Handling a bounce, end to end
A bounce says the message did not reach the recipient, and the first job is to decide whether that is permanent or temporary. A permanent bounce suppresses the recipient and changes what the person's record claims about reaching them. A temporary one leaves it alone. Neither produces an automatic loop of re-sends, and neither is a reason to invent a different one.
The flow, stage by stage#
- Trigger — a message that was accepted for delivery is returned or reported as undeliverable.
- External event — the receiving side tells the sending side, either in a returned message or in a provider notification.
- Authentication and workspace resolution — the report is matched to the mailbox that sent the original, and the mailbox names the workspace.
- Ingest — the report is normalised to two facts: which address failed, and whether the failure is permanent.
- Canonical record — the outcome is written against the message that was sent, so the thread shows what happened rather than showing a message that appears delivered.
- Classification and reasoning — permanent or temporary. Everything downstream depends on this one decision, so it is made from what the report says rather than from a guess.
- Knowledge, memory and rules — the person's record is updated where the failure is permanent, because contactability is a fact about the relationship.
- Autonomy and approval — unaffected. A bounce is an outcome, not a permission question.
- Action through a provider — none. Nothing is sent in response to a bounce, and no address is probed to find a working one.
- Result — for a permanent failure, a suppression carrying
bounceas its origin; for a temporary one, a failed message with its reason and nothing else. - Relationship, timeline and memory — the failure is visible on the person, so the next person to open the record knows the address does not work.
- Audit, usage and Needs You — the outcome is recorded, and repeated failures against one mailbox surface as an operational problem rather than as a run of unlucky messages.
Permanent and temporary#
| Classification | Typical meaning | What changes | Retry |
|---|---|---|---|
| Permanent | The recipient does not exist, or the receiving side refuses it outright | A suppression with bounce as its origin; it is no longer treated as reachable | Never — it is wrong, and repetition cannot make it right |
| Temporary | The mailbox is full, or the receiving server declined for now | The message is recorded as failed with its reason; the recipient keeps their standing | Not automatic — the work that produced the message is what may run again later |
| Unclassifiable | The report does not say clearly | Recorded as a failure without a suppression | A person decides; nothing is assumed in either direction |
Suppressing on a temporary failure would delete a real relationship because somebody's inbox was full on a Tuesday. Not suppressing on a permanent one leaves a workspace sending into an address that has not existed for a year, collecting failures that make its other mail less likely to arrive. The classification is therefore the whole of the work.
What a bounce is not permission to do#
- Not permission to guess another one. A bounced address is evidence that it is wrong, not a hint about what the right one is. Connect does not construct an address from a name and a domain, and a bounce does not change that.
- Not permission to re-send the same message harder. A clean failure can be sent again once its cause is fixed; a repeated attempt against a permanent failure is a message to nobody and a mark against the mailbox.
- Not a reply to write. Nothing outbound is generated by a bounce, and there is nobody at the failing address to receive it.
- Not a reason to delete the person. The relationship, the history and the company link survive; what changes is that one address does not work.
What you see, and what to do#
Open the thread. A bounced message shows its outcome rather than reading as delivered.
Result You are not left comparing your sent folder against a silence to work out what happened.
Read the classification before acting.
Result Permanent and temporary need opposite responses, and acting on the wrong one is how a good address gets suppressed or a dead one gets hammered.
For a permanent bounce, look for another route to the person on their record.
Result A person can hold identities on more than one channel, and the timeline shows which of them has actually worked.
For repeated temporary failures across many recipients, treat it as a mailbox health question.
Result A pattern spread across recipients is about the sender or its provider, not about the people being written to.
Failure modes#
- The bounce cannot be matched to a message
- The address is still recorded as having failed. Matching to the original is a convenience; the failure is the fact.
- A bounce arrives long after the send
- It is recorded when it arrives. Nothing already delivered elsewhere is affected, and nothing is re-sent to compensate.
- An address bounces and later works
- Clearing the suppression is a deliberate act with a visible origin, the same as clearing any other.
- A sequence continues after a bounce
- Check whether the failure was classified as temporary. A permanent one stops the sequence through the suppression rather than through the sequence's own logic.
Questions#
Does Connect try a different address after a bounce?
No. An address that failed is evidence about itself and nothing else. Connect does not construct or guess one from a name and a domain, and a bounce is not treated as an invitation to start.
Why was a full mailbox not suppressed?
Because that failure is temporary. Suppressing on it would remove a working relationship because somebody was over quota for a day, which is a permanent response to a passing condition.
Is the person deleted when their address bounces permanently?
No. The person, their history and their company link all stay. What changes is that one address is no longer treated as a way to reach them, and their record shows why.