Deliverability Case Study: "Smells Like Spam Spirit"
The original runs on Kurt Cobain's shrug: a mumbled verse, a chorus that erupts, then "oh well, whatever, never mind." This parody gives that shrug to a spam filter. The narrator isn't a villain plotting against senders. It is a gatekeeper that has heard a billion knocks at the door, adds up the evidence on each one, and feels nothing either way. That indifference is technically accurate. A mailbox provider's filter doesn't hold grudges or hand out punishments. It scores signals and routes mail, and it never explains itself.
Here is the technical breakdown of the deliverability concepts detailed in the song:
Verse 1: The Gatekeeper Can't Read Your Mind
"Fire up your sends, invade my space / You want the inbox, win the race"
"Oh no, your motives are obscure"
- The Deliverability Context: The sender in this verse treats the inbox as a prize: hit send at full volume and get there first. The filter sees it differently. A sudden burst of mail from a sender with little history ("fire up your sends") looks the same as a compromised account or a spammer's first run. "Your motives are obscure" is the most honest line in the song. A filter has no access to intent. It never sees that you're a legitimate brand with a great offer, only observable evidence: who is connecting, from where, at what rate, with what authentication, and how recipients reacted to your last messages.
- The Strategy: Stop trying to win the inbox and start being predictable. Consistent volume, a consistent sending identity and a steady cadence give the filter a history to judge you by. A sender with no track record gets treated as an unknown, and an unknown arriving in bulk gets treated with suspicion.
Pre-Chorus: EHLO, Where the Judging Starts
"EHLO, EHLO, EHLO — oh no"
"EHLO, EHLO, EHLO — blocked out"
- The Deliverability Context: Cobain's "hello, hello, how low" becomes the command that opens every SMTP session.
EHLO (Extended HELO, defined in RFC 5321) is the sending server introducing itself by hostname before any message is transferred. The pun points at something many senders miss: filtering starts before your email exists. When the connection opens, the receiving side already knows your IP address and can look it up against DNS blocklists such as Spamhaus ZEN. At EHLO it learns the name you claim, and then it can check whether your IP has a reverse DNS (PTR) record and whether that name resolves back to the same IP. All of that happens before your subject line has been read.
- A Technical Nuance: RFC 5321 lets a receiving server verify that the EHLO name matches the connecting IP, but says it must not reject a message solely because that check fails. So a sloppy EHLO (a bare
localhost, an unqualified name, a hostname that doesn't resolve) won't usually trigger a bounce on its own. It becomes one more point against you in the score, which is the paranoia this song is about. Google's sender guidelines do require valid forward and reverse DNS for sending IPs, so that part isn't optional.
- "Oh no" vs. "blocked out": The pre-chorus changes after the first chorus. "Oh no" is the filter getting suspicious; "blocked out" is the verdict. A rejection at the SMTP level comes back as a
5xx code (permanent) or a 4xx deferral (try later). The sender at least finds out. Remember that when you get to Verse 3.
Chorus: Every Signal Adds Up
"Your domain's weak / Your reputation / Your headers stink / Authentication, yeah"
- The Deliverability Context: The chorus is a charge sheet, and it describes how modern filtering actually works. No single item is the verdict. The filter adds each one to a running total:
-
"Your domain's weak": Domain reputation is built from history. A newly registered domain, or one that has only ever sent in irregular bursts, has nothing to vouch for it. Domain reputation and
IP reputation are tracked separately, so a clean IP can't save a weak domain.
-
"Your reputation": The aggregate of past behavior: complaint rates (keep them under 0.10%, never approaching 0.30%), bounce rates (under about 2%), and how recipients engage with your mail.
-
"Your headers stink": Headers are evidence. A missing or malformed
Message-ID or
Date, a display name that impersonates a brand the address doesn't belong to,
Received chains that don't make sense, or bulk mail with no
List-Unsubscribe header all add to the score.
-
"Authentication, yeah": The receiver records its
SPF,
DKIM and
DMARC results in an
Authentication-Results header. If those results fail or don't align with your visible From domain, that's the loudest line on the charge sheet.
- The Fix: Treat every signal as cumulative, because it is. Senders often fix the item that tripped the last bounce and ignore the four minor ones. Those minor ones together are what push a borderline message into junk.
