If your Zap keeps turning off, start here
Zapier turns a Zap off on its own when two things are true. It errors 95 percent of the time it runs. And it has run more than 20 times in the past 7 days.
So a Zap that keeps switching itself off is usually failing first. To get it back:
- Open Zap history and look at the last few runs. The error is usually right there.
- Fix what it says. Often it's a connection that needs signing in again, or something that changed in the app on the other end.
- Turn the Zap back on with the switch at the top left of the Zap editor.
- Replay the runs that failed, from Zap history. On the Free plan you can replay runs that errored. Runs that were held need a paid plan.
On a Team or Enterprise plan you get warned before it switches off. Zapier emails the account owner and waits 24 hours on Team, 72 on Enterprise.
Why Zapier turns a Zap off
The rule is there so a broken Zap doesn't keep failing forever. It used to be stricter. Zapier's product note from September 2023 says Zaps turned off as soon as they hit that line, and the grace period came with that update.
You can override it. In the Zap editor, open the advanced settings and find Error ratio override. Pick "Keep running if errors occur" and the Zap stays on even if it errors over and over. It's available on Free, Professional, Team and Enterprise.
I wouldn't do that on anything that matters. A Zap that errors all the time is doing the job badly, and staying on just hides it. And with the override on, Zapier says your Zaps never get the grace period.
A Zap can also stop for reasons that aren't errors. Zapier's own list:
- You hit your plan's task limit, so it holds every action until the cycle resets.
- A step uses a premium feature your plan doesn't include.
- Your plan changed or your trial ended, so it holds the runs your new plan doesn't cover.
- On Team or Enterprise, you lost access to an app connection the Zap uses, so it pauses.
- On Enterprise, an admin restricted an app the Zap uses, so those steps are held.
How to get told when a Zap turns off
Start with the emails you already get. By default, Zapier emails your Zapier account address about every error, the moment it happens.
Check two things. That the address is one you read. And that nobody changed the frequency to Never. If somebody else set up your account years ago, those emails may be going to them.
Zapier also has an Alerts page. It lists Zaps with errors, held runs, broken connections, and Zaps that were turned off.
If you want it pushed to you somewhere else, there's a built-in app called Zapier Manager. Its "Zap Turned Off" trigger fires when a Zap you own is paused or turned off. Another fires when a Zap you own hits an error.
One more is worth setting up. If you hit your task limit, every Zap in the account is held, including the ones that would warn you. Zapier Manager has a trigger for when task usage reaches a number you pick. Set it below your limit.
When a Zap isn't triggering at all
All of that is about Zaps that fail.
There's a second kind of stop. The Zap never runs.
The form stops sending. Or the app on the other end quietly stops passing anything along. No run starts, so nothing fails. Every rule above is counting failures.
If the trigger's connection itself breaks, you'll usually hear about it. Zapier shows broken connections on its Alerts page. The silent case is when nothing arrives at all.
That happened to me. It wasn't a Zap. The tool doesn't matter here.
I went looking for my leads and they just weren't in my inbox.
I asked my AI what happened, it said the job wasn't running, and I had it fixed.
It wasn't dramatic, but it took me four days to notice.
It started with a password change. No error. No warning.
Afterward, I wanted to know what setting would have caught it. On November 16, 2026, I searched the help pages for Zapier, Make and n8n for an alert that fires when nothing runs. I found none built in. Tools change, so check yours.
Error alerts catch failures. Only a heartbeat catches silence.
Make and n8n have the same blind spot
The help pages I read for Make and n8n describe alerts the same way. They watch for runs that fail.
Make. In a scenario's settings, under Errors before deactivation, you set how many errors in a row Make accepts before it switches the scenario off. The default is 3. A scenario that starts with an instant trigger, like a webhook, switches off at the first error. So does a scenario that hits a connection validation error.
Make emails you by default when an error stops a scenario from running, and when it switches a scenario off because of errors. You can turn on warnings too, per organization, from your profile's email preferences.
