Email deliverability blog

Most guides show recipients the Not spam button. This one is for senders: what Yahoo publishes about authentication, reputation, complaints and deferrals.

No mailbox provider lists AI authorship as a spam signal. Their own ML filters judge identity, reputation and reactions. Here is where AI really adds risk.

Knowing what a bounce means is the easy part. Bounce management is the pipeline behind it: capture, classify, retry, suppress, alert and trace to the source.

"Outlook" is three filters wearing one name: Outlook.com's consumer filtering, Microsoft 365's tenant protection, and the recipient's own junk settings. What each one checks and which parts a sender controls.

Submission, transfer, IP and domain identity, encryption, feedback. A map of the machinery every message crosses, who runs each layer, and where to go deeper on each.

A 550 can mean the address is dead or that the receiver refuses you. The enhanced code and text decide which, and every variant has a different fix.

Gmail asks bulk senders for a DMARC record and accepts p=none, but not unaligned mail. The requirement, the error codes that name it, how Gmail enforces your policy, and why to go further.

Compliance gets bulk mail accepted; operations keep it accepted. Consent records, return processing, pacing, cadence and the loops a high-volume sender runs.

Receipts, resets and alerts fail one user at a time. Nine practices that follow a triggered message from the event that fires it to the bounce that comes back.

Two sets of numbers get called Gmail sending limits: the caps on what one account can send, and the 5,000-a-day line that decides which rules apply to your mail. What each is, and what neither is for.

Monitoring mode protects nothing; reject with unfinished authentication destroys your own mail. The staged migration between them: evidence, testing mode, quarantine, and rollback.

The gzipped XML in your rua mailbox answers one question per row: who sent as your domain, and did they authenticate? Every field decoded, plus the pass/fail combinations and what each means.

Filters grade the exact domain in your mail, and subdomains are graded separately. How to split marketing, transactional and corporate mail across subdomains without orphaning reputation.

Microsoft documents three general-purpose ways for apps and devices to send through Microsoft 365, plus an internal-only bulk option, and they are not interchangeable. The decision tree, the documented limits, and the ESP threshold.

Workspace gives a domain three SMTP doors out: the Gmail server, the relay service, and the restricted relay. What each one is for, the limits Google documents, and when the answer is an ESP.

One stream is triggered, expected, and exempt from unsubscribe rules; the other is scheduled, promotional, and consent-bound. Where the line runs, and why crossing it costs deliverability.

The 'Undeliverable' reply is a structured document with the answer inside it. Where the diagnostic line hides, what the codes mean, and the fix each common bounce-back text asks for.

Microsoft's free portal shows IP owners what Outlook.com thought of yesterday's mail: filter verdicts, complaint rates, trap hits. How to register, what each panel means, and what to do with red.

Password resets, receipts, shipping notices: the mail a user's own action triggers. What counts, what doesn't, why the law treats it differently, and why it deserves its own infrastructure.

Every email is relayed; the question is by whom, with whose reputation, and under what authentication. The mechanics of relaying, and how to pick a relay without wrecking deliverability.

Every WooCommerce order fires emails through WordPress's unauthenticated default mailer. Why 'where is my confirmation?' tickets happen, and the transport fix that ends them.

An order confirmation is the most wanted email a store sends — when even that folders, the sending domain setup is unfinished. The store-side checks for both mail streams.

Sending through Google's own servers doesn't exempt you from Google's own rules. SPF, admin-console DKIM, DMARC, and the bulk-sending line Workspace was never meant to cross.

Connecting a domain takes minutes; earning a reputation takes months. The HubSpot-specific checklist: domain connection, list provenance, engagement, volume shape.

A rep's one-to-one email and a bulk campaign leave Salesforce through different doors and fail differently. Which stream is failing, why, and the fixes per stream.

Out of the box, WordPress hands email to the web server's PHP mail() — an unauthenticated identity no filter trusts. The diagnosis, and the SMTP and API fixes.

The newsletter that reached everyone in spring lands in spam by fall. The causes are almost always in the account: authentication, list age, engagement. The checks, in order.

Mailbox providers publish exactly what they require of bulk senders. Meeting those requirements, plus consent and warm-up, is the entire legitimate playbook.