Verse 2: The Guard Has Seen It All
"I flag the ones who push too hard / Your broken auth won't fool my guard"
"My little filter's seen it all / And watched your finest campaigns fall"
- The Deliverability Context: "Push too hard" is rate. Mailbox providers throttle senders whose volume outruns their reputation. Gmail, for example, defers with
421 4.7.28 when it detects an unusual rate of unsolicited mail from an IP or domain. Retrying harder makes it worse. "Broken auth" is the quieter killer: a DKIM key rotated in your ESP but never published in DNS, an SPF record pushed past its 10-DNS-lookup limit into a permerror, a third-party sender signing with its own domain instead of yours. To the filter, broken authentication looks exactly like no authentication. Gmail answers unauthenticated mail with 5.7.26.
- "Your finest campaigns": This is the indifference at its sharpest. The filter has scored billions of messages and doesn't care how good your design was, how sharp the copy was, or how long the team worked on the launch. Content is a minor input next to identity and reputation. A beautiful campaign from a sender with a weak reputation still falls.
- The Strategy: Fix the boring infrastructure before polishing the creative. Monitor DMARC aggregate (
rua) reports so broken authentication shows up in a report, not in your campaign results. Back off when you see 4xx deferrals instead of pushing more volume into them.
Verse 3: The Irony, and the Silent Verdict
"And I forget what made you pass / Oh yeah, your tricks don't fool me fast"
"I find it hard, it's hard to trust / Oh well, whatever — straight to junk"
- The Irony: The filter admits it gets things wrong. Some mail passes that shouldn't, and the filter can't say why: "I forget what made you pass." Machine-learning filters don't keep a list of reasons. A verdict comes out of thousands of weighted features, and passing yesterday guarantees nothing today. Spammers exploit that gap with fresh domains, rotating IPs and shuffled content. Those tricks can get through for a while, but not for long ("don't fool me fast"). Once recipients start hitting "Report spam," the filter learns, and anything that shares that sender's fingerprint gets caught too. The lesson for legitimate senders: one campaign that landed in the inbox doesn't prove your program is healthy. It may only mean the filter hasn't caught up yet.
- "Straight to junk": Notice what the filter does not do here: it doesn't block. The pre-choruses say "blocked out," an SMTP rejection the sender can see. "Straight to junk" is worse in a way, because the receiving server answers
250 OK, your ESP records a delivery, and the message goes into the spam folder where nobody reads it. Your delivery rate says 99%. Your inbox placement rate can be far lower, and nothing in your sending logs will show it.
- The Resolution: Measure placement separately from delivery. Watch click-through rate (CTR) by mailbox provider (a sudden drop at one provider is often the first sign of junk-folder routing), and use seed-list placement testing to estimate where mail lands.
Outro: A Denial
"A denial, a denial..."
The original's closing chant becomes the filter's final word, repeated with no explanation. That's how rejection feels from the sending side: a status code, a line of terse text, and no conversation. The denial does contain information, though. Enhanced status codes like
5.7.26 (authentication),
4.7.28 (rate) and
5.7.1 (policy) tell you which part of the charge sheet failed.
The filter isn't your enemy and it isn't your friend. It's a bored gatekeeper adding up evidence. You can't change its mood, only what it sees.
Most senders picture the spam filter as an adversary. It's closer to a bureaucrat: it doesn't hate you, it just processes you. It starts adding up evidence when your server says
EHLO, keeps adding through your headers and authentication, and finishes with a verdict it won't explain. You can't argue with it or impress it. You can only control what it sees. Here is how to give it less to hold against you.
Pass the Door Check Before You Say a Word
The SMTP session is judged before your content ever arrives. Make your connection look like it belongs to a real, well-run mail server.
- Match Forward and Reverse DNS: Every sending IP needs a PTR record that resolves to a hostname, and that hostname must resolve back to the same IP (forward-confirmed reverse DNS, or FCrDNS). Google's sender guidelines require valid forward and reverse DNS.
- Use a Real EHLO Hostname: Your MTA should introduce itself with a fully qualified domain name that matches its PTR, not
localhost or a cloud provider's default machine name. RFC 5321 says a mismatch alone shouldn't cause a rejection, but filters still score it.
