First, separate hard, soft and block bounces
The headline bounce rate mixes three different events. Hard bounces are permanent: the mailbox or the whole domain does not exist. Soft bounces are temporary: a full mailbox, a message too large, a server that did not answer in time. Blocks are refusals aimed at the sender rather than the address, and many platforms file them under one of the other two headings, so the reason text matters more than the label the platform gives it.
Export the bounces from the last few sends and add three columns: the bounce type, the recipient's mailbox provider, and the list source the address came from. Then read the pattern rather than the total. Hard bounces spread across providers but concentrated in one source point at that source. Hard bounces among new sign-ups point at the form. The same addresses reappearing on consecutive reports point at suppression. Refusals concentrated at one provider, arriving across every source and quoting authentication, policy or reputation, point at the sender. Keep this split separate from placement: mail that was accepted and then filed as junk is a different symptom, covered in emails going to spam.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Hard bounces jump on the first send after a quiet period or a platform move | Last engagement date and original source of the bounced addresses | Stale addresses mailed after a long gap or migrated without history | Re-permission the old segment in small batches; suppress non-responders |
| Hard bounces concentrated in one recently added file | Bounce rate by list source and import date | Purchased, rented or event list with unverified addresses | Stop mailing the source and remove it; do not buy or rent lists |
| New subscribers bounce on their very first message | Bounced welcome messages; misspelt domains among them | Sign-up forms accepting mistyped addresses without confirmation | Add confirmation to open web forms; prompt for obvious domain typos |
| The same addresses bounce on send after send | Whether last week's hard bounces were mailed again; suppression sync across tools | Hard bounces not suppressed after the first failure | Suppress on first hard bounce and sync the list to every sender |
| Refusals quoting authentication, often at one provider first | Reason text; SPF, DKIM and DMARC results on a test message | Authentication failures rejected by the receiving server | Correct the records and signing; confirm alignment with the From domain |
| Refusals quoting policy or reputation at one provider, across all sources | That provider's postmaster view; recent complaint history | Provider blocks after reputation damage, reported as bounces | Pause mail to that provider's segment; fix the reputation cause; ramp back |
Fix, verify and prevent each cause
Addresses that went stale
Addresses decay. People change jobs, abandon free mailboxes and let domains lapse, so a segment that has not been mailed for a long time, or a list carried over from an old platform without its bounce and engagement history, returns a burst of permanent failures on its first send.
Fix: hold the old segment back from regular sends and mail it in small batches with a plain re-permission message, as the list building and consent guide recommends for any list whose origin or currency is uncertain; suppress everyone who bounces or does not respond.
Verify: each batch shows fewer hard bounces than the one before, and the main list's rate returns to its usual level once the old segment is separated.
Prevent: when migrating platforms, carry the suppression list, the last engagement date and the consent record for every address, and never let a segment go unmailed for so long that its first send becomes a test of whether it still exists. When a whole list is already in that state, the tutorial on restarting email to a dormant list sequences the clean-up, the re-permission message and the staged first sends.
A bought, rented or event list
A file nobody built through a sign-up form is the fastest way to a bounce spike. Purchased and rented lists carry abandoned and invented addresses, and badge scans or sign-up sheets from events carry handwriting errors and addresses given casually. Beyond the bounces, the people on such lists did not agree to hear from the sender.
Fix: stop mailing the source, suppress its bounces and complaints, and either remove it or send one re-permission message to the addresses that remain; the list building guide sets out why bought lists almost never pass the consent test.
Verify: the bounce rate on sends without the source falls straight back, and the source's own figures stop appearing in the next report.
Prevent: record a source on every import, refuse files without a consent record, and replace event scanners with a form the attendee completes themselves.
Typos at the point of sign-up
When welcome messages bounce, the address was wrong from the moment it was typed. Misspelt domains, missing characters and addresses entered on a phone keyboard are common, and a form that accepts anything resembling an address will store them. Bots filling forms add more.
Fix: switch on confirmation for forms on the open web, so an address joins the list only after its owner clicks a link, and add a gentle prompt for well-known domain misspellings; the list building guide covers when double opt-in earns its friction. For deciding which forms need confirmation and which can stay single opt-in, the single versus double opt-in comparison has a decision table by sign-up source.
Verify: welcome-message bounces fall to a trickle, and unconfirmed addresses expire instead of accumulating.
Prevent: review new sign-ups by form every week, so a form that starts producing bad addresses, often after a redesign or a bot attack, is caught within days.
