skip to content

DMARC

DMARC requires an SPF or DKIM pass aligned with the visible From domain, publishes a policy for mail that fails, and asks receivers to report. Interviewers ask why passing SPF alone is not enough.

on this pageshow

explore

questions

28

What is the difference between DMARC aggregate reports and DMARC failure reports, and which one does a domain owner rely on?

level: juniorimportance: must knowfreq 48%

answer

  1. statistics versus samples
  2. one periodic, one near-immediate
  3. XML summary against an ARF message
  4. privacy decides who sends which

basics

~20 s

Aggregate reports (RFC 9990, requested with rua) are daily XML summaries of SPF, DKIM and alignment results per sending IP, with no message content; failure reports (RFC 9991, requested with ruf) are per-message ARF samples, and most receivers send none.

solid answer

~50 s

An **aggregate report** is an XML document, normally gzip-compressed, that a receiver mails to each `rua` address for a reporting period; RFC 9989 says receivers SHOULD send them at least once every 24 hours. It counts messages per source IP and gives the disposition, the DKIM and SPF alignment verdicts and the raw results, for passing and failing mail alike. A **failure report** goes to a `ruf` address soon after a DMARC failure, in the Abuse Reporting Format (RFC 5965 with RFC 6591, updated by RFC 9991), and carries the failed message's header block or the whole message. Because that can include personal data, RFC 9989 notes that most receivers refuse to send them, and those that do often redact. So a domain owner builds monitoring on aggregate reports and treats failure reports as an occasional diagnostic extra. Both formats left RFC 7489 for their own documents in May 2026.

go deeper

for a junior

Recall the two report types, the tag that requests each, and the one-line difference: aggregate reports count, failure reports sample. Knowing which one can carry message content is the core of the answer.

for a middle

Explain format and timing: gzip-compressed XML per reporting period against an ARF message per failure, and why aggregate reports cover passing mail as well as failures.

for a senior

Show that you know failure reports are scarce in practice and that monitoring is built on aggregate data; mention that RFC 9989 removed the ri and rf tags in May 2026.

for a principal

Frame the trade-off: aggregate data is enough for a sender inventory at little privacy cost, while failure reports buy per-message detail at a legal and handling cost the organisation has to own.

## Two kinds of DMARC feedback DMARC lets the owner of a domain publish a policy for mail whose visible `From` domain (the RFC 5322 From, called the **Author Domain**) does not pass. The same DMARC record can ask receivers for feedback, and the specification defines two kinds. Since May 2026 they live in separate documents: **RFC 9990** for aggregate reports and **RFC 9991** for failure reports, which together with RFC 9989 obsolete the single 2015 text, RFC 7489. A record asks for aggregate reports with the `rua` tag and for failure reports with the `ruf` tag; if a tag is absent, receivers MUST NOT generate that kind of report. ## Aggregate reports: the statistics An aggregate report answers one question for one reporting period: *which systems sent mail using my domain, and how did that mail authenticate?* - **Format:** an XML document defined by the XML Schema in RFC 9990, with the root element `feedback`. - **Packaging:** attached to an email message; the XML SHOULD be gzip-compressed (`application/gzip`, file extension `.xml.gz`), otherwise it is sent as `text/xml`. - **Cadence:** RFC 9989 says receivers SHOULD send aggregate reports at least once every 24 hours; the period typically covers one UTC day. - **Content:** report metadata, the policy the receiver found, and one or more `record` elements, each with a `source_ip`, a message `count`, the applied disposition, the DKIM and SPF alignment verdicts and the raw DKIM and SPF results. - **Scope:** messages that passed as well as messages that failed, so the owner sees every stream, not only the problems. RFC 9990 states that this data contains no individual mail addresses, no IP addresses of individuals and no message content. ## Failure reports: the samples A failure report answers a different question: *what did this particular failing message look like?* It is generated soon after a receiver sees a DMARC failure rather than on a schedule, and it uses the **Abuse Reporting Format** (ARF, RFC 5965) with the authentication-failure extensions of RFC 6591, as updated by RFC 9991. An ARF report is a `multipart/report` message with three parts: 1. a human-readable description; 2. a machine-readable `message/feedback-report` part with fields such as `Feedback-Type: auth-failure`, `Auth-Failure: dmarc`, `Identity-Alignment`, `Source-IP` and `Reported-Domain`; 3. the original message in full (`message/rfc822`) or a copy of its entire header block (`text/rfc822-headers`). That third part is both the value and the problem: it can hold recipients, subjects, bodies and attachments. RFC 9989 records that most receivers refuse to send failure reports because of privacy concerns, and RFC 9991 notes that many large-scale providers limit or disable them; those that do send often redact. The record's `fo` tag tells receivers which failures the owner would like reported, but report generators "may choose to adhere" to it - it is a request, not a switch. ## Side by side | | Aggregate report | Failure report | |---|---|---| | Defined in | RFC 9990 | RFC 9991, format from RFC 5965 and RFC 6591 | | Requested with | `rua` | `ruf`, shaped by `fo` | | Timing | periodic, at least daily (SHOULD) | soon after a failure, under a rate limit | | Unit | counts per source IP and result | one message, or a group of similar ones | | Content | domains, results, counts | header block or the whole message | | Availability | sent by receivers that implement reporting | sent by few receivers, often redacted | ## Why aggregate reports carry the programme A domain owner preparing for enforcement needs to know every system that sends with its name. The aggregate report is built for that: each source IP appears with its volume and with whether it passed aligned, so a forgotten application server or an outside sending service shows up as a row. Among the things RFC 9989 says a domain owner MUST do to fully participate in DMARC is set up a mailbox to receive aggregate reports and analyse them. Failure reports can explain one failure quickly, but because few receivers send them, a plan that depends on them sees only a small, unrepresentative slice of the traffic. ## What changed in May 2026 - The reporting text moved out of RFC 7489 (Informational) into RFC 9990 and RFC 9991, both on the Standards Track. - The `ri` tag (a requested aggregate interval, default 86400 seconds in RFC 7489) and the `rf` tag (a requested failure-report format) were removed by RFC 9989; a receiver that follows it ignores them as unknown tags. - Aggregate reports gained elements such as `np`, `testing` and `discovery_method`, and a reported DKIM result must now name its selector.