- Check Blocklists Before You Wonder Why: Look up your sending IPs and domains against Spamhaus ZEN and DBL whenever deferrals spike. A listing that blocks you at connection time can never be fixed by improving content.
Stop Feeding the Charge Sheet
Filters combine many weak signals into one score. Every small defect you remove lowers the total.
- Clean Up Your Headers: Send a valid
Message-ID and Date, use a From display name that matches the brand the address belongs to, and include RFC 8058 one-click unsubscribe (List-Unsubscribe plus List-Unsubscribe-Post) on bulk mail. Gmail and Yahoo have required it since February 2024.
- Keep Authentication Unbroken, Not Just Configured: SPF (Sender Policy Framework), DKIM (DomainKeys Identified Mail) and DMARC (Domain-based Message Authentication, Reporting, and Conformance) need to pass and align on every stream. Most breakage happens during changes: a new ESP, a rotated DKIM selector, a new sending tool added to SPF. Re-verify authentication after every infrastructure change.
- Read Your DMARC Aggregate Reports:
rua reports show which sources are failing authentication on your domain. They're how you catch "broken auth" before a campaign does.
Earn a Reputation Instead of Racing for the Inbox
"Fire up your sends" is how new and recovering senders get throttled. Reputation follows consistent behavior.
- Ramp Volume Gradually: New IPs and domains need a warmup: start around 200–500 messages a day to your most engaged recipients, roughly double every 2–3 days, and expect 4–8 weeks to reach full volume.
- Respect Deferrals: A
4xx response, like Gmail's 421 4.7.28 rate-limit deferral, means slow down. Configure your MTA to back off exponentially instead of hammering the same provider.
- Hold the Thresholds: Keep spam complaints under 0.10% (0.30% is where serious filtering begins) and hard bounces under about 2%. Suppress recipients who haven't engaged in 90–120 days so they stop dragging your engagement down.
Measure Placement, Not Delivery
"Straight to junk" never shows up as a bounce. A 250 OK only means the receiving server accepted the message, not that it reached the inbox.
- Separate the Two Numbers: Delivery rate counts what was accepted. Inbox placement rate (IPR) counts what reached the inbox. Only seed-list placement testing estimates the second.
- Segment Metrics by Provider: A CTR drop concentrated at one mailbox provider is a junk-folder signal. Don't use open rates for this. Apple Mail Privacy Protection inflates them, and they will stay high while your mail sits in spam.
- Don't Trust One Good Send: A campaign that reached the inbox doesn't prove your program is healthy. Filters learn from complaints after delivery, so judge your trend over weeks, not one result.
Read the Denial
When the filter does block you, the bounce tells you why. Most senders throw that information away.
- Parse Enhanced Status Codes:
5.7.26 points to authentication, 4.7.28 to rate, 5.7.1 to policy or content. Group bounces by code and by provider instead of logging them as one "failed" number.
- Never Retry a
5xx Indefinitely: Permanent failures need investigation and suppression. Temporary failures should retry with backoff within a bounded queue lifetime (48–72 hours at most).
Conclusion
The spam filter in this song isn't out to get you. It's indifferent, and that's good news. An indifferent system responds to evidence, so you can change the result by changing what it sees: a clean connection, intact headers, passing authentication, steady volume and a reputation that recipients' behavior backs up. Give it nothing to hold against you, and "oh well, whatever" works in your favor.
Your Gatekeeper Checklist:
- Confirm every sending IP has a PTR record, a matching forward lookup, and an EHLO hostname that agrees with both.
- Re-verify SPF, DKIM and DMARC alignment after every ESP, selector or sending-tool change, and review
rua reports weekly.
- Audit bulk mail headers for a valid
Message-ID, an honest display name and RFC 8058 one-click unsubscribe.
- Configure exponential backoff on
4xx deferrals and stop retrying 5xx bounces.
- Track inbox placement and per-provider CTR separately from delivery rate.
- Group bounces by enhanced status code so every denial tells you what to fix.
Educational content. Email deliverability evolves rapidly. Platform rules (Gmail, Yahoo, etc.), engagement signals, and ESP behaviours change frequently, and real-world issues often involve conflicting signals, data quality problems, and failure modes that general best practices can’t anticipate. Content on this site is provided for informational purposes only and does not replace a thorough analysis by a qualified deliverability professional.
Terms of Use