DMARC aggregate reports are one of the best ways to discover who is sending mail that claims to use your domain. They can reveal an undocumented platform, a forgotten relay, a spoofing campaign, or an authentication change soon after it happens.
That makes them valuable operational evidence. It does not make them a complete log of every message sent, received, delivered, or displayed for your domain.
The difference matters when an administrator is investigating an incident. A report can support a conclusion about the mail stream observed by one receiver during a reporting period. It usually cannot prove what happened to a particular message or explain everything that happened before or after that receiver saw it.
The DMARC aggregate reporting specification describes reports as receiver-generated, periodic feedback about messages relevant to a domain owner. They are normally sent daily or more frequently, and the XML groups messages into records with common characteristics.
A record can contain details such as:
none, quarantine, or rejectFrom domain used for DMARC evaluationFrom domainThat is a compact view of an observed mail stream. It is intentionally much smaller than a copy of every message. Aggregation is what allows large receivers to provide useful feedback without sending a separate report for every delivery attempt.
A DMARC report answers “what authentication-related traffic did this receiver observe?” more reliably than it answers “what happened to this individual email?”
The count value is easy to overinterpret. It is the number of messages represented by that record for which the reported policy evaluation applied. It is not a list of message IDs, and it is not a universal counter shared by all receivers.
Suppose a report contains a record with:
<source_ip>192.0.2.44</source_ip>
<count>250</count>
<policy_evaluated>
<disposition>none</disposition>
<dkim>pass</dkim>
<spf>pass</spf>
</policy_evaluated>This supports the statement that the reporting system associated 250 messages with that source and result during the relevant period. It does not identify the 250 messages, tell you their subjects or recipients, or prove that every one reached an inbox.
RFC 9990 also clarifies that report counts can be based on messages received even when other filtering agents ultimately block them. A count is therefore not an inbox-delivery count. It is evidence of messages processed by the reporting receiver's mail system under the report's grouping rules.
Aggregate reporting is particularly good at answering questions about domain-level authentication operations.
The source IP and authentication identifiers can expose legitimate senders that were missing from the sender inventory. A new source may be a newly approved service, an old integration that was never documented, a forwarder, or an attacker.
The report is a starting point for classification, not an automatic verdict. An IP address may belong to shared infrastructure, and a DKIM signing domain may identify a service rather than the business team that configured it. Confirm the source with sending-system owners and provider records.
Trends in aligned SPF and DKIM results can show whether a migration, selector rotation, or new sending stream is authenticating as intended. A sudden increase in failed or unaligned records is a useful signal to investigate before enforcement is tightened.
Reports can also show that a source is passing authentication with a different identity than expected. That often points to a missing alignment setting, a bounce-domain change, or a provider configuration that was only partially completed.
Unrecognized sources with failing authentication can indicate messages claiming to use the domain without authorization. Grouping by source, date range, and authentication outcome helps separate a single noisy event from a recurring pattern.
This is an estimate of the traffic observed by participating receivers. It is not a census of every spoofed message on the Internet. Receivers that do not send reports, messages that never reach a participating receiver, and reporting failures are outside the dataset.
The report includes the policy configuration observed by the reporting system and the disposition applied to matching records. That helps establish whether a policy change reached a receiver and whether local policy affected the final disposition.
The DMARC specification allows a reporting entity to send multiple reports when it observes policy changes during a reporting period. It may also report only the final configuration while the period contains traffic evaluated under an earlier configuration. Treat a report period that overlaps a DNS or DMARC policy change with extra care.
The missing fields are as important as the fields that are present.
A report shows that a receiver saw traffic claiming to use your domain. It does not prove that your application or outbound MTA originated it. For an accepted sender, compare the report with the sender's message logs, queue records, and provider event data.
For a spoofed message, the source may have no relationship to your infrastructure at all. The fact that it used your From domain is precisely why it appeared in the report.