  • Does a receiver send DMARC aggregate reports for a domain whose DMARC record has no rua tag?
    No. RFC 9989 says that if the `rua` tag is not provided, Mail Receivers MUST NOT generate aggregate feedback reports for the domain, and the same rule applies to `ruf` and failure reports. A record without `rua` still carries a policy, but the owner gets no feedback at all, which defeats the purpose of starting in monitoring mode.
  • Can a domain owner still ask for a different reporting interval or failure-report format in its DMARC record?
    Not under RFC 9989 (May 2026). The `ri` tag, a requested interval that defaulted to 86400 seconds in RFC 7489, and the `rf` tag, a requested failure-report format, were removed. A receiver following RFC 9989 ignores them as unknown tags; aggregate reports SHOULD come at least every 24 hours, and failure reports use ARF as updated by RFC 9991.

An aggregate report is a bank statement: it lists each payee with totals, enough to spot a charge you do not recognise, but no receipts. A failure report is a photocopy of one disputed receipt - far more detail, including whatever personal information is written on it, so it is handed out sparingly.

  • Aggregate reports include copies of the messages that failed DMARC.
  • Failure reports arrive once a day as a summary of every failure.
  • Any receiver that sends aggregate reports will also send failure reports when asked.
  • Publishing ruf guarantees a sample of every forged message.
  • The ri tag still sets how often receivers send aggregate reports.
open as a page

Under RFC 9989, what must a DMARC policy record at _dmarc.example.com contain to be valid, and what does the smallest useful one say?

level: juniorimportance: must knowfreq 46%

basics

~20 s

A DMARC record must begin with v=DMARC1, spelled exactly, or receivers ignore it, and should carry p=none, quarantine or reject. The smallest useful one is v=DMARC1; p=none; rua=mailto: plus a report mailbox: monitoring with aggregate reports.

open as a page

Under DMARC, what do the policies none, quarantine and reject ask a receiving mail server to do with a message that fails DMARC?

level: juniorimportance: must knowfreq 62%

basics

~10 s

They state the From domain owner's preference for failing mail: none asks for no change, quarantine for treating it as suspicious, reject for refusal during the SMTP session. The receiver's local policy still decides.

open as a page

Why does a domain whose DMARC record publishes p=none get no protection against messages forging its From address?

level: juniorimportance: must knowfreq 58%

basics

~20 s

p=none expresses no handling preference, and RFC 9989 says a discovered p=none MUST NOT change how receivers handle mail, so a forged message is treated as if the domain had no DMARC policy. It buys visibility through reports, not protection.

open as a page

What did Gmail's and Yahoo's 2024 bulk-sender rules require of a sending domain's DMARC, and does p=none satisfy them?

level: juniorimportance: must knowfreq 50%

basics

~20 s

From February 2024 both providers require bulk senders to authenticate with SPF and DKIM, publish a DMARC record (p=none accepted) and send mail whose From domain aligns with SPF or DKIM. p=none meets the record rule but stops no forgery.

open as a page

Under RFC 9989 DMARC, what does a message need to earn a DMARC pass, and why is an SPF or DKIM pass alone not enough?

