Step 1: hours returned
Count how often the task runs, how long it takes manually, and how long a human still spends reviewing the automated output. Review time is the term most ROI estimates forget.
Hours returned = runs × (manual minutes − review minutes) ÷ 60Worked example (illustrative).
Illustrative: a report task runs 20×/month, takes 45 minutes manually, needs 5 minutes of review after automation. Hours returned = 20 × (45 − 5) ÷ 60 = 13.33 hours/month.
Step 2: modeled monthly value
Multiply hours returned by an hourly value — then subtract the automation's recurring cost (subscriptions, compute, maintenance time). The hourly value is an assumption, not money received: freed hours only become cash if you cut spending or redeploy the time to revenue work.
Modeled monthly value = hours returned × hourly value − recurring costStep 3: payback period
Divide the one-time setup cost by the modeled monthly value. That is the simple payback — how many months until the automation has 'paid for itself' on paper.
Payback (months) = setup cost ÷ modeled monthly valueWorked example (illustrative).
Illustrative: $400 setup ÷ $350 modeled monthly value = 1.14 months payback. Treat this as a planning figure — exceptions, outages, and review can erase it.
What the model excludes
- Revenue growth — do not add speculative sales to a labor estimate.
- Error costs — what happens when the automation is wrong, and how often.
- Maintenance drift — APIs change, formats change, edge cases accumulate.
- Taxes and financing effects.
Assumptions and limits
- Hourly value is an assumption, not automatically received money.
- The model is a planning tool; exceptions and outages can erase the estimated benefit.
- Examples are illustrative arithmetic.
Common questions
Is time saved the same as money saved?
No. Time returned only becomes a cash saving if spending actually falls — fewer hours paid, or the freed time redeployed to work that earns. Otherwise it is capacity, which is valuable but not cash. The model also excludes revenue growth, so speculative sales are never added to a labor estimate.
What is a good payback period for automation?
Shorter is better, but the honest answer depends on risk: a 2-month payback on a fragile automation can be worse than a 6-month payback on a robust one. Always price review, maintenance, and failure modes alongside the payback. The worked example divides $400 setup by $350 modeled monthly value for 1.14 months — a planning figure, not a promise.
Should review time count against automation ROI?
Yes — always. Every automation needs human review somewhere, and review minutes come straight off the hours returned. Automations with heavy review needs can have negative ROI despite impressive demo numbers. The hours-returned formula subtracts them directly, as in the worked example, where a run takes 45 minutes manually and 5 minutes of review.
Published · Updated . Automation planning guides for founders and developers. Models are planning tools, not guarantees.