At 9:14 on a Tuesday morning, an invoice leaves a bookkeeping service on behalf of a small firm called Example Co. Somewhere between that server and the customer’s inbox, in a span of time too short to measure with a wall clock, the receiving mail server will ask four questions about it, look up three DNS records, and reach a verdict. The customer will see none of this. She will see an invoice, or she will see nothing.
Most explanations of DMARC start with the acronym and work outward. This one follows the invoice. It is easier to understand a border crossing by watching one traveler go through it.
The message, and its two return addresses
Here is what matters about the invoice as it leaves the bookkeeping service:
MAIL FROM:<bounces@mail.ledger-service.net> (envelope sender, invisible to the reader)
From: Example Co Billing <billing@example.com> (what the reader sees)
DKIM-Signature: v=1; a=rsa-sha256; d=example.com; s=ls1;
h=from:to:subject:date; bh=...; b=...Notice that there are two sender addresses and they disagree. The envelope sender, also called the Return-Path or MAIL FROM, belongs to the bookkeeping service; it is where bounces go. The From header belongs to Example Co; it is what the customer reads. This arrangement is normal. It is also precisely the arrangement a forger uses, which is why the receiving server is about to become curious.
Question one: is this server allowed to send for the envelope domain?
The receiver takes the domain from the envelope sender, mail.ledger-service.net, and looks up its SPF record. The connecting IP address is on the list. SPF passes.
But look at what was proven. The bookkeeping service has confirmed that its own server may send mail for its own domain. Nothing has been said about example.com. SPF checks the envelope sender, never the From header, so on its own it would wave through any message whose sender bothered to publish an SPF record for a domain he controls. Forgers bother.
Question two: is the signature genuine?
Next the receiver reads the DKIM-Signature header. The d= tag names the signing domain, example.com, and the s= tag names the selector, ls1. The receiver fetches the public key from DNS at ls1._domainkey.example.com, recomputes the hash of the signed headers and the body, and checks the signature against the key. It verifies. DKIM passes.
That key is in Example Co’s DNS because someone at Example Co put it there when setting up the bookkeeping service. It is the firm’s way of saying, in advance, that this service may sign on its behalf. DKIM proves two things: the signed parts of the message were not altered in transit, and whoever signed it holds the private key for a selector published under d=.
Question three: what does the From domain say about itself?
Now DMARC begins. The receiver extracts the domain from the From header, example.com, and asks DNS for a TXT record at _dmarc.example.com:
"v=DMARC1; p=quarantine; rua=mailto:reports@example.com; adkim=r; aspf=r"If the From address had used a subdomain, say billing@invoices.example.com, the receiver would look first at _dmarc.invoices.example.com. Finding nothing there, it would fall back to the record at the organizational domain, example.com, and apply that record’s sp tag if it has one, or its p tag if it does not. If there is no record at either name, DMARC simply does not apply, and the message is judged by the receiver’s ordinary spam filtering. You can see exactly which record a receiver would find for any name with the DMARC checker.
Question four: does anything that passed belong to the From domain?
This is the step that makes DMARC more than the sum of SPF and DKIM. The receiver compares the domains that authenticated against the domain in the From header.
| Check | Result | Authenticated domain | From domain | Aligned? |
|---|---|---|---|---|
| SPF | pass | mail.ledger-service.net | example.com | No |
| DKIM | pass | example.com | example.com | Yes |
SPF passed, but for someone else’s domain, so for DMARC purposes it counts for nothing. DKIM passed for example.com, which matches. Under relaxed alignment, the default, the two domains only need to share the same organizational domain, so a signature from mail.example.com would also have aligned. Under strict alignment they must match exactly. Alignment has its own guide, because it is where most legitimate mail goes wrong.
DMARC needs one aligned pass. Not two. The invoice gets through on its signature alone.
The rule is an OR: SPF passing with alignment, or DKIM passing with alignment. The invoice has one. DMARC passes, and the policy is never consulted.
The verdict, and what happens when it goes the other way
Suppose instead that a stranger had sent the same invoice from a rented server, with billing@example.com in the From line and a new bank account in the body. His SPF might pass for his own domain; he might even sign with his own key. Neither aligns with example.com. DMARC fails, and now the receiver reads the policy:
p=none: deliver as usual, and note the failure in the report.p=quarantine: treat the message as suspicious. In practice, the spam folder.p=reject: refuse the message during the SMTP conversation. It is never delivered.
Two qualifications. If the record carries a pct tag, the policy is applied to only that percentage of failing messages, which is how careful people move toward p=reject without betting everything on one DNS change. And the receiver always has the last word. The standard allows it to override your policy when it has reason to, for example when it recognizes a message as coming through a known forwarder or mailing list. Your record is a strongly worded request. Receivers generally honor it, but the enforcement is theirs.
The outcome is written into a header on the delivered message, where you can read it:
Authentication-Results: mx.receiver.example;
spf=pass smtp.mailfrom=mail.ledger-service.net;
dkim=pass header.d=example.com header.s=ls1;
dmarc=pass (p=QUARANTINE) header.from=example.comPaste the full headers of any message into the email header analyzer and it will pick out these lines and explain each verdict.
The complication: forwarding
Now suppose the customer forwards all her mail automatically from an old address to a new one. The forwarding server passes the invoice along, and the new receiver runs the same four questions. This time SPF fails outright, because the connecting IP belongs to the forwarder, which is on nobody’s list. The DKIM signature, however, travels inside the message. As long as the forwarder did not alter the signed headers or the body, it still verifies, it still aligns, and DMARC still passes.
If the forwarder is a mailing list that adds a footer or rewrites the subject line, the signature breaks too, and a perfectly honest message fails DMARC. This is why a DMARC failure is not, by itself, evidence of forgery, and why signing with an aligned DKIM key matters more than any other single step. A sender that relies on SPF alone will fail every time its mail is forwarded.
Afterward: the report
The receiver does one more thing, later. Once a day, typically, it gathers every message it saw with example.com in the From line, groups them by sending IP address, and sends an XML summary to the address in the rua tag. The invoice becomes one count in one row: this IP, this many messages, SPF result, DKIM result, aligned or not, and what was done about it. No message content is included, and no recipient addresses.
The forger’s attempt appears in the same report, on a different row. That is the quiet half of DMARC: you get an account, from the receivers themselves, of who is using your name. The reports are XML and arrive in bulk, so most people hand them to a tool. The free DMARC report analyzer will read one file; DMARC monitoring reads all of them, every day, and keeps the history. If you are setting this up for the first time, the reporting setup guide shows where the address goes.
That is the whole mechanism. Two checks that existed already, one comparison that did not, a policy, and a report. The customer got her invoice. She has no idea how close she came to getting a different one.