What it can't see: a scenario that's on and simply gets nothing.
n8n. You build an error workflow, starting with the Error Trigger node, and attach it to a workflow in Workflow Settings. It runs if an execution fails, including when the trigger node itself errors.
You can't test it with a manual run. n8n says the Error Trigger only runs when an automatic workflow errors. To test it, put a Stop and Error node in a workflow that's active and running on its own trigger, and let it fail.
n8n's Insights shows production runs and failures for the last 7 days. Every plan gets the summary. The per-workflow dashboard is on paid tiers. Either way it's something you have to open. It doesn't alert you.
What it can't see: a workflow that never starts. And if you host n8n yourself and the server goes down, nothing inside n8n can tell you.
Add a heartbeat
A heartbeat is a small sign that an automation did its job. It lands somewhere you already look. When it doesn't show up on time, that's your alarm.
The established name for this is a dead man's switch. Healthchecks.io describes it plainly: it "keeps silent as long as pings arrive on time" and "raises an alert as soon as a ping does not arrive on time."
Put the heartbeat at the very end of the automation, after the real work. A lead saved. An invoice sent. If it sits at the start, a run that saved nothing still looks healthy.
I'd pick one of two ways to build it.
The sheet. One Google Sheet with a row per automation you care about. Columns for the name, how often it should run, and when it last ran. Each automation gets one more step at the end that updates its own row with the time it ran. Zapier, Make and n8n can all update a Google Sheets row.
Then something has to read it. The simplest version is you, on your weekly look (step 5 below). Anything older than its rhythm is late.
If you'd rather get an email, build a daily check that reads every row and emails you every morning, late or not. If it only emails when something's late, a check that died looks exactly like "all fine." When the morning email doesn't come, that's your alarm too.
- In n8n: a Schedule Trigger on a Days interval, the Google Sheets node's Get Row(s) with no filter so every row comes back, and an If node using Date & Time "is before" to mark the late ones.
- In Make: a scenario scheduled Daily that reads the rows, with a filter using Make's date and time operators.
- In Zapier: this one takes more work. A lookup step returns only the first row that matches, and a filter compares against a date you give it. So you'd need one small check Zap per automation, with a Formatter step to work out the cutoff. It also needs a paid plan, because the Free plan allows one trigger and one action.
The outside service. A dead man's switch service does the same job from outside your tools. Each automation pings it at the end of a run. The service expects the ping on a schedule and emails you when one is late. If you host n8n yourself, this is the one I'd use, because it still works when your server is down.
I haven't built these in a live account for this article. Every piece comes from the tools' own help pages, and the Zapier version has more moving parts than the other two. Whichever you build, test it the way step 6 says.
Low-volume automations need a longer window. A form that fires twice a week will look late most days. Set the check to the real rhythm or it cries wolf, and then you stop reading it.
Seven steps to check if your automations are running
I turned this into a framework before I made the video: list, ask, alert, heartbeat, owner, break, retire. Steps 1 to 3 fit in about an hour if you run ten or so automations. That's my estimate, not a measurement.
List every automation
Open each tool. Don't work from memory.
Zapier, Make, n8n, your CRM's own workflows, email rules and auto-replies, form handlers, booking tools, anything a contractor set up. One row each: what starts it, what it should produce, where that lands, how often, and which account it runs under.
When to do it: one sitting, this week.
Ask who would notice
If this stopped today, who would notice, and when?
Write the honest answer. A customer in an hour. Me at month-end. My accountant at tax time. Nobody. Then a rough cost per day of silence. Rank by days-until-noticed times cost per day.
When to do it: right after the list.
Turn on the error alerts
And write down what they can't see.
Check that each tool's error emails go to an address someone reads. Beside each row, write the blind spot: a trigger that never fires sends no error.
When to do it: same sitting.
Give the top rows a heartbeat
One place for all of them.
The sheet or a monitor service. Each heartbeat says how often it should arrive and what counts as late.
When to do it: start with the top two rows.