Delisting is a three-step job: confirm the listing, fix the cause, then follow each operator's documented removal process. The procedures, blocklist by blocklist.

The forbidden-words list is the most persistent myth in email marketing. Where it came from, the little truth left in it, and what modern filters actually score.

A bounce is a receiving server explaining, in code, why it refused your message. The causes, from dead addresses to reputation blocks, and what each one asks you to do.

Most of what a deliverability consultant sells is a documented procedure: authentication, infrastructure, reputation, list, placement. Here is the procedure, with the tools per step.

Industry averages get quoted as if they were physical constants. Where the figures actually come from, why methodology moves them, and the number that matters instead.

Domain authentication, audience decay, complaint creep, shared-pool mechanics — the ordered diagnostic for Mailchimp senders whose campaigns stopped inboxing.

Spam placement on Klaviyo is almost always sender-side, and fixable from inside the account: sending domain, list hygiene, engaged segments. The checks, in order.

Password resets arrive instantly while the newsletter lands in junk. The campaign-specific causes — bulk thresholds, list decay, cadence, template patterns — and the fixes.

Spam placement produces no bounce and no alert — you have to go looking. The free checks, the manual seed-test method, and the honest limits of each.

RFC 7208 gives every SPF evaluation a budget of ten DNS lookups, and includes spend it recursively. Counting rules, the failure it causes, and three ways to restructure.

Permerror doesn't judge your mail — it means receivers couldn't interpret your SPF record at all. The five causes, in order of likelihood, and the repair sequence.

Softfail is SPF's deliberately hedged verdict: probably not authorized, but don't reject. Where it comes from, how receivers and DMARC treat it, and when to fix it.

Five protocols carry every proof of identity in email. What each one does, which are mandatory, which are optional, and how they depend on each other.

Every check that protects inbox placement, sorted by when it runs: one-time setup, before every send, and the monthly monitoring loop. Work top to bottom.

A dedicated IP isn't an upgrade — it's a responsibility transfer. The real decision criteria: volume, consistency, isolation needs, and whether you can staff the reputation you'd own.

Deliverability is whether your mail reaches the inbox rather than the spam folder or nowhere. The actors, the inputs, the metrics and where to start — the pillar guide.

Bounce rate is the loudest list-quality signal receivers can see. What counts in the numerator, where the 2% convention comes from, and the reduction plan that works.

Spam-folder advice mostly arrives after the damage. Prevention is a different job: a handful of standing practices that keep the filtering evidence from accumulating at all.

A fresh IP has no history, and receivers throttle what they don't know. The ramp, the per-provider behavior to expect, and how to keep an IP warm once it is.

Every bounce poses one question: is this address dead, or is delivery merely delayed? The answer decides whether you suppress, retry or investigate — and the labels sometimes lie.

Every DKIM setup form asks for a selector, and no error message explains it. It's the label that tells receivers which of your public keys verifies a given signature.

The warning doesn't mean DMARC is missing — it means your policy is p=none, so failures are observed but never acted on. The fix is a staged move to enforcement.

Between 'send' and the inbox, Gmail runs a message through connection checks, authentication, reputation and engagement models. Here's the pipeline, stage by stage.

dmarc=fail has about six real-world causes, and one of them accounts for most cases. How to read the evidence and pin yours down.

The p= tag is the only part of DMARC with teeth. What none, quarantine and reject do at the receiver, and the staged rollout to enforcement.

Aggregate rua reports are the only free receiver-side view of your authentication. The XML, block by block, and what disposition none means.

SMTP never verified senders, so identity got bolted on through DNS. The authentication stack, layer by layer, and what receivers now require.

Deliverability is never finished. The five standing disciplines, with cadences, that keep inbox placement from decaying between crises.

Nine steps, in the order that actually moves inbox placement: authentication, list repair, engagement, and the monitoring that keeps gains from eroding.

Eight steps from no record to verified: prerequisites, building the record, publishing the TXT entry, and confirming receivers see it.

Three DNS records, three different questions answered. How SPF, DKIM and DMARC divide the work of proving your mail is really yours.

Plan on two to eight weeks, depending on target volume. What warm-up actually builds, a conservative ramp, and the mistakes that reset the clock.

Spam placement always has a cause. Here are the seven that account for nearly every case we see, ordered by how often each one turns out to be it.