There is generally no message ID, recipient address, final mailbox, spam-folder result, or per-message delivery status in an aggregate record. A receiver can count a message while a later filter rejects it, or accept it into a system without placing it in the inbox.
Use the receiving provider's message trace, delivery event, bounce, or complaint data when the question is about a specific message. Use the original message headers when the question is about the path or authentication result of that message.
Aggregate DMARC reports do not contain the subject, body, links, attachments, or local account that submitted a message. They cannot tell whether an authenticated message was a normal campaign, a compromised account, or a malicious request sent by an authorized system.
DMARC authenticates domain-level identifiers. It does not authenticate the intent of a sender or evaluate content. A high volume of aligned mail can still be unwanted or abusive.
The source IP in a record is the connecting system observed by the reporting receiver for that traffic. It is not necessarily the original application, mailbox, or first outbound MTA. Forwarders, gateways, relays, and shared provider infrastructure can sit between those systems.
The report also does not contain the complete Received header chain. To reconstruct a message path, preserve and inspect the original message source or the relevant MTA traces.
Reports depend on receiver participation, the presence of a valid reporting request, successful report generation, and successful delivery to the report destination. A quiet report day may mean low traffic, but it may also mean that a receiver did not report, a report was delayed, or ingestion failed.
Do not interpret the absence of a report as proof that no mail was received. Monitor report freshness and ingestion health separately from authentication results.
Report processing needs to expect imperfect telemetry. A receiver can retry delivery and send the same report again. Two reports can cover overlapping periods. Separate reporting systems can describe related traffic independently. The current aggregate reporting specification provides report identifiers to help consumers recognize duplicates, but it does not turn every report into a globally deduplicated event stream.
Store the report ID, reporting organization, policy domain, covered time range, attachment fingerprint, and ingestion time. Use those values to detect exact duplicates and flag partial overlaps for review. Do not blindly add every count value together when periods overlap.
There is a similar caution around policy changes. DNS caching and receiver processing do not change everywhere at the same instant. A report can contain traffic evaluated under a previous policy even though the current policy has already changed, or it can summarize the receiver's final observed configuration for a period that was not uniform.
The most reliable investigations combine several evidence sources instead of forcing aggregate reports to answer questions they were not designed for.
| Question | Best evidence to start with |
|---|---|
| Who is sending as the domain? | DMARC aggregate reports, DNS, provider inventory |
| Is SPF or DKIM alignment failing? | DMARC reports, original headers, authentication logs |
| Did our system generate this message? | MTA or application logs, provider event logs |
| Was this message accepted or delivered? | Receiver trace, SMTP response, bounce or delivery event |
| Did it reach the inbox or spam folder? | Recipient-provider telemetry or controlled mailbox tests |
| Was the message harmful or unwanted? | Content and URL scanning, complaints, abuse reports, user reports |
| Was an account or sending service compromised? | Authentication logs, access logs, volume anomalies, incident records |
The report is often the best first clue because it gives broad visibility across senders and receivers. It is rarely the final piece of evidence for a message-level incident.
Treat DMARC data as a time-series signal with a known scope.
Keep normalized report history long enough to see changes in the sending estate, but retain raw email and attachments according to a deliberate operational and privacy policy. Aggregate does not mean risk-free: source IPs and traffic volumes can still reveal information about mail operations.
DMARC aggregate reports are a map of authentication-related traffic observed by reporting receivers. Maps are powerful because they show patterns that an individual message log cannot: an undocumented sender across many destinations, a new spoofing source, or a slow drift in alignment.
They are not a complete mail log. They do not enumerate every message, preserve every header, record every filtering decision, or prove inbox delivery. Treating them as a complete history creates false certainty and can lead to the wrong remediation.
Use DMARC reports to inventory senders, measure authentication coverage, spot abuse, and understand policy effects. Then join those signals with system logs and message-level evidence when the question becomes specific. The strongest conclusion is not “the report says this definitely happened.” It is “the report supports this interpretation, and these other records confirm the rest.”