One owner, one weekly look
Pick a day and five minutes.
In a one-person business the owner is you. Write it down anyway. The list is what someone else reads if you're away.
When to do it: put it on the calendar.
Break each alarm once on purpose
An alarm that's never fired isn't an alarm yet. It's hope.
Switch the automation off, or send it a bad test record, and see if the alert and the missing heartbeat both reach you. Where breaking it is unsafe, test a copy.
When to do it: after setup, and again after any change.
Retire what nobody needs
Pause first. Delete later.
If nobody can name what a row is for, pause it and wait one full cycle. If nothing complains, delete it. Give everything you keep a review date.
When to do it: quarterly.
This is the list as a table you can copy into a sheet.
| Automation | Who notices, and when | Alert goes to | Heartbeat | Owner | Review date |
|---|---|---|---|---|---|
| Contact form to CRM | Nobody for days | me@ (read daily) | Sheet row, daily | Me | Feb 1 |
| Invoice reminders | Me at month-end | me@ | Sheet row, weekly | Me | Feb 1 |
| Review request after a job | Nobody | None yet | Not worth one | Me | Retire? |
Where this is overkill
Two cases where I'd skip most of it.
- A customer sees the result within minutes. They're your heartbeat. That works if you're fine hearing it from a customer.
- You have one automation and you see its output every day. Skip the list. Just check its error emails go somewhere you read.
Do this one thing this week
Take the automation at the top of your list. Turn it off for an hour and see if anything tells you.
If nothing does, you just found your silent stop before a customer did.
Questions people ask
Why does my Zap keep turning off? Zapier turns a Zap off when it errors 95 percent of the time and has run more than 20 times in 7 days. Team and Enterprise accounts get an email and a 24 or 72 hour grace period first. Fix the error Zap history shows, then turn the Zap back on.
How do I get notified when a Zap is turned off? Zapier lists turned-off Zaps on its Alerts page. To have it sent to you, use Zapier Manager's "Zap Turned Off" trigger in a small Zap that messages you.
Why is my Zap not triggering? Zapier's own list: the Zap isn't turned on, the items existed before you turned it on, there's a connection problem with the trigger app, or the wrong trigger event is selected. If none of those fit and nothing shows in Zap history, the app on the other end may have stopped sending. That's the case a heartbeat catches.
Does Make tell you when a scenario is deactivated? Yes. By default Make emails you when it switches a scenario off because of errors, and when an error stops a run.
Can n8n alert me when a workflow doesn't run? Its error workflow runs when an execution fails, including a trigger error. I found nothing built in for a workflow that simply never starts, so use a heartbeat.
What is a heartbeat, or dead man's switch? A signal an automation sends when it finishes. A separate check expects it on a schedule and alerts you when it doesn't arrive.
Watch the video
I made a video on the same seven steps, How to Tell When an Automation Stops Running. It's 4 minutes 42 seconds, and it's the fastest way to see the whole thing once.
Build the system behind your own automations
How to Framework walks through writing your own rules, one domain at a time.
Explore How to FrameworkSources
- Zapier Help, Zap is not running, updated November 11, 2026, read November 16, 2026
- Zapier Help, More time to fix Zap workflows that error (product update, September 2023), read November 16, 2026
- Zapier Help, View and manage your Zap history, read November 16, 2026
- Zapier, Zapier Manager triggers, read November 16, 2026
- Zapier Help, Manage notifications when errors occur, Alerts page, Zap isn't triggering, Replay Zap runs and the Free plan, read November 16, 2026
- Zapier Help, Filter and path rules in Zaps and Schedule Zaps, read November 16, 2026
- Make Help Center, Scenario settings, Overview of error handling, Manage your email preferences, Schedule a scenario and Filtering, read November 16, 2026
- n8n Docs, Error handling, Error Trigger, Insights, Schedule Trigger, Google Sheets and If, read November 16, 2026
- Healthchecks.io, Documentation, read November 16, 2026
- My own record of the outage: four days from a password change to finding the missing lead emails
