIM keys and removal of unused selectors
- Recording DNS changes; DMARC, SPF and DKIM records having a known owner
- Quickly comparing users’ “the email didn’t reach the customer” reports with report data
DMARC strengthens protection against unauthorized use of your From domain; whether to apply the policy is the receiving server’s decision. It doesn’t cover fraud using look-alike domains, display name impersonation or a compromised real account. Process controls for that gap are in the BEC and fake invoice fraud guide, and user-side defense layers in the reducing phishing risk in corporate email article. The reference for the protocol’s official definition is RFC 9989; for reporting, RFC 9990.
In setups with a large domain inventory and scattered sending sources, report analysis and service alignment can be carried out together within our email security and phishing protection service.
Frequently asked questions
How many reports should be collected before moving to p=reject?
Coverage decides, not the calendar. Infrequent sources such as month-end invoicing, quarter-end statements and seasonal campaigns need to appear in the reports at least once. Depending on the organization’s business cycle, this can take weeks or months. The criterion: no unexplained legitimate source should remain in the reports.
SPF passes but DMARC fails; why?
SPF passed for the envelope sender (Return-Path) domain, but that domain isn’t aligned with the one in the From header. Most third-party services use their own bounce domain. The fix is one of two routes: configure the service to use a custom Return-Path with your own subdomain, or have it sign DKIM with your own domain.
Are forwarded emails lost after p=reject?
SPF most likely breaks on forwarding. If the message wasn’t modified, the DKIM signature stays valid and DMARC passes via DKIM. Mailing lists that add a tag to the subject line or a footer can break DKIM; at that point, the recipient’s ARC support is decisive. That’s why aligned DKIM should be required for every source.
Are subdomains protected when the main domain has reject?
If there’s no sp tag, subdomains inherit the main domain’s p policy. If a subdomain has its own _dmarc record, that record applies. For subdomains that need a different policy, write the sp tag explicitly and inventory the _dmarc records on subdomains.