A field-by-field walk-through of a DMARC aggregate report
A DMARC aggregate report is one receiver's account of one reporting period, usually a single UTC day. It has a header that says who sent the report, which period it covers, and what DMARC policy the receiver saw, followed by one record per sending IP. Each record answers four questions: how many messages that IP sent using your From domain, what SPF and DKIM said, whether either produced an aligned pass, and what the receiver did. The part people misread most is the difference between the raw SPF and DKIM results (auth_results) and the DMARC-aligned results (policy_evaluated). A source can pass SPF and DKIM and still fail DMARC.
This is the deep version, element by element. For the shorter guide, see DMARC reports explained. If you're not sure which report type you're looking at, start with Aggregate (RUA) vs forensic (RUF) reports.
Before you open it: the email and the file
The report arrives as an email whose subject follows this pattern:
Subject: Report Domain: example.com
Submitter: mail.receiver.example
Report-ID: <sample-ridtxt@example.com>
Report Domain is your policy domain. Submitter is the receiver that sent the report.
The attachment is the DMARC XML, usually compressed, and it should be gzip (.xml.gz). The filename packs in the receiver, your domain, and the start and end of the period as Unix timestamps: receiver!policy-domain!begin!end.
One caution before you double-click. Report attachments can be compressed or XML "bombs" built to exhaust whatever opens them, and report data can be forged. Treat these files as untrusted input, especially from a "reporter" you've never heard of. Real report mail is itself required to pass DMARC with alignment, which helps.
The report format at a glance
Everything sits inside a <feedback> element with three parts:
report_metadata: who reported and for when.policy_published: your DMARC record as this receiver saw it.record: one or more, each describing mail from one source.
report_metadata
org_name: the reporting organization, such asgoogle.com.email,report_id: contact and a unique ID.date_range:beginandendas Unix timestamps (seconds since 1970, UTC). Typically one UTC day. Convert them before you compare against your own logs.error(optional): problems the receiver hit with your record, such as invalid tag values or multiple DMARC records. Don't skip it. It can be the first sign your record is broken.
policy_published
This is your DMARC record as that receiver read it: domain (the policy domain), p, and, depending on the format, sp, np, adkim, aspf, fo, testing (your t tag), and discovery_method. Tags you didn't set show their defaults. The tags themselves are explained in DMARC tags explained: p, sp, np, adkim, aspf, rua, ruf, fo, and t.
Use this block to confirm a DNS change has been picked up. If you changed your policy partway through a day, that day's reports can mix old and new policy.
You may still see <pct> here. It comes from the older report format, which Google's and Microsoft's own samples still follow. pct is historic, so don't read its presence as advice to use it. If your record still has pct, remove it (pct is historic: how to phase in DMARC enforcement now).
record: row
source_ip: the IP that connected to the receiver. That's the last hop, so forwarded mail shows the forwarder's IP, not yours.count: how many messages matched this row.policy_evaluated: the DMARC verdict for those messages (next section).
policy_evaluated: the DMARC verdict
dkimandspfare eachpassorfail, and they're alignment-aware.dkimis the result of the DKIM identifier alignment test, andspfis the same for SPF. Sospf>failhere doesn't mean SPF failed. It means SPF didn't produce an aligned pass.dispositionis the action the receiver applied under DMARC:none,quarantine, orreject. The current format addspass, meaning no action because the message passed DMARC under an enforcing policy. Older-format reports don't usepass. Disposition doesn't tell you whether the mail landed in the inbox or spam.reason(optional) appears when the receiver did something other than what your policy asked. The current types arelocal_policy,mailing_list,trusted_forwarder,policy_test_mode(yourt=y), andother. Older reports also useforwardedandsampled_out, the latter tied to the historicpct.
Policies are requests. A reason is the receiver telling you it decided differently, and why.
identifiers
header_from: the domain in the From header. This is what DMARC protects.envelope_from: the SPF (MAIL FROM) domain. It can be empty for bounces.envelope_to: optional, often absent.
auth_results: the raw results
These results are uninterpreted with respect to DMARC. They're what SPF and DKIM said before anyone checked alignment:
- Zero or more
dkimentries:domain(thed=value),selector(thes=value, required in the current format), andresult. There are none if no DKIM signature was checked. - At most one
spfentry:domain, optionalscope, andresult.
DKIM results can be none, pass, fail, policy, neutral, temperror, or permerror. SPF adds softfail.