level: juniorimportance: must knowfreq 68%

basics

~20 s

A DMARC pass needs at least one SPF or DKIM pass for a domain aligned with the From header's domain. A pass for an unrelated domain proves nothing, because anyone can pass SPF or DKIM for a domain they control.

open as a page

What does an RFC 9990 DMARC aggregate report contain, and how does it reach the domain owner's rua address?

level: middleimportance: must knowfreq 38%

basics

~20 s

An RFC 9990 aggregate report is XML, usually gzip-compressed and mailed daily to the rua address: report metadata, the policy the receiver saw, and per source IP a message count, the disposition, aligned DKIM and SPF verdicts and the raw results.

open as a page

Under RFC 9989, what does each tag in the DMARC record v=DMARC1; p=reject; sp=quarantine; np=reject; adkim=s; rua=mailto:[email protected] request, and which defaults fill the gaps?

level: middleimportance: must knowfreq 38%

basics

~10 s

p=reject covers example.com, sp=quarantine its existing subdomains and np=reject subdomains that do not exist; adkim=s asks for strict DKIM alignment; rua names the aggregate-report mailbox. Absent tags default: aspf=r, fo=0, t=n, psd=u.

open as a page

Under RFC 9989 DMARC, how do you stage a domain from p=none to p=reject, and what must be true before each step?

level: middleimportance: must knowfreq 42%

basics

~20 s

Publish p=none with rua, use the aggregate reports to fix every legitimate sender until it passes aligned, then move to quarantine and finally reject, rehearsing each step with t=y, which asks receivers to apply one level less.

open as a page

A retailer's order confirmations, sent through an outside commerce platform, pass SPF and DKIM yet fail DMARC; what is the likely cause, and how is it fixed?

level: seniorimportance: must knowfreq 50%

basics

~20 s

Both passes cover the platform's own domains, its bounce domain in MAIL FROM and its DKIM d=, so neither aligns with the retailer's From domain. Fix it with an envelope subdomain of the retailer's domain or a signature with the retailer's d=.

open as a page

In a DMARC record's rua= and ruf= tags, how are report destinations written, and what does the fo= tag change?

level: middleimportance: should knowfreq 24%

basics

~20 s

DMARC's rua= and ruf= each take comma-separated URIs, normally mailto: addresses, with commas and ! inside a URI percent-encoded. fo= (0, 1, d or s, colon-joined) picks which failures produce a failure report and is ignored without ruf=.

open as a page

Under RFC 9989 DMARC, with a record only at the parent domain, which tag sets the policy for existing versus non-existent subdomains?

level: middleimportance: should knowfreq 32%

basics

~20 s

A subdomain without its own DMARC record is covered by the organizational domain's record: sp if the subdomain exists, np if its lookup returns NXDOMAIN. A missing np falls back to sp, a missing sp to p.

open as a page

In DMARC identifier alignment, how does relaxed mode differ from strict mode, and what do the adkim and aspf tags each govern?

level: middleimportance: should knowfreq 40%

basics

~20 s

Relaxed alignment, the default, accepts an authenticated domain with the same organizational domain as the From domain; strict requires the identical name. adkim sets the mode for DKIM's d= domain and aspf for SPF's MAIL FROM domain, independently.

open as a page

DMARC aggregate reports from dozens of receivers show an unfamiliar source IP sending as your From domain - how do you tell a forgotten legitimate sender from forgery?

level: seniorimportance: should knowfreq 34%

basics

~20 s

Join every receiver's rows for that IP and read them as evidence: steady volume with raw DKIM or SPF passes for an outside service's domain suggests a forgotten sender; no signature, failures everywhere and scattered bursts suggest forgery. Confirm ownership internally.

open as a page

Several of a group's DMARC records still carry pct=25 or pct=0; what does that tag mean under RFC 9989, and how should the records be rewritten?

level: seniorimportance: should knowfreq 20%

basics

~20 s

RFC 9989 removed DMARC's pct, rf and ri, so conforming receivers ignore them: p=reject; pct=25 is read as full reject. Replace pct=0 with t=y, delete pct=100, rf and ri, and choose a single policy where pct was in between.

open as a page

Under RFC 9989 DMARC, what does a receiver apply to failing mail when the record carries t=y, and which older tag did t replace?

level: seniorimportance: should knowfreq 21%

basics

~10 s

With t=y a receiver applies the policy one level lower: reject becomes quarantine, quarantine becomes none, and none is unaffected; reports are unchanged. t replaced RFC 7489's pct, whose partial percentages were applied inaccurately.

open as a page

When an association's domain moves to DMARC p=reject, why do members' posts through outside discussion lists start bouncing, and what do list operators change?

level: seniorimportance: should knowfreq 26%

basics

~20 s

A list resends and often edits the post, so neither SPF nor DKIM passes aligned with the member's From domain, and receivers honouring reject refuse it. Lists respond by rewriting From to their own address and signing it.

