A weekly deliverability review that takes twenty minutes

Reputation moves slowly and then stops working all at once. The teams that avoid outages are not the ones with the most tools; they are the ones with a fixed review they actually run. This is that review, with the thresholds that turn an observation into an action.

Daily, two minutes

• Bounce strings, not bounce rate. Open your ESP's log and read the actual SMTP responses from yesterday. On Mailgun that is *Sending → Logs*; on SendGrid, the Email Activity feed; on Amazon SES, your configuration set's event stream. One new recurring string from one provider is the earliest signal you will get.
• Deferrals. A rising share of 421 or 451 responses means a receiver is throttling you. Act by holding volume, not by retrying harder.
• Any mailbox that stopped sending. On mailbox-based outreach, an authentication failure (535 5.7.8, a revoked OAuth grant) silently removes a sender from the rotation and nothing looks broken.

Weekly, fifteen minutes

Run it in this order, because each step changes how you read the next.

1. Authentication. Confirm SPF still resolves within 10 DNS lookups, the DKIM selector you sign with is still published, and DMARC is unchanged. Someone edits DNS every week in every company.
2. Google Postmaster Tools. Check the user-reported spam rate and the SPF/DKIM/DMARC pass rates for each sending domain. Google's published bulk-sender threshold is a spam rate below 0.3%; treat anything sustained above 0.1% as a problem to fix this week. The domain reputation chart no longer exists, so spam rate and auth rates are what you have.
3. Microsoft SNDS, if you own the sending IPs. Nothing useful here for shared ESP pools or for 365 mailbox sending.
4. Rates against thresholds.

1. Blocklists. Check the sending domain and the IP. Know which lists you are querying: SORBS was shut down in June 2024, so a tool still reporting SORBS results is reporting nothing, while Validity's Heatwave blocklist launched on 3 September 2026 and is new enough that a first listing can surprise you.
2. Seed placement, per stream. One test per sending domain and provider, with the real production message. Compare to last week's, not to an absolute target.

Monthly, thirty minutes

• List decay. Re-verify, suppress role addresses, and drop or isolate the unengaged segment. B2B addresses rot continuously as people change jobs.
• Stream audit. Confirm transactional, marketing and outreach are still on separate subdomains and that nothing new has been pointed at the wrong one.
• DMARC aggregate reports. Read a month of them. They are the only place you will find out that a forgotten tool is sending as your domain.
• Volume consistency. Plot daily sends per domain. Spikes and gaps both cost you, and they are usually accidental.

Escalation rules, decided in advance

Write these down before you need them, because in the moment the temptation is always to keep sending.

• Gmail spam rate above 0.3%: stop marketing sends to unengaged segments immediately and send only to the most engaged cohort until it falls.
• Hard bounce rate above 5% on a single send: stop the sequence, re-verify the remaining list, resume at a third of the volume.
• A new 550 5.7.1 policy rejection pattern from one provider: pause that stream, check blocklists and authentication, do not retry at volume.
• A domain blocklisting: delist first, then resume at a low ramp. Sending through a listing extends it.
• An unexplained placement drop with clean authentication: look at what changed in the last seven days — a template, a tracking domain, a list import, a volume jump. It is almost always one of those four.

What not to do

• Do not build a dashboard before you have a habit. The review is the product; the tooling only saves minutes.
• Do not watch a single aggregate "deliverability score". It hides the per-provider split where the actual failure lives.
• Do not change three things after a bad week. You lose the ability to attribute the recovery.
• Do not skip the review during a good month. The slope you catch early is the one that is cheap to fix.

MailPilot automates the parts of this that are measurement rather than judgement: daily per-mailbox reputation tracking, SPF, DKIM and DMARC checks per sending domain, and real per-provider inbox placement through its seed network, for mailboxes connected over OAuth (Gmail, Outlook) or SMTP and IMAP (SendGrid, Amazon SES, Mailgun, Postmark, custom SMTP). The escalation decisions stay yours.