Hard bounces that are mailed again
A permanent failure should be mailed once. If the same addresses appear on bounce report after bounce report, suppression is failing: the platform's automatic suppression is off, a second sending tool never received the list, or an import re-added addresses that had already failed. Providers notice senders that keep mailing addresses that do not exist, and the M3AAWG sender practices treat bounce handling as basic operator hygiene.
Fix: confirm hard bounces are suppressed on first failure, agree a rule for when repeated soft bounces count as hard, and run the bounce and suppression steps of the email list hygiene SOP after every send.
Verify: no address on last week's hard-bounce export appears in this week's send log, and a seeded test address that bounced is refused by every system.
Prevent: keep one global suppression list, synced to the CRM and every tool that can send, and check the sync whenever a new tool is connected.
Authentication failures refused at the door
Some providers refuse mail that fails authentication rather than filing it as spam, and the refusal arrives as a bounce whose reason mentions SPF, DKIM or DMARC. This is not a list problem at all, and suppressing the addresses would remove good subscribers.
Fix: read the authentication results on a real message, correct the records, and make sure the platform signs with your own domain and that the passing domain aligns with the From address, following the deliverability guide; Google's sender guidelines document authentication requirements for all senders and stricter ones for bulk senders. Where aggregate reports name the failing system, DMARC failing works through it source by source.
Verify: a fresh test passes and aligns, and the authentication refusals disappear from the next send.
Prevent: make authentication part of switching on any system that sends as the domain.
Blocks after reputation damage
When complaints or a volume jump damage reputation badly enough, a provider may stop accepting mail from the sender for a while, and the platform reports the refusals as bounces. The pattern is distinctive: refusals at one provider, spread across every list source, with reason text about policy, rate or reputation.
Fix: pause mail to that provider's addresses, find and remove the cause, which is usually a complaint spike from one source or a whole-list send after a long segmented period, then return with the most engaged segment and step volume up, as the deliverability guide describes for warming.
Verify: the provider accepts the first small batches without refusals, and its postmaster view stops deteriorating.
Prevent: watch complaints by provider and source after every send, so damage is caught before it turns into refusals.
Prevention: confirmation, staged imports and automatic suppression
Most bounce problems are let in at one of three doors, and each door can be closed with a standing rule rather than a clean-up. The first door is the form: confirmation on open web forms means an address is real and wanted before it is mailed, and an expiry on unconfirmed addresses stops half-subscribers from piling up. The second door is the import: every file goes into a holding segment with its source recorded, is mailed in small batches, and joins the main list only when its bounces and complaints look like the rest of the list. The third door is failure itself: a hard bounce suppresses the address everywhere at once, permanently, so it can never be re-imported by a later file.
Around those rules sits a light monitoring habit. After each send, read bounces by type, provider and source, not only the campaign total, and treat a rise in one cell of that grid as the alert. Once a month, run the full pass in the email list hygiene SOP, which adds inactive segments, re-permission and a consent sample. Put the bounce rate by type beside the complaint rate and the active share of the list as list-health indicators (KPIA metric tied directly to an objective.) in the monthly report, next to the click-through (CTRShare of impressions that result in a click.) and conversion figures (CVRShare of visits or clicks that convert.) the business already reads, so that a list starting to decay is visible before a provider starts refusing it. The fundamentals hub sets this routine alongside authentication and segmentation.
A software company's bounce spike after a platform move
A software company moves its 62,000-address list to a new email platform in May. The first newsletter on the new platform returns a 6.8% bounce rate against a usual 0.4%. Split by type, 5.9% are hard bounces and almost all sit in a 14,000-address segment imported from the old CRM, which held contacts from trade shows between 2019 and 2022 and had never been mailed by the old platform. A further 0.6% are refusals at one provider quoting DKIM: the new platform is still signing with its own domain. The team suppresses the 3,700 hard-bounced addresses, moves the rest of the CRM segment into a holding group for a re-permission message in batches of 1,000, and switches on signing with the company's domain. The next newsletter, sent without the holding group, bounces at 0.3%. The re-permission batches confirm 2,100 people; the remaining 8,200 are suppressed.
Common mistakes
Reading the total bounce rate without splitting it by type. Suppressing addresses that were refused for an authentication fault. Migrating a list without its bounce history. Treating verification as a replacement for consent. Mailing an event file with the main list on its first send. Letting a second sending tool ignore the suppression list.
