Skip to main content

Bounce rate and suppression

A bounce is a receiving mail server refusing a message because of the address it was sent to. It is the clearest signal you get about list quality, and it is the number that decides whether your mail keeps arriving: mailbox providers watch bounce rates to judge whether a sender is careless, and a careless sender gets filtered long before anyone tells them.

Your bounce rate is bounces divided by everything you sent in the window. A message counts once, however many attempts it took. When a receiving server puts a message off and accepts it on a later attempt, that message is delivered, not bounced.

Only refusals that are the recipient's fault count as bounces. A server refusing your content or your reputation, a server that gave no reason, and a failure on the Epostix side are not bounces, and none of them moves the bounce rate in campaign reports or bounce alerts. See who is at fault.

Where you see it​

Delivered, bounced and failed for the window

Overview shows it three ways: as one of the four measures, as the red segment under Delivered volume, and per domain in the Sending domains table.

A bounce records a rejection or non-delivery report from a receiving server. The email can also have a failed status. Do not add raw failed and bounced counts as if they were disjoint. A failed message says why in its failure: bounced is a refusal of the address, and every other reason, such as suppressed or attachment_too_large, is a failure that no receiving server judged. See email.failed for the full list, and analytics for how API counts are measured.

The thresholds​

Epostix flags a domain at two levels.

Bounce rateWhat happens
Under 2%No bounce-rate warning at this threshold; inspect unexpected bounces anyway
2% or higherThe domain is marked Bounce rate high and the figure turns amber
5% or higherThe domain is marked Bouncing heavily, the figure turns red, and it is raised in the attention band at the top of Overview

Bounce rate per sending domain

These thresholds are per domain, not per workspace. One bad domain is flagged on its own even when your overall rate looks healthy. That is the point, because reputation is tracked per domain.

Who is at fault​

Every refusal is read for who the receiving server held responsible. That verdict, the fault, decides whether the refusal is a bounce, whether it suppresses the address, and whether you hear about it.

FaultMeaningCounts as a bounceCan suppressYou see it
RecipientThe address cannot take mail: it does not exist, is disabled or full, or its domain accepts no mailYesYesYes, as Bounced
SenderThe server refused the message or your sending reputation: content, blocklists, authentication, rate limits or policyNoNoYes, as Refused by receiver
UnknownThe server refused the message without saying why, or no mail server of the domain could be reachedNoOnly an unreachable domainYes, as Reason not stated
PlatformThe failure was on the Epostix side, such as a connection or configuration problem of oursNoNoOnly as a failed message

A recipient fault is either permanent, a hard bounce, or temporary, a soft bounce. A hard bounce says retrying will not help. A soft bounce says the address exists but could not take the message now, such as a full mailbox.

Hard bounces are the ones that damage reputation. A high hard-bounce rate says you are mailing addresses you never confirmed.

A sender fault says nothing about the address, so the next message to it may arrive. A run of them at one provider is the sign to look at your content, your authentication and your sending volume rather than your list.

Platform failures never reach your bounce rate or produce email.bounced, because there is nothing for you to fix. Epostix retries most of them while the message can still be delivered. A message that cannot be fails with a failure.reason of platform_error and a failure.fault of platform, and its reason says only The message could not be delivered.

Reasons​

Each refusal also carries a reason, in webhooks and in the API.

ReasonFaultWhat the server said
mailbox_unknownRecipientThe mailbox does not exist
mailbox_disabledRecipientThe mailbox exists but has been disabled
mailbox_fullRecipientThe mailbox is over its quota
domain_unroutableRecipientThe domain has no working mail server, or declares that it accepts no mail
blocklistSenderYour sending IP or domain is on a blocklist
reputationSenderYour sending reputation is too poor
rate_limitedSenderYou are sending faster than the server accepts
policySenderThe message breaks a policy of the receiving server
contentSenderThe content was judged spam or unsafe
authenticationSenderSPF, DKIM or DMARC did not pass
sender_addressSenderThe From or bounce address was refused
domain_unreachableUnknownNo mail server of the domain could be reached until delivery was given up
unclassifiedUnknownNothing in the reply says why

What Epostix does automatically​

You do not have to prune your list by hand. Bounces suppress addresses on their own. A recipient fault suppresses, and so does a domain whose mail servers stayed unreachable. A sender fault, any other unknown fault and a platform failure never do.

