# n8n Go-Live Checklist: 10 Checks Before a Workflow Touches Real Data

- **url:** https://easify-ai.com/n8n-go-live-checklist-before-workflows-touch-real-data/
- **published:** 2026-11-16
- **topics:** technologies-deep-dive
- **publisher:** Easify AI, part of Easify Technologies

A workflow that works in testing is not the same as a workflow that is ready for production. In testing you are watching. In production, nobody is, so every failure needs to announce itself.

These ten checks are the ones we run before an n8n workflow is allowed near real customers, orders or invoices.

## The checklist at a glance

| # | Check | Area |
|---|---|---|
| 1 | An error workflow is set | Failure handling |
| 2 | Retries on calls to other services | Failure handling |
| 3 | Each node's error behaviour is deliberate | Failure handling |
| 4 | Failed executions are saved | Visibility |
| 5 | Steps are safe to re-run | Data safety |
| 6 | Tested with real-shaped data | Data safety |
| 7 | Credentials are production ones, with least access | Security |
| 8 | Volume and timeouts are checked | Performance |
| 9 | The workflow is documented | Ownership |
| 10 | A named owner gets the alerts | Ownership |

## Failure handling

1. **Set an error workflow.** Create a workflow that starts with the **Error Trigger** node, then point every production workflow at it in the workflow settings. That error workflow should send an alert and log what failed.
2. **Add retries to calls to other services.** Use **Retry On Fail** on nodes that call APIs, with a pause between tries. Timeouts and rate limits usually pass by themselves.
3. **Decide each node's error behaviour.** n8n can stop the workflow, continue, or route the error to its own output. Choose deliberately: an invoice step should probably stop; a "nice to have" enrichment step can continue.

```flow
trigger | Workflow runs        | Orders, leads, invoices | n8n
tool    | Retry on temporary errors | Pause between tries | n8n
tool    | Error workflow       | Logs item, node and message | PostgreSQL
human   | Owner alerted        | Re-runs the failed items
output  | Nothing lost         | Every failure accounted for | Microsoft Teams
```

> **Watch out:** a retry on a step that creates records can create duplicates. Make the step safe to repeat first (check 5), then add retries.

## Visibility

4. **Save failed executions.** Check the workflow settings so failed production runs are kept. When something breaks, you want to open the run and see the exact data and error, not guess.

## Data safety

5. **Make steps safe to re-run.** Before writing a record, check whether it already exists (by order number or email, for example). Then re-running a failed batch never creates duplicates.
6. **Test with real-shaped data.** Empty fields, accented names, very long text, unexpected formats. Most production failures come from data nobody tested with.

## Security

7. **Use production credentials with the least access needed.** A workflow that only reads invoices should not use an account that can delete them. Keep credentials in n8n's credential store, never pasted into nodes.

## Performance

8. **Check volume and timeouts.** Run the workflow with a realistic batch size. Large lists may need splitting into smaller batches, and long-running workflows need a sensible timeout so they don't hang.

## Ownership

9. **Document the workflow.** One short note: what it does, what triggers it, what it touches, and how to re-run it. Sticky notes inside the workflow work well.
10. **Name an owner.** Every production workflow needs one person who receives its alerts and decides what happens next. A failure alert that goes to everyone gets read by no one.

> **Tip:** run this checklist again whenever a workflow changes significantly, not just at first launch.

## Summary

- **Failures must announce themselves:** an error workflow, retries and saved executions.
- **Re-runs must be safe:** check before you write, and test with messy data.
- **Access must be minimal**, and volume tested before launch.
- **Someone must own it**, with a short note on how it works.

If you'd like this set up across all your workflows, our [reliable n8n automations use case](/use-cases/reliable-n8n-automations/) shows how a typical build works.

## FAQ

### What is an error workflow in n8n?

A separate workflow that starts with the Error Trigger node and runs whenever a workflow that points to it fails. You set it in the failing workflow's settings, and use it to send alerts and log the failure.

### Should every n8n node retry on failure?

No. Retries suit temporary problems such as timeouts and rate limits on calls to other services. Retrying a step that writes data can create duplicates unless the step is safe to repeat.

### How do I stop failed executions disappearing?

In the workflow settings, make sure failed production executions are saved, so you can open a failed run and see the data and the error. Keep successful runs only as long as you need them.
