The pre-send email deliverability checklist.

How to use this checklist

A checklist is useful only if a failed item changes what you do. Run this one close to launch, with the production sending domain, From address, ESP or CRM, tracking setup, and final template. Put a name beside each failure. DNS belongs to whoever controls the domain; audience quality belongs to the campaign owner; links and rendering usually belong to marketing operations.

Do not wait for every item to be perfect. Separate hard blockers from warnings and context. Failed authentication on the real message is a blocker. A new but legitimate sending domain is context that calls for lower volume and closer monitoring. One image without alt text is a quality issue, not proof the campaign will land in spam.

Identity and domain: checks 1–6

Start with the identity a recipient and a mailbox provider will see. The visible From domain should be owned by your organization, recognizable to the audience, and consistent with the message. Confirm the actual sending route rather than relying on a setup screen that says the domain is verified.

Pass means the delivered test message shows the expected domains and authentication results. If your ESP uses a different return path or signing domain, decide whether it aligns under DMARC. A DNS record that exists but is not used by the delivered message does not pass this stage.

• 1. The visible From domain is owned, monitored, and expected by recipients.
• 2. SPF authorizes every legitimate service sending for the relevant return-path domain.
• 3. The production message contains a valid DKIM signature with the expected signing domain.
• 4. At least SPF or DKIM aligns with the visible From domain for DMARC.
• 5. A DMARC record exists and its reporting mailbox is monitored before enforcement is increased.
• 6. Sending IPs have valid forward and reverse DNS when you control that infrastructure.

Audience and consent: checks 7–11

Deliverability begins before the message is written. Know how every address entered the audience, what the recipient was told, and how recently the relationship was active. Purchased, scraped, guessed, or poorly documented data creates complaint and trap risk that no subject-line edit can repair.

Pass means the audience has a defensible source, invalid addresses have been suppressed, and the campaign targets people likely to recognize the sender. For a large dormant segment, reduce the initial audience and win back engagement deliberately rather than using the full database as a test.

• 7. Every address has a documented consent or lawful outreach basis appropriate to the message.
• 8. Hard bounces, prior complaints, and unsubscribes are suppressed across every sending system.
• 9. New addresses are validated at collection or before a high-risk send.
• 10. The segment excludes people who have shown sustained non-engagement unless this is a controlled reactivation campaign.
• 11. The sender, topic, and frequency match what the recipient reasonably expects.

Infrastructure and sending behavior: checks 12–16

A clean list can still run into trouble when volume jumps without warning or unrelated traffic shares one reputation. Compare planned volume with the recent baseline. If the domain, IP, mailbox, or provider is new, start with people most likely to welcome and engage with the message.

Pass means the route is stable, the sending stream is identifiable, and the team has a response plan for unusual bounces or provider blocks. Transactional messages should not be dragged down by a risky promotional experiment if your platform lets you separate those streams.

• 12. The planned volume and cadence are consistent with recent legitimate traffic.
• 13. Marketing and transactional streams are separated where their risk and importance differ.
• 14. TLS is used and message formatting follows normal internet-message standards.
• 15. Provider responses, deferrals, and hard-bounce codes are being retained and reviewed.
• 16. Google Postmaster Tools, Microsoft SNDS, or equivalent provider data is connected when volume and IP ownership make it available.

Message, links, and HTML: checks 17–21

Read the final message as a skeptical recipient. The identity and purpose should be obvious without loading images. Links should go where their labels promise, use domains the reader can understand, and avoid public shorteners or unnecessary redirect chains. Open the exact URLs after tracking has been applied; the pre-tracking draft is not the final message.

There is no universal list of forbidden words. Mailbox filters use many signals and recipient context. Focus on clarity: no fake reply prefixes, no misleading urgency, no hidden unsubscribe path, and no template that is effectively one giant image.

• 17. The sender name, subject, and opening line accurately describe the message.
• 18. Every customer-facing link resolves over HTTPS and no placeholder URL remains.
• 19. Tracking uses a branded or understood domain with no avoidable shortener or redirect chain.
• 20. The email remains understandable when images are blocked and meaningful images have alt text.
• 21. Bulk or marketing mail includes visible unsubscribe handling and one-click unsubscribe where provider rules require it.

Placement and launch decision: checks 22–25

The last check is an observed send through the real route. Use seed inboxes that represent the providers and account types that matter to your audience. Keep the message unchanged. A good result is not just a high aggregate percentage; it is enough matched seeds, no unexplained provider cluster in spam, and no hard setup failure in the evidence.

Set the launch threshold before seeing the result. For a high-value campaign, you might delay if a major audience provider shows repeated spam placement or if too many seeds remain missing to interpret the test. Record the baseline and the decision. That history is more useful than a screenshot with no sender, date, or context.

• 22. A production-like message was sent through the exact platform and sender route planned for launch.
• 23. The seed mix covers the mailbox providers and personal or professional segments that matter to the campaign.
• 24. Pending, Promotions, Spam, and Missing are interpreted separately rather than forced into one pass/fail score.
• 25. The launch owner has recorded the result, any accepted risk, and the next monitoring or retest time.

A final placement test is where preparation becomes evidence.

What to do when a check fails

Fix in risk order. Start with a broken or misaligned sender identity, then list and consent problems, then broad reputation or infrastructure issues, then provider-specific behavior, and finally message-level details. If the campaign is already generating complaints or high hard-bounce rates, pause it; do not use a placement test as permission to keep sending harmful traffic.

Change one meaningful variable at a time when possible. Rerun the same message after the fix and compare provider-level outcomes. If everything changes at once, an improved result may feel good but teach you nothing you can repeat.