Keeping automations from breaking silently: monitoring and ownership

The worst thing about a broken automation is that nothing happens. No error appears on anyone's screen. Leads stop arriving in the CRM, orders stop reaching accounting, or welcome emails stop sending, and nobody notices for days or weeks. This guide covers the simple habits that make sure you find out quickly.

Why automations break

Automations depend on several systems staying the same. Common causes of failure:

Expired connections. A password change, a removed user or an expired login token disconnects an app.
Renamed or deleted fields. Someone edits a form or a CRM property and the mapping breaks.
App changes. A vendor updates its API or changes how a connector works.
Plan limits. You hit a usage limit and runs stop until the next billing cycle.
Staff leaving. The automation was built under their account, which gets deactivated. See Offboarding a staff member.
Bad data. An unexpected value, such as a blank required field, makes a step fail.

Build an automation register

The foundation of everything else is knowing what you have. Keep a shared list with:

Column	Example
Name	Website enquiries to CRM
What it does	Creates CRM contact and follow-up task for each contact form entry
Where it runs	Zapier, CRM native integration, custom code
Apps involved	Website form, CRM, project tool
Owner	Named person
Account it runs under	A shared business account, not a personal one
Criticality	High, medium, low
How failures are reported	Email alert to owner and ops inbox
Last checked	Date

The asset inventory tracker works well for this, or add a tab to your tool audit.

Give every automation an owner

The owner does not have to have built it. They are responsible for knowing it exists, receiving its alerts, and making sure it is fixed when it breaks. When an owner leaves or changes role, reassign their automations as part of the handover.

Turn on failure alerts

Zapier, Make and most native integrations can notify you when a run fails. Check:

Alerts are switched on for every automation that matters.
They go to at least two people, or a shared inbox or chat channel.
Someone acts on them. Alerts that are ignored train people to ignore alerts.

Watch for "nothing happened"

Failure alerts only catch runs that start and then fail. They miss the case where the trigger never fires. For critical flows, add a check that the expected result is still happening.

Compare counts. Each week, compare form entries on the website with new contacts in the CRM, or orders in the store with sales in accounting.
Use a daily heartbeat. Some teams set a scheduled test that runs through the automation and alerts if the expected result does not appear.
Watch for silence. If a flow that normally runs every day has not run for two days, that is worth a look. Some monitoring tools can alert on this.

Test after every change

Most breakages follow a change somewhere else. Make it a habit to re-test related automations after:

Editing a website form or CRM field
Changing a password or user on a connected app
Switching plans or adding a new tool
A website update or redesign. See Form testing.

A simple routine

0 of 4 done

Weekly (10 minutes). Check the run history of high-criticality automations for errors. Compare one or two key counts.
Monthly. Submit test entries through key forms and confirm they arrive everywhere they should.
Quarterly. Review the register. Remove automations nobody needs, confirm owners, check accounts and connections still belong to current staff.
On every staff departure. List automations and connections under that person's accounts and move them before deactivating anything.

Saved in this browser only.

When something breaks

Stop the damage. Pause the automation if it is creating bad data.
Find the gap. Use the run history to see when it started failing and what was missed.
Fix the cause, not just the symptom. Reconnect the account, correct the mapping.
Replay or catch up missed items. Many platforms let you re-run failed runs. Otherwise, process them by hand.
Note what happened in the register, and add a check if it could happen again.

Common mistakes

Alerts to one person's inbox, which nobody reads during their holiday.
Automations under personal accounts.
No register, so nobody knows what to check.
Ignoring warnings about approaching usage limits.