Free worksheet
The Automation Decision Sheet
Work out which of your repetitive tasks are worth automating — and which to leave alone.
Read this first
Most guides like this assume the answer is yes. This one does not. We build and run these automations ourselves and publish what they cost, and on our own numbers most small repetitive tasks are not worth automating. This sheet is designed to reach that answer honestly when it is the right one.
— or just work through it here.
Step 1 — Inventory
List only the tasks where information moves between two systems. Thinking, deciding and talking to people are not automation candidates.
| Task | How often | Minutes each | Who does it | What breaks if missed |
|---|---|---|---|---|
Step 2 — Score each task
A. Time cost per month
Times per month × minutes each.
B. Error cost
What one mistake costs — a lost lead, a wrong address, a re-shipped parcel, an awkward apology. Score 1 (mildly annoying) to 5 (costs real money or a customer). This is the factor people underweight.
C. Setup cost
Honestly: a simple two-step automation is an afternoon including testing. A multi-step one with error handling is two or three days for someone who has not done it before.
D. Running cost
Not just the subscription. We measured this: a workflow that processes a list costs roughly one unit of work per item, plus two. A 10-item order cost 12 units where a 1-item order cost 3 — four times as much, and your customers decide which you get, not you.
Retries cost too: recovering 20 failed calls consumed 14 extra units in our test.
E. Maintenance and risk
Every automation is a thing that can break, usually silently. In our tests, a workflow whose alert step sat after the failing step produced 0 notifications when it died. Budget real attention for this, or do not build it.
F. Privacy and risk check
Does this task touch customer personal data — names, emails, addresses? If so, that data will pass through, and often be stored in, a third party’s logs. Check their retention setting before you connect anything real, and never test against live customer records.
Step 3 — Decide
| If… | Then |
|---|---|
| Under 30 min/month and error cost 1–2 | Leave it alone. You will spend longer building and maintaining it than doing it |
| A native free option already exists | Use that. Google Forms writes to Google Sheets with no tool. Looker Studio does reporting free. Check before you buy |
| Over 2 hours/month and error cost 3+ | Strong candidate. Build it — with an error handler |
| It touches customer personal data | Not yet. Test with fake records first |
| You cannot say who gets told when it breaks | Not yet. Answer that first |
| It needs a judgement call | Not automatable. Automate collecting the information the judgement needs instead |
Step 4 — Before it touches anything real
- Tested end to end with fake records
- Tested what happens when a required field is empty — in our test an empty submission was accepted silently, every field stored as null, and no error was raised
- Tested the same input twice — without a duplicate check, two identical submissions produced 2 rows instead of 1
- Error handler is a separate workflow, not a step inside this one
- A named human gets told when it fails
- You know the monthly cost at realistic volume, not best case
- You know what a half-finished run leaves behind — ours left 4 of 10 rows written, with nothing marking the order incomplete
Step 5 — The page to forward
Automation proposal
Task: ______________________________
Currently costs: ______ min/month · Error risk: ___ /5
Proposal: ______________________________
Setup: ______ hours · Running cost: ______ /month at ______ volume
What breaks if we do not: ______________________________
Who gets alerted on failure: ______________________________
Recommendation: Build / Leave alone / Use the free option
Every figure in this sheet comes from workflows we built and ran ourselves against test data. Full write-ups and the limits of what we measured are on the workflow pages, and our testing method is here. There are no affiliate links in this document.