An annotated example
This is a trimmed sample of the current report format:
<feedback xmlns="urn:ietf:params:xml:ns:dmarc-2.0">
<report_metadata>
<org_name>Sample Reporter</org_name>
<date_range>
<begin>302832000</begin>
<end>302918399</end>
</date_range>
</report_metadata>
<policy_published>
<domain>example.com</domain>
<p>quarantine</p>
<sp>none</sp>
<np>none</np>
<testing>n</testing>
<discovery_method>treewalk</discovery_method>
</policy_published>
<record>
<row>
<source_ip>192.0.2.123</source_ip>
<count>123</count>
<policy_evaluated>
<disposition>pass</disposition>
<dkim>pass</dkim>
<spf>fail</spf>
</policy_evaluated>
</row>
<identifiers>
<envelope_from>example.com</envelope_from>
<header_from>example.com</header_from>
</identifiers>
<auth_results>
<dkim>
<domain>example.com</domain>
<result>pass</result>
<selector>abc123</selector>
</dkim>
<spf>
<domain>example.com</domain>
<result>fail</result>
</spf>
</auth_results>
</record>
</feedback>
Reading it: 192.0.2.123 sent 123 messages with From: example.com. SPF failed, but DKIM passed with d=example.com, which aligns with the From domain. DMARC needs only one aligned pass, so these messages passed, and the receiver took no action (disposition pass) even though the policy is quarantine. The SPF failure alone didn't matter.
Now the classic problem case. In this sample (older format), the row shows policy_evaluated spf = fail, while auth_results shows SPF pass for bounces.vendor.example and header_from is example.com. SPF passed, just for the vendor's bounce domain, so it didn't align. Microsoft calls this gap between "authentication pass and alignment pass" the most important insight in aggregate reports. In that sample, DKIM is signed with example.com, so DMARC still passes. Take that DKIM signature away, and the same mail fails DMARC with every raw check green.
Reading one record, step by step
- Find
header_from. Is it your domain or a subdomain? A report for your policy domain can include records for subdomains that have no record of their own. - Check
policy_evaluated. Ifdkimorspfispass, those messages passed DMARC. Bothfailmeans DMARC failed. - If it failed, open
auth_results. - SPFpassbut alignedfail: SPF passed for another domain, typically a vendor's bounce domain. - DKIMpasswith adomainthat isn't yours: the vendor signs with its own domain. The fix is custom DKIM at the vendor, withd=set to your domain (Third-party mail fails alignment). Alignment rules are in Identifier alignment for SPF and DKIM. - Everythingfailornonefrom an IP you don't recognize: it could be spoofing, a legitimate service nobody has set up, or a forwarder. Don't decide yet (Legitimate mail vs spoofing in reports, Reports show sources we do not recognize). - Note
dispositionand anyreason. An override isn't your policy "not working". It's the receiver telling you what it chose. - Weigh by
count, across days and across reporters. One report is one receiver's view of one day. Different receivers can see different paths and results for the same source.
What a report can't tell you
- Inbox or spam placement. It isn't in the format, and adding it is discouraged. See Authentication vs inbox placement.
- Mail rejected early. Messages rejected early in the SMTP transaction, such as on an SPF
-allhard fail, will never appear in aggregate DMARC reports. - Receivers that don't report. Sending reports is a SHOULD with known opt-outs (Why DMARC reports do not arrive).
- The original sender behind a forwarder. You see the last hop (Forwarded mail fails DMARC).
- Service names. You get IPs and domains. Mapping them to real services is the inventory work in How to inventory sending sources.
When to stop reading by hand
Reading a few reports by hand is the best way to learn the format. It doesn't scale. Reports are meant to be machine-parsed, and the XML is not human readable in bulk. You can use a third-party reporting service or build your own parser. What a good visual view should show is in What visual DMARC analysis should show you.
Mistakes worth avoiding
- Reading only
auth_resultsand missing that nothing aligned. - Reading
policy_evaluatedspf=failas "SPF is broken." It's an alignment result. - Judging from one day or one reporter.
- Forgetting the timestamps are UTC epoch seconds.
- Ignoring
error, which can reveal duplicate or invalid records. - Ignoring
reason, so a receiver override looks like a broken policy. - Adding an unknown IP to SPF to make a failing row go away. That can authorize a spoofer (Legitimate mail vs spoofing in reports).
Free DMARC account
Turn raw DMARC reports into a readable dashboard
Point your rua address to our platform and read the results by sender and source instead of opening XML files.
- Results grouped by sending source
- SPF, DKIM and alignment results per sender
- The original XML reports are still there to download
FAQ
What do dkim and spf under policy_evaluated mean?
Whether each produced an aligned pass for DMARC. They aren't the raw SPF and DKIM results; those are in auth_results.
Why does SPF show pass in one place and fail in another?
SPF passed for a different domain, often a vendor's bounce domain, so it didn't align with your From domain.
Why is disposition none when my policy is reject?
In older-format reports, none also means no action was taken because the message passed DMARC. The current format adds pass for that case. Otherwise, the receiver applied an override (check reason), or your record has t=y and the receiver followed testing mode. Receivers make the final call.
Can I see whether mail went to spam?
No. Aggregate reports don't include placement.
How do I read a DMARC report from Google or Microsoft?
The same way as any other. The Submitter in the subject and org_name in the XML name the receiver, and a Microsoft aggregate report uses the same elements as Google's. Both companies' published samples still follow the older report format, so expect to see pct, and don't expect np, testing, or a pass disposition until a reporter moves to the current format. Microsoft 365 only sends aggregate reports when the receiving domain's MX points directly to Microsoft 365.
Is there an easier way than reading XML?
Yes: a reporting service or your own parser. See What visual DMARC analysis should show you.