Email Deliverability Basics: Reputation, Warm-Up, and Bounces
How sender reputation is built and lost, why new IPs and domains need warm-up, how to handle bounces and complaints automatically, and the list hygiene that keeps mail in the inbox.
Deliverability is the difference between mail that is accepted by the receiving server and mail that actually reaches the inbox. Authentication records prove who you are; deliverability is about whether mailbox providers trust you, and that trust is earned through reputation, built gradually through warm-up, and kept through careful handling of bounces and complaints. This guide covers the mechanics a sender on Amazon SES, Google Workspace, or a self managed server needs to understand before volume grows.
How reputation is built and lost
Every large mailbox provider keeps a score for the sending IP address and, increasingly, for the sending domain. The inputs are broadly the same everywhere: the share of recipients who mark a message as spam, the share of addresses that bounce, whether recipients open and reply or delete without reading, whether SPF, DKIM, and DMARC pass, whether the sending volume is consistent or spiky, and whether the IP or domain appears on public blocklists. Google Postmaster Tools and similar programmes at other providers give a partial view of your standing.
Domain reputation follows you when you change providers or IPs, which is why authentication with your own domain matters and why a new domain should not be thrown away after a bad campaign. IP reputation is shared on a shared pool, which is fine for most senders because the provider polices the pool, and is entirely yours on a dedicated IP, which is better only once you send enough to keep the score alive. On Amazon SES the shared pool is the default; dedicated IPs are an option for senders with steady, meaningful volume.
Reputation is lost faster than it is built. A single send to an old list can produce enough complaints and unknown-address bounces to move a domain from inbox to spam for weeks.
Warm-up: earning volume gradually
A new IP address or a new domain has no history, and mailbox providers treat sudden high volume from an unknown source as a spam signal. Warm-up means starting with a small daily volume to your most engaged recipients and increasing it step by step over days or weeks, watching bounce and complaint rates at each step and holding or stepping back if they rise. The exact schedule depends on your list and the providers you send to; volume should grow no faster than the positive signals that justify it.
Amazon SES handles part of this for you. New accounts start in the sandbox and then with a modest sending quota that AWS raises as sending quality is demonstrated, and dedicated IPs can be warmed automatically by SES, which routes a rising share of your mail through the new address over a set period. Google Workspace mailboxes have per user sending limits set by Google that suit normal business correspondence and are not intended for bulk sending, so a newsletter should go through a dedicated service rather than a mailbox.
Warm up the domain as well as the IP. If you move transactional mail to a new subdomain, let it run for a while before you add marketing volume, and send the first campaigns to recipients who opened recently.
Bounces and complaints
A hard bounce means the address is permanently undeliverable: the domain has no mail server or the mailbox does not exist. Remove hard bounced addresses immediately and never send to them again. A soft bounce is temporary: a full mailbox, a server that is down, or a greylisting deferral. Retry a few times over a day or two, then suppress. Sending repeatedly to addresses that do not exist is one of the clearest spam signals.
A complaint is a recipient pressing the spam button. Mailbox providers that operate feedback loops send the complaint back to the sender, and services such as SES receive it and publish it to you as an event. Suppress the address at once and look at what the message was, because a cluster of complaints usually points to a campaign, a segment, or a template rather than to individuals.
Automate all of this. On SES, subscribe an SNS topic to the identity’s bounce and complaint notifications or use a configuration set with event publishing, and write the handler that updates your own suppression table. SES also keeps an account level suppression list and monitors each account’s bounce and complaint rates against thresholds documented by AWS; an account that exceeds them can be placed under review or have sending paused.
| Signal | What it means | Action |
|---|---|---|
| Hard bounce | Address or domain does not exist | Suppress permanently, investigate the list source |
| Soft bounce | Temporary failure at the receiving side | Retry with backoff, suppress after repeated failures |
| Complaint | Recipient marked the message as spam | Suppress immediately, review the campaign that caused it |
| Unsubscribe | Recipient withdrew consent | Honour promptly and keep a record |
| Blocklist listing | IP or domain flagged by a reputation service | Stop sending, find the cause, follow the operator’s delisting process |
| Deferral | Receiver is rate limiting or greylisting | Slow down, keep the retry queue, do not add volume |
List hygiene and consent
Most deliverability problems begin with the list. Addresses collected years ago, scraped from websites, or bought from a vendor produce bounces, complaints, and hits on spam traps, which are addresses maintained by blocklist operators to catch senders who mail without consent. There is no technical fix for a bad list; the only fix is not to send to it.
- Collect addresses with clear consent and record the time, source, and what the person agreed to receive.
- Use double opt-in where the audience or the applicable law expects it, so that typos and fake addresses never enter the list.
- Remove or re-permission addresses that have not engaged for a long period rather than continuing to send.
- Validate address syntax at signup and reject obvious role addresses if they are not appropriate for the content.
- Keep transactional and marketing lists separate and never move an address from one to the other without consent.
- Honour unsubscribes and complaints across every system that sends as the domain, not only the one that received them.
Consent is also a legal requirement in most jurisdictions, and the rules in the transactional versus marketing email guide apply regardless of how good the infrastructure is.
Content and sending practice
Authenticate every sending domain with SPF, DKIM, and DMARC, as described in SPF, DKIM, and DMARC setup; unauthenticated bulk mail is now refused or filtered by the large providers. Use a consistent From name and address so recipients recognise you. Keep the ratio of text to images sensible, avoid link shorteners and mismatched link text, and make sure every link resolves to a domain with a reputation you control. Include a visible unsubscribe link in marketing mail and the List-Unsubscribe headers that let mailbox providers offer one-click unsubscribe. Send at a steady cadence rather than in rare, large bursts, and segment so that the least engaged recipients receive the least mail.
Finally, monitor. Track delivery, bounce, complaint, and open rates per identity and per campaign, register your domains with the postmaster tools that providers offer, and check blocklists periodically. A problem noticed within a day is a paused campaign; a problem noticed a month later is a domain to rebuild.
Where to go next
The Amazon SES accounts page lists SMTP limit options, and the Google Workspace account page lists mailbox options for business correspondence. Each is subject to its provider’s terms and acceptable use policy, and any sending must be permission based and comply with anti-spam law.
To understand how SES quotas and production access interact with warm-up, read Amazon SES explained. If you send from a VPS, port 25 and cloud email sending explains why direct delivery is restricted and how relays fit in. The rest of the series is in the Cloud Library.