Somebody, at some point, decided that the most important security information a domain owner will ever receive should be delivered as a gzipped XML file attached to an email with no body. One imagines a committee. One imagines the committee was pleased. The format has not changed since, and so anyone who wants to know who is sending mail as their domain must first learn to read a document designed with a parser in mind and a person as an afterthought.
It is not difficult once the layout is clear. An aggregate report has three parts: who is reporting, what policy they found, and what they saw. This guide walks through a complete report, element by element, and marks the handful of fields that deserve your attention and the larger number that do not. For the broader picture of what reports are for, see understanding DMARC reports.
A complete report
Here is a small but realistic aggregate report, with one record:
<?xml version="1.0" encoding="UTF-8"?>
<feedback>
<report_metadata>
<org_name>receiver.example</org_name>
<email>noreply-dmarc@receiver.example</email>
<report_id>8842190357716203991</report_id>
<date_range>
<begin>1794787200</begin>
<end>1794873599</end>
</date_range>
</report_metadata>
<policy_published>
<domain>example.com</domain>
<adkim>r</adkim>
<aspf>r</aspf>
<p>quarantine</p>
<sp>quarantine</sp>
<pct>100</pct>
</policy_published>
<record>
<row>
<source_ip>203.0.113.25</source_ip>
<count>142</count>
<policy_evaluated>
<disposition>none</disposition>
<dkim>pass</dkim>
<spf>fail</spf>
</policy_evaluated>
</row>
<identifiers>
<header_from>example.com</header_from>
</identifiers>
<auth_results>
<dkim>
<domain>example.com</domain>
<selector>mail2026</selector>
<result>pass</result>
</dkim>
<spf>
<domain>bounce.mailer-platform.net</domain>
<result>pass</result>
</spf>
</auth_results>
</record>
</feedback>Real reports contain many record blocks, one per distinct combination of source and result. Everything else appears once. If you have a file of your own to hand, the free DMARC report analyzer will lay it out as a table while you follow along.
report_metadata: who is talking, and about when
| Element | Meaning | Matters? |
|---|---|---|
org_name | The organization that received the mail and wrote the report | Yes. It tells you whose users were being mailed |
email | A contact address for the reporter | Rarely |
report_id | A unique identifier for this report | For de-duplication only |
date_range | Start and end of the period, as Unix timestamps in UTC | Yes. All counts refer to this window |
The timestamps are seconds since January 1, 1970. The pair above covers the whole of November 16, 2026, UTC. A report is a statement about a period that has already ended, which is the first reason DMARC data is always at least a day old. Optional extra_contact_info and error elements sometimes appear here; the latter is worth a glance, since a reporter may use it to say it had trouble fetching your record.
policy_published: what the receiver found in your DNS
This block echoes the DMARC record the reporter retrieved for your domain during the period. It is not what you intended to publish; it is what they saw.
domain: the domain whose DMARC record was applied. For mail from a subdomain without its own record, this is the organizational domain.pandsp: the policy for the domain and for its subdomains.adkimandaspf: alignment mode for DKIM and SPF,rfor relaxed orsfor strict.pct: the percentage of failing mail the policy applies to.fo: failure reporting options, if present.
Check this block after any change to your record. If you moved to p=reject on Monday and Thursday’s reports still say quarantine, either DNS caching is longer than you thought or the change did not take. Compare it with what the DMARC checker shows now.
row: the count and the verdict
Each record describes a group of messages that shared the same source IP and the same results. The row holds the essentials.
source_ip and count
source_ip is the address that delivered the mail to the reporter, IPv4 or IPv6. count is how many messages are in the group. The address is all the identification you get; attaching a name to it takes a reverse DNS lookup and a look at the authentication domains further down. An IP that keeps failing is also worth checking against blocklists, which blacklist monitoring does daily for the sources you own.
policy_evaluated
This is the most important block in the file, and the most frequently misread.
disposition: what the receiver actually did with these messages under DMARC:none(no DMARC action),quarantine, orreject.dkim:passorfail, meaning DKIM passed and aligned with the From domain, or did not.spf:passorfail, meaning SPF passed and aligned, or did not.
These two are DMARC’s own verdicts, alignment included. If either says pass, the messages passed DMARC. In the sample, dkim is pass and spf is fail: DMARC passed on the signature, and the disposition is accordingly none.
The report gives every result twice: once as it happened, and once as DMARC judged it. The difference between the two is alignment.
reason: when the receiver overrides you
Sometimes policy_evaluated contains a reason element with a type and an optional comment. It appears when the disposition differs from what your policy called for.
| Type | Meaning |
|---|---|
forwarded | The receiver judged the message to have been forwarded and let it through |
mailing_list | It came through a mailing list, which explains the broken authentication |
trusted_forwarder | It arrived from a forwarder the receiver trusts |
sampled_out | Your pct value excluded this message from the policy |
local_policy | The receiver’s own rules took precedence; the comment may say which |
other | Something else; read the comment |
identifiers: the domains on the message
header_from: the domain in the From header. This is the domain DMARC was evaluated for. Watch for subdomains here; mail fromnews.example.comappears in the report forexample.comwhen the subdomain has no record of its own.envelope_from: the domain of the envelope sender (Return-Path), when the reporter includes it.envelope_to: the recipient domain, included by some reporters.
auth_results: what SPF and DKIM actually said
Here are the raw results, before any question of alignment.
Each dkim entry gives the signing domain (the d= value), the selector, and a result such as pass, fail, none, temperror, or permerror. There can be several entries when a message carried several signatures. The spf entry gives the domain that was checked, normally the envelope sender’s, and a result that may also be softfail or neutral.
Now the sample makes sense. SPF passed in auth_results, but for bounce.mailer-platform.net, which does not align with example.com; hence spf is fail in policy_evaluated. DKIM passed for example.com, which does align. This is the signature of a third-party service correctly set up to sign with your domain while using its own bounce address, and it is entirely healthy. The mechanics are in DMARC alignment explained.
The selector is a small gift: it often identifies the sending service outright, and it tells you which DNS record that mail depends on. If a selector you see here fails to resolve in the DKIM checker, you have found a problem before your recipients did.
Reading a record in thirty seconds
For each record, in order:
- Look at policy_evaluated. If
dkimorspfispass, DMARC passed. Move on unless you want both. - If both are fail, look at auth_results. Raw passes for another domain mean a legitimate service that needs configuring. Raw failures or no signature at all mean forwarding, modification, or a stranger.
- Identify the source_ip. Is it a service you use?
- Check the count. Three messages are an anecdote. Three thousand are a finding.
- Note the disposition and any reason. That is what happened to the mail.
Then repeat for every record in every report from every receiver, every day, and add them up by sender. That last instruction is where enthusiasm for hand-reading tends to expire. DMARC monitoring performs the addition and names the senders; the sending sources documentation explains how the naming is done, and the reporting setup guide shows how to point your rua tag at it. The XML will go on arriving either way. It would be a pity for nobody to read it.

