DMARC aggregate report

Also known as: RUA, aggregate report

A DMARC aggregate report (RUA) is an XML file a mailbox provider sends — typically once a day — summarising all the mail it saw claiming to be from your domain. The format is specified by RFC 9990. It does not contain message content; it is a statistical roll-up: sending IP, message count, SPF/DKIM results, and the disposition applied.

Requesting reports

You request them with the rua= tag in your DMARC record:

v=DMARC1; p=none; rua=mailto:dmarc@yourdomain.com

Three practical details that the tag alone doesn’t tell you:

  • They arrive gzipped. The XML should be GZIP-compressed and attached as application/gzip; a receiver that sends it uncompressed uses text/xml, which is allowed but discouraged, because a large report can exceed the receiving server’s size limit and simply never be delivered. The filename carries the reporting receiver, your domain, the start and end timestamps, and sometimes a unique id — mail.receiver.example!yourdomain.com!1013662812!1013749130.xml.gz.
  • Daily is what you get, not what you asked for. A reporting period normally covers a single UTC day starting at 0000 UTC, which is why the timestamps rarely line up with your own midnight. There used to be a tag for requesting an interval, ri=; providers sent daily reports regardless, and the current specification has removed it.
  • A mailbox outside your organizational domain has to opt in. An agency or vendor address only works if that domain publishes a _report._dmarc record authorising it; without one, conforming receivers discard the request silently — no bounce, no error, no reports. An address under your own domain needs nothing. This is the single most common reason a perfect-looking record produces nothing; see reports won’t arrive if rua= is wrong, or the agency treatment if you collect for domains you don’t own.

What’s inside a record

Each <record> groups messages by sending source and From: domain, and reports what the receiver did with them alongside the alignment outcome:

<record>
  <row>
    <source_ip>203.0.113.10</source_ip>
    <count>42</count>
    <policy_evaluated>
      <disposition>none</disposition>
      <dkim>pass</dkim>
      <spf>fail</spf>
    </policy_evaluated>
  </row>
  <identifiers>
    <header_from>yourdomain.com</header_from>
  </identifiers>
  <auth_results>
    <dkim>
      <domain>yourdomain.com</domain>
      <selector>s1</selector>
      <result>pass</result>
    </dkim>
    <spf>
      <domain>yourdomain.com</domain>
      <result>fail</result>
    </spf>
  </auth_results>
</record>

All three children are required, and the two outside <row> are the ones tooling most often forgets: <identifiers> carries the From: domain the policy was looked up on, and <auth_results> carries what SPF and DKIM actually returned, before DMARC asked whether either one aligned.

Reading these by hand across many domains does not scale — which is what a self-hosted analyzer like DMARC Analyzer is for. See how to read a DMARC aggregate report.

Providers don’t all implement the schema identically

The schema leaves a lot optional. <policy_published> has to carry only the domain and p; sp, np, adkim and aspf turn up only when the receiver echoes them. A <record> can arrive with no <envelope_from>, and with <auth_results> carrying no <dkim> child at all — the element is required but its DKIM results are not. And the superseded schema made the DKIM <selector> optional, so an older reporter can tell you a signature failed without telling you which key — the current one requires it.

File names are the other trap: the unique id is optional and not every receiver builds the name exactly as specified, so tooling that reads the filename instead of the XML inside it breaks on the next provider. Parse the document, and treat every optional element as absent until the report proves otherwise — a parser written for one provider’s exact output quietly drops rows from another’s.