The 5-minute automation ROI test (and when to walk away)

Every week someone asks whether a task is "worth automating." The answer takes five minutes and three numbers — and roughly a third of the time, the right answer is no.

The formula

monthly hours   = hours per run × runs per month
monthly saving  = monthly hours × loaded hourly cost
payback (months)= build cost / monthly saving

Use the loaded hourly cost (salary + benefits + overhead), not the wage. For most small teams that's $40–80/hour, or ¥100–200/hour in China.

Then apply the only threshold that matters:

| Payback | Verdict | |---------|---------| | < 3 months | Build the smallest version now | | 3–6 months | Worth it if the rules are stable | | 6–12 months | Shrink the scope first | | > 12 months | Don't build it (yet) |

And a sanity filter before the math: the task must be repetitive, rule-based, and costly when it goes wrong. Miss any of the three and the formula lies.

Worked example 1: lead intake for a local service business

A clinic gets 20 form leads a month. Each takes ~10 minutes to copy into the CRM, and leads sometimes sit until the next morning.

Manual time: 20 × 10 min = ~3.3 hours/month
Loaded cost: $45/hour → $150/month
Build: $490 fixed
Payback: 3.3 months

On the numbers alone it's borderline. But add the thing the math can't see: a single missed booking is $200–$500. If the automation prevents one lost job a quarter, it pays for itself. This is why the conversation should start with "how many leads do you lose?" rather than "how many hours?"

Worked example 2: invoice entry at a bookkeeping firm

Thirty invoices a day, ~3 minutes each, 20 working days.

Manual time: 30 × 3 min × 20 = 30 hours/month
Loaded cost: $35/hour → $1,050/month
Build: $2,200 (extraction + review queue + write-back)
Payback: 2.1 months

But the honest version includes the review queue: the person still spends ~20 minutes a day reviewing low-confidence fields. Net saving is closer to 25 hours/month, and payback moves to ~2.5 months. Still an obvious build. The cost that isn't in the formula — a wrong entry discovered during month-end — is what makes the review step non-negotiable.

Worked example 3: a weekly ops report

Someone exports CSVs from three tools every Monday, reconciles them in a sheet, and pastes a summary into Slack. It takes 4 hours a week.

Manual time: 4 × 4.33 = 17.3 hours/month
Loaded cost: $50/hour → $867/month
Build: $1,200
Payback: 1.4 months

This is the easiest "yes" in the catalogue, which is why reports are a great first project: low risk, fast payback, and the client sees the output every Monday.

The hidden costs people forget

Maintenance. APIs change, tokens expire, pages move. Budget 1–3 hours a month, or buy a care

plan. A pipeline nobody maintains becomes a liability within a quarter.

Tooling. Self-hosted n8n is ~$10–20/month of server; Zapier/Make at volume can cost more than

the labour it saves. Put this in the formula.

Failure cost. If the automation silently drops orders, the "saving" is negative. Error workflows

and alerts are part of the build, not a nice-to-have.

Your own time. A $490 build that takes you three weeks is not a $490 build. Estimate in failure

modes, not nodes: every system written to, every approval step, every pagination edge adds hours.

When to walk away

It happens once a quarter: the per-run saving never recovers the build.
The rules are subjective ("decide which leads are good"): you'd be encoding judgement, and the

exceptions will eat the savings.

The process is about to change: automate the stable version, not the one being re-designed.
Nobody owns the exceptions: a queue no one reviews is worse than a manual process.
Compliance requires a named human signature: automate the prep, keep the signature.

Saying "don't build this" is not a lost sale. It's the reason the same client comes back with the two tasks that are worth it — and refers you to someone who needs the same honesty.

Run it in five minutes

Fill in three numbers — hours per week, loaded hourly cost, build budget — and the calculator returns the payback, the monthly saving and a verdict, including "don't build it":

Automation ROI calculator

If the payback is under six months, the next step is a one-page scope with a written definition of done. If it's over twelve, save your money. Either way, you now have a defensible answer instead of a gut feeling.