Let’s talk
GUIDE

How do I find out when an n8n workflow fails silently?

Give every published workflow an error workflow: a separate workflow that starts with the Error Trigger node and sends an alert. You set it in the workflow's settings, one error workflow can serve many, and it runs when an automatic execution fails. It does not run when you test by hand in the editor.

That catches runs that fail. The worse case is a run that succeeds and does the wrong thing: an empty value, a branch nobody handles, an answer in the wrong shape. Make those fail on purpose with the Stop And Error node, so they reach the same alert.

Updated Sep 30, 2026

How it works

  1. Keep failed runs

    In the workflow settings, keep "Save failed production executions" on, so there is something to open when the alert arrives.

  2. Create one error workflow

    A new workflow with the Error Trigger first and an alert after it (email, chat). It receives the workflow name, the last node executed, the error message and a link to the execution.

  3. Point each workflow to it

    Open each workflow's settings and choose it under "Error workflow". Test it by making a published run fail, since manual runs do not trigger it.

  4. Turn wrong results into failures

    Where a value must exist or an answer must have a shape, check it and route the bad case to Stop And Error, or to a person, instead of letting the run finish quietly.

  5. Replay the failed run

    Open the failed execution and use "Debug in editor": n8n loads its data into the workflow so you can fix the node and run it again with the same input.

What it costs to run

No separate list price: the error workflow is an ordinary workflow on the same n8n plan, and it runs only when something fails. What you keep to investigate depends on the execution settings above.

What we learned building it

From Betterlane's own lead handling: three n8n workflows built on n8n Cloud in September 2026 and tested with a seed address. Not a client system.

A form that "worked" read nothing
Our approval form returned its answers keyed by the field labels ("Your reply", "Decision"), not the field names we read. Every run succeeded with empty values. Only a live test with a real edit showed it.
Distrust the answer before acting on it
In reply triage, an answer from the model that is not valid JSON becomes a "model_failed" case and goes to a person. It never becomes an empty draft that looks sent.
Keep an alarm outside the workflow tool
If n8n does not accept a new lead (down, trial lapsed, misconfigured), our own admin site still sends its alert, so an outage of the automation cannot silence the lead.
Make silence itself visible
A line every evening at 20:00 Dubai counts leads, answers and the median time to first answer. A day of zeros is a question someone asks, not a gap nobody sees.

When it is not worth it

  • If nobody reads the alert channel, an error workflow only moves the silence somewhere else. Decide who receives it first.
  • For a one-off import you watch while it runs, the executions list is enough.

Questions

Why did my error workflow not run when I tested?

The Error Trigger only runs when an automatic execution fails. A run you start by hand in the editor does not trigger it.

Does the error workflow need to be published?

No. n8n says a workflow that uses the Error Trigger does not have to be published.

Can one error workflow cover all our workflows?

Yes. Set the same error workflow in each workflow's settings; the alert says which workflow and which node failed.

What if the trigger itself fails?

Then the workflow never started, so the error data carries the trigger error instead of an execution link. The alert still arrives.

Sources

Have a client workflow that fails and nobody finds out until the client does?

See the service: For agenciesLet’s talk

All guides