Automation Audit Checklist for Rules That Quietly Fail
The automation that worries me is not the one that crashes with a red error. It is the one that stopped working in March, sent nothing, warned no one, and was only noticed in June when a customer asked why their invoice never arrived.

The automation that worries me is not the one that crashes with a red error. It is the one that stopped working in March, sent nothing, warned no one, and was only noticed in June when a customer asked why their invoice never arrived. An automation audit checklist is the cheapest fix for that problem: a short, repeatable review that catches silent failures before they cost you a client or a weekend.
I run the one below every quarter on my own setup, which is about 40 rules spread across Zapier, email filters, iPhone Shortcuts and a couple of chatbot flows. It takes roughly 90 minutes. Here is the full process.
Why automations fail without telling you
Most automations are built when everything is fresh. The form has the fields you expect, the spreadsheet has the column names you typed, the account has a valid password. Over months, small things drift:
- A form field gets renamed from "Phone" to "Mobile number" and the mapping goes blank.
- An app connection token expires after 60 or 90 days and the trigger stops firing.
- A teammate leaves and the rule ran under their account.
- A filter condition like "subject contains Invoice" misses the new template that says "Your bill".
- A free plan quietly hits its monthly task limit on the 22nd.
None of those produce a dramatic failure. The workflow runs, finds nothing to do, and reports success. That is why waiting for error emails is not a strategy.
Step one: build a one-page inventory
You cannot audit what you cannot list. Open a plain spreadsheet with these columns and fill one row per automation:
| Column | What to write | Example |
|---|---|---|
| Name | Plain description | New web lead to CRM and Slack |
| Tool | Where it lives | Zapier |
| Owner account | Whose login runs it | office@ shared account |
| Trigger | What starts it | Website contact form submitted |
| Expected volume | Normal runs per month | 30 to 50 |
| Last checked | Date of your last audit | Mid August |
| If it fails | Who notices and how fast | Lost lead, maybe never noticed |
The last column is the one that matters. Sort by it. Anything where the honest answer is "nobody would notice" goes to the top of the audit.
If you keep files and prompts tidy already, this step is easier. The same naming discipline from naming files so automation rules can find them works for naming automations: verb first, then the system, then the outcome.
The automation audit checklist itself
Go row by row and answer these seven questions. Each one targets a specific way rules fail.
- Did it run as often as expected? Compare the last 90 days of run history with your expected volume. A lead form that normally fires 40 times a month and fired 3 times is broken, even if all 3 succeeded.
- Is every connection still authorised? Open the connected accounts page in each tool. Reconnect anything with a warning icon or an expiry date within the next month.
- Does it run under the right account? Move anything owned by a personal or departed login to a shared service account.
- Do the field mappings still match? Submit a real test entry and follow it all the way through. Look for blank fields at the far end, not just a green tick.
- Are the filters still true? Re-read every condition. Subject lines, sender addresses and tag names change more than people expect.
- Are you near a plan limit? Check task or run counts against your plan. If you used 85 percent last month, you will run out in a busy one.
- Is it still needed? Delete anything that serves a process you no longer run. Fewer rules means fewer surprises.
Write the date and a one word result, like ok, fixed or deleted, in the Last checked column. That record is what makes the next audit quick.
Add a heartbeat for the critical few
Quarterly checks catch slow drift, but some automations are too important to leave for three months. For your top three or four, add a heartbeat: a simple signal that proves the rule is alive.
The easiest version is a weekly count. A scheduled step every Monday at 09:00 counts how many times the automation ran in the past seven days and sends you one line, such as "Lead intake: 11 runs this week". If the number is zero or far below normal, you know immediately. Most tools can do this with a schedule trigger and a spreadsheet lookup. For webhook based flows, add a step that writes one row to a log sheet each time the hook fires, then count those rows.
Why I ignore the advice to automate the audit too
A popular suggestion is to replace manual audits with a monitoring tool that watches every automation and alerts on failures. Those tools are useful, but I do not think they replace the human review, for a simple reason: they only see errors. The failures that hurt most are runs that succeed while doing the wrong thing. A monitor will not tell you that your lead now lands in the CRM with an empty phone number, or that a filter is routing half your invoices into the wrong folder.
There is also the cost. Dedicated monitoring often runs 20 to 50 dollars a month, which is more than many small teams pay for the automation tool itself. A spreadsheet and 90 minutes a quarter is cheaper, and it forces someone to look at what the rules actually produce. That judgement step is the same argument we make in why the human in the loop still matters. Use alerts for crashes, and use your own eyes for correctness.
A realistic quarterly schedule
Put a 90 minute block in your calendar for the first Tuesday of January, April, July and October. Spend the first 20 minutes updating the inventory, 50 minutes on the checklist for your highest risk rows, and the last 20 on fixes and deletions. If you have more than 50 automations, split the list in two and audit half each time, starting with the "nobody would notice" group.
After two or three rounds, the audit gets faster, because most rows come back ok and you know where the fragile ones live.
Frequently asked questions
How often should I audit my automations?
Every three months suits most small teams and solo workers. Critical rules, such as lead capture or payment notices, deserve a weekly heartbeat check on top of that. If you change tools or forms often, audit after each change as well.
What is the most common reason an automation stops working?
Expired or disconnected app connections are the most frequent cause, followed by renamed fields in forms or spreadsheets. Both make the rule run without doing anything useful, which is why they go unnoticed for weeks.
Should automations run under a personal account?
It is better to use a shared service account owned by the business. When a person leaves or changes their password, rules tied to their login break or keep running somewhere nobody can see. A shared account keeps ownership clear and access easy to hand over.
More in Tips and Habits
Tips and Habits
Document Templates That Stop You Rebuilding the Same File
Last March I sent a client a project update with another client's name in the footer.
Tips and Habits
Voice Dictation Habits That Make First Drafts Faster
I keep a rough log of how long writing tasks take me, and the entry that bothered me for most of last year was a weekly client...
Tips and Habits
Saved Replies That Answer Common Customer Emails in Seconds
You answer the same customer emails every week: where is my order, can I change my delivery address, do you ship abroad, how do...