open as a page

When a forwarder rewrites a message's envelope sender with SRS, why does the copy pass SPF yet fail DMARC's SPF alignment, and what can still give it a DMARC pass?

level: seniorimportance: should knowfreq 26%

basics

~20 s

SRS puts an address at the forwarder's domain in MAIL FROM, so SPF passes for the forwarder, which is unaligned with the original From domain. Only an aligned DKIM signature that still verifies after forwarding can give the copy a DMARC pass.

open as a page

Your legal team asks what personal data DMARC failure reports would bring in - how do you decide whether your domain requests them at all?

level: principalimportance: should knowfreq 18%

basics

~20 s

Request DMARC failure reports only for a defined diagnostic need: they can carry recipients, subjects and whole messages, most receivers send none, and aggregate reports already find senders. If you do, time-box them and control the ruf mailbox.

open as a page

As the lead for a holding company whose example.com is at DMARC p=reject, how do you decide sp, np and whether product subdomains run their own records?

level: principalimportance: should knowfreq 12%

basics

~20 s

Set np=reject because legitimate mail should never come from names that do not exist, keep sp no weaker than the subdomains' real readiness, and publish explicit records at every sending subdomain. Use psd=n only when a unit truly owns its subtree.

open as a page

Should an association's DMARC policy go to p=reject when its members post from that domain to outside mailing lists, and how do you decide?

level: principalimportance: should knowfreq 22%

basics

~20 s

No single answer fits: RFC 9989 advises against reject for domains whose users post to lists. Weigh forgery pressure, measured list exposure and goals like BIMI; a common outcome is quarantine on the people domain, reject on a bulk subdomain.

open as a page

Should a domain owner tighten DMARC alignment from the relaxed default to strict with adkim=s and aspf=s, and what does that decision buy and cost?

level: principalimportance: should knowfreq 15%

basics

~20 s

Usually not across the board: RFC 9989 notes nearly all owners find relaxed sufficient. Strict, with explicit records per From domain, closes subdomain-based passes and organizational-domain disagreements, but envelope subdomains and parent d= signatures stop aligning.

open as a page

Under RFC 9990 DMARC, why do receivers ignore an rua address on an outside report processor's domain, and what record authorizes it?

level: middleimportance: nice to knowfreq 16%

basics

~10 s

When the rua host's organizational domain differs from the policy record's, receivers MUST look for a TXT record starting v=DMARC1 at <policy-domain>._report._dmarc.<destination-host>, published by the report processor; without it the address is ignored.

open as a page

Under the BIMI draft, what DMARC policy must a domain publish before receivers will show its logo, and what else does a logo need?

level: middleimportance: nice to knowfreq 16%

basics

~10 s

BIMI requires DMARC quarantine or reject on both the organizational and From domains, no sp=none, and mail that passes DMARC, plus a default._bimi record pointing at an SVG logo and each receiver's own approval.

open as a page

In DMARC's SPF alignment, which SPF-checked domain is compared with From, and what replaces it on a bounce with a null MAIL FROM?

level: middleimportance: nice to knowfreq 18%

basics

~20 s

Only the MAIL FROM domain that SPF validated counts. When the reverse-path is null, as on a bounce, RFC 7208 defines the MAIL FROM identity as postmaster@ the HELO name, so that name is aligned; a separate HELO check never counts.

open as a page

An auditor finds DMARC records with a misspelled sp= value, two TXT records at one _dmarc name, and v=DMARC1 in second place; what does an RFC 9989 receiver do with each?

level: seniorimportance: nice to knowfreq 13%

basics

~20 s

A DMARC record not starting with v=DMARC1 is ignored; two DMARC records at one name are both discarded; an invalid sp (or p, or np) makes the receiver act as if p=none when rua is valid, and apply no DMARC otherwise.

open as a page

Under RFC 9989's DNS tree walk, whose DMARC record governs mail From app.shop.example.com when shop.example.com and example.com both publish one, and when would a PSL-based receiver disagree?

level: seniorimportance: nice to knowfreq 15%

basics

~20 s

example.com's record, through its sp or np: with no psd tag, the tree walk takes the name with the fewest labels as the organizational domain. A psd=n at shop.example.com would stop the walk there, which a Public Suffix List receiver ignores.

open as a page

What does an ARC seal (RFC 8617) record when a list or forwarder relays a message, and why can't it alone turn a DMARC fail into a pass?

level: seniorimportance: nice to knowfreq 14%

basics

~20 s

An ARC set records the authentication results an intermediary saw on arrival, signed and chained to earlier sets. A receiver may honour a valid chain from a sealer it trusts to override a DMARC fail, purely as local policy.

open as a page