Six Hops and a Verdict: Reading the Journey Buried in a Received Header
Photo: BrokenSphere, CC BY-SA 3.0, via Wikimedia Commons
When a message lands in Junk, most of us get one piece of information: a flag. No reasoning, no appeal, no notes. That flag is the end of a process that might have involved five machines, two providers and eleven minutes of queuing — and almost all of that story is still sitting in the message, in plain text, waiting to be read.
The Received: header is the closest thing email has to a flight recorder. Every gateway that touches a message is supposed to stamp it, and those stamps stack up in a defined order. Read them the wrong way round and you'll conclude the wrong server is at fault. Read them properly and you can usually point at one hop and say: that's where the score changed.
Start at the bottom, not the top
Here's the bit that trips people up. Each server prepends its own Received: line to the top of the header block. So the message you're looking at in your mail client has the newest hop at the top and the oldest at the bottom. Chronological order runs upwards.
That means the first line you see was added by your own mailbox provider — the last gate the message passed. The bottom line is the sending side, and it's the one most likely to be forged or simply unhelpful. Typical structure:
Received: from mx3.mailhost.example (mx3.mailhost.example [198.51.100.24])
by inbound.mailprovider.co.uk with ESMTPS id 4Xy2...
for <[email protected]>; Sat, 21 Nov 2026 09:14:02 +0000
Four fields earn their keep: from (who claims to be handing over), by (who accepted it), the IP in brackets, and the timestamp. The with field tells you the protocol — ESMTP, ESMTPS, ESMTPSA, LMTP — and the id is what you quote when you raise a ticket with your provider.
Build the map before you build a theory
Pick a single message. Copy the full headers into a text file. List every Received: line, oldest at the top of your list. For each, note the timestamp. The gaps between timestamps are often more interesting than the hops themselves.
A short hop — under a second — is normal machine-to-machine handover. A hop of thirty seconds to a few minutes suggests a queue. A hop of several minutes to an hour, immediately followed by a successful delivery, is the classic signature of greylisting: the receiving server bounced it with a temporary 4xx, the sender retried, and the second attempt sailed through.
Worth knowing: greylisting is still alive in 2026, particularly on smaller UK hosts and on mail destined for addresses with thin traffic. It's cheap, it's unglamorous, and it still catches a lot of low-effort bulk.
The X-headers do the talking
The Received: chain tells you where a message went. The X- headers tell you what each stop thought of it. These are non-standard, which is exactly why they're useful — every vendor invents its own.
Things to look for:
X-Spam-Score/X-Spam-Status— SpamAssassin-family scores. A score that jumps between two hops means a filter added points, not that the sender got worse.X-Spam-Bar— the+++++version. Crude, but quick to eyeball.X-MS-Exchange-Organization-SCL— Microsoft's Spam Confidence Level.5–6is spam-filtered,9is high confidence. If a message showsSCL:1from Microsoft and still landed in Junk, the decision was made somewhere else downstream.Authentication-Results— the actual SPF, DKIM and DMARC verdicts as evaluated by a named host. Note theauthserv-id; it tells you who's speaking.X-MailerandX-Originating-IP— frequently absent, occasionally gold. A consumer webmail domain claiming anX-Mailerof a bespoke PHP library is a contradiction worth chasing.
The relay you didn't know you had
The header that causes the most raised eyebrows is an X- header naming a service nobody in the room recognises. Common culprits: an old marketing platform still wired into a contact form, a filtering appliance inherited from a previous MSP, a mail archiver, or a scanner that still routes outbound mail through itself.
If you find one, don't panic — but do find out whether it's still trusted. A forgotten relay with a permissive IP range is a standing invitation, and it will never appear on anyone's asset register.
A worksheet you can actually fill in
During an incident, structure beats intuition. Print this, or copy it into a spreadsheet, and work bottom-up:
| Hop | Timestamp (UTC) | From host | By host | IP | Gap (s) | Auth result | Spam score | Note |
|---|---|---|---|---|---|---|---|---|
| 1 | — | sending side | ||||||
| 2 | ||||||||
| 3 | ||||||||
| 4 | ||||||||
| 5 | last gate |
Fill the Auth result column from Authentication-Results, not from SPF records you looked up separately afterwards. You want what the filter saw, at the moment it saw it.
Reading the result
Three patterns cover most real incidents.
The score jumps at hop three and stays there. The filter at hop three made the call; everything after was just transport. Go and ask that vendor which rule fired, and take the message ID with you.
Authentication passes at the edge and fails at the mailbox. Something rewrote the message in transit — a forwarder, a mailing list, an alias. Look for X-Forwarded-* and a second Authentication-Results block.
No jump at all, but it still went to Junk. Either a reputation signal that lives outside the headers (IP or domain history) or a user-level rule. Check the mailbox rules before you blame the stack. In our experience it's the rule more often than anyone wants to admit.
The verdict is one bit. The headers are the argument. Learn to read the argument and you'll stop guessing — and you'll stop raising tickets against the wrong vendor.