Email Automation Mistakes: Workflows That Backfire
Ten failures including the two that only appear months later, and how to catch each before a subscriber does.
On this page
The short version
- Ten failures, and two of them only appear months after the flow was built — which is exactly when nobody is looking at it.
- Automation fails silently. A broken trigger sends nothing and reports success; a stale message sends the wrong thing and reports success.
- Every one of these is caught by the same habit: read every live flow end to end, as a subscriber would receive it, on a schedule.
Automated flows are invisible by design. Once built they run without anyone seeing them, which is their whole value and the reason their faults persist.
Nothing in the reporting distinguishes a flow that is working from one that is sending the wrong thing to the wrong people. Both show sends, opens and clicks.
Ten failures and how each shows up
| Failure | What it looks like | Catch it with |
|---|---|---|
| No exit condition | Customers still being sold what they bought | Write the exit with the trigger |
| Exit never tested | Same as above, but you believed it worked | Perform the exit action and check |
| Two flows colliding | Two messages in one morning | A per-subscriber frequency cap |
| Trigger fires on a weak signal | Recipients cannot connect it to anything | Require the action to cost something |
| No re-entry rule | One subscriber in the log many times | Once per subscriber, period or object |
| Batched instead of immediate | Welcome email arrives next morning | Subscribe with a real address, watch the clock |
| Campaigns sent during a flow | Reads as two different senders | Suppress campaigns for anyone in a flow |
| Unsubscribe not honoured in a flow | Mail after opting out | Check every flow respects the global rule |
| Stale content | A price, feature or person that changed | Read every flow end to end, quarterly |
| Broken trigger | Send volume quietly at zero | Alert when a flow's volume drops to zero |
Why unsubscribes get missed inside flows
Most platforms honour an unsubscribe globally, and most is not all. Some treat flow messages as a separate stream, some classify commercial flows as transactional by default, and some only apply the suppression at enrolment rather than at each send.
That last case is the dangerous one: a subscriber enrolled before opting out continues receiving the rest of the sequence, because the check happened once at the start.
The test is direct. Enrol in a flow, unsubscribe, and confirm nothing further arrives. Nobody does this, and it is the only way to know.
The review that catches nine of the ten
Once a quarter, open every live flow and read it end to end as a subscriber would receive it — not the flow diagram, the actual messages in order.
It takes about an hour for a programme with four flows and it finds the stale content, the missing exits, the message referencing a person who left, and the sequence whose tone no longer matches how you write now.
The tenth — a broken trigger sending nothing — is the one a read-through will not catch, because there is nothing to read. That one needs an alert on send volume, which is a few minutes of setup.
The order to fix them in
Anything that mails people who opted out comes first, because it is a compliance matter as well as a courtesy.
Then anything that contradicts what a subscriber knows — selling them what they bought, or reminding them about a completed order. Those cost trust immediately.
Then collisions and timing. Then stale content, which is the largest job and the least urgent, since a slightly out-of-date message is better than a wrong one.
Flow audit
- Every flow has an exit condition, and it has been tested
- Unsubscribing mid-flow stops the remaining messages
- A per-subscriber frequency cap spans all flows and campaigns
- Campaigns are suppressed for anyone inside a flow
- Each behavioural flow has a re-entry rule
- The first message of every flow was timed with a real address
- Every flow read end to end within the last quarter
- An alert fires if any flow's send volume reaches zero
Frequently asked questions
How often should flows be reviewed?
Quarterly for commercial flows, twice a year for the rest, and immediately after any change to pricing, products or the people named in them.
What is the single most common fault?
A missing or untested exit condition. It produces the errors recipients notice most — being sold something they already own — and it is the easiest to prevent by writing the exit at the same time as the trigger.
Should old flows be switched off or deleted?
Switched off and kept, with a note on why. A deleted flow takes its history with it, and the reason it existed is usually the thing someone needs six months later.