An address that does not exist is suppressed permanently, on the first occurrence. That covers an unknown mailbox, a domain that a receiving server reports cannot take mail, and a domain that publishes a null MX record to say it accepts none. There is no tolerance here, because a second attempt at an address that does not exist is the exact behaviour providers penalise.

Every other bounce is tolerated, then escalates. That covers soft bounces, a domain whose mail servers could not be reached, and the hard bounces that may not last, such as a disabled mailbox. An unreachable domain escalates like a soft bounce, but only as far as a temporary suppression: on its own it never becomes permanent, and it never counts toward your bounce rate, because an outage at the receiving end is nobody's sending mistake. One on its own changes nothing, because a full mailbox or a server having a bad morning is not a reason to give up on an address. An address that keeps failing is suppressed temporarily, and one that is still failing after that becomes permanent.

A bounce reported after delivery suppresses only when the report clearly names the failed recipient. Otherwise it is recorded and shown, and the address stays as it was.

While an address is temporarily suppressed it cannot escalate further, so a single bad week does not push a recoverable address to permanent.

The counts and the windows behind this are tuned as sending conditions change, so treat the behaviour as described here rather than as fixed numbers to plan around.

note

Suppression is per address, not per campaign. Once an address is suppressed for bouncing, nothing you send reaches it until the suppression lifts, including transactional mail.

A bounce suppression belongs to your workspace. It comes only from mail you sent, and it never suppresses the address for anyone else. A delivery to the address lifts a temporary suppression. A permanent one stays.

Recent hard bounces on Epostix​

Epostix keeps one short list across all senders to protect the shared sending IPs. When an address hard bounced on Epostix in the last 14 days, a message you send to it is not sent. It fails as Bounced, with the source suppression_list and the reason of the original bounce. It counts toward your bounce rate, and it suppresses the address in your workspace. A message is not held when you delivered to the address after that bounce.

The list never says who the other sender was, and it does not appear in your suppression list.

Complaints​

A complaint is a recipient marking your message as spam. Mailbox providers that run a feedback loop pass those reports on, and Epostix matches each one to the message it is about and raises email.complained. Gmail and iCloud do not report individual complaints, so a recipient there marking you as spam is never seen.

A complaint suppresses the address in your workspace as soon as the report is processed. What it stops depends on what the recipient complained about:

Complaint aboutWhat stops
A marketing messageMarketing mail to that address. Transactional mail, such as receipts and password resets, still goes through
A transactional messageEvery message to that address

If the address is a contact in your audience, the contact is also unsubscribed.

You cannot remove a complaint suppression. It appears in your suppression list with the reason complaint, and removing it is refused with a validation_error. The recipient has told their provider that your mail is unwanted, and mailing them again is what providers penalise hardest. If you believe a complaint was recorded against the wrong address, contact support.

Some reports can be matched to your message but cannot be proven to come from the provider. Epostix still suppresses the address for those, but they do not raise email.complained and do not count toward your complaint rate.

Epostix sets the Feedback-ID and X-Postmaster-Msgtype headers on every message. Providers use them to group complaint and spam-rate statistics, so their values identify your workspace, the stream (marketing or transactional) and the campaign, without naming your domain. If you send either header, or X-Mailru-Msgtype, through the API or SMTP, your value is dropped.

When the rate is high​

Work down this list in order.

  1. Check which domain. The Sending domains table attributes bounces per domain. A workspace-wide rate hides the fact that one domain is usually responsible.
  2. Check what changed. The Delivered volume chart shows bounces per day. A rate that jumped on one day is a bad import or a single campaign; a rate that climbed steadily is list decay.
  3. Check the domain is still authenticated. A domain whose SPF or DKIM record has stopped resolving will bounce at providers that reject unauthenticated mail. Its DNS health panel says so directly.
  4. Stop sending to unconfirmed addresses. Purchased, scraped, or years-old lists are the usual cause. Nothing in the platform can repair a list that was never opt-in.

Bounce rate responds slowly. Once providers have judged a domain, cleaning the list stops the damage but does not immediately undo it. Recovery is measured in weeks of clean sending, which is the strongest argument for keeping the rate low while it already is.