← Journal
A Simple Framework for Calculating Automation ROI
Stop guessing whether an automation project will pay back. A four-variable model, a back-of-envelope calculation, and the mistakes to avoid.
Every automation pitch deck has an ROI slide. Almost none of them are honest. Here is the calculation we actually use before we take a project on.
The four variables
- H: hours per week currently spent on the task
- W: fully-loaded hourly cost of the person doing it
- C: one-time build cost
- M: ongoing monthly cost (infra, APIs, maintenance)
Annual return is straightforward:
annual_savings = (H * 52 * W) - (M * 12)
payback_months = C / (annual_savings / 12)
If payback is under nine months, it's a strong candidate. Between nine and eighteen, it depends on strategic value. Over eighteen, walk away or rescope.
The mistakes that kill the model
- Counting time, not money. "This saves 10 hours a week" is meaningless if those 10 hours go to more Slack, not more revenue.
- Forgetting exception handling. The happy path is 30% of the work. The other 70% is edge cases, and edge cases have a price.
- Ignoring depreciation. APIs change. Models deprecate. Budget 20% of build cost annually just to keep the thing alive.
A worked example
A compliance review process: 12 hours/week at $85/hour fully loaded, $40K build, $600/month to run.
annual_savings = (12 * 52 * 85) - (600 * 12) = $45,840
payback_months = 40000 / (45840 / 12) ≈ 10.5 months
Strong candidate. And after year one, it's pure margin.
If you have a workflow in mind and want an honest read, we'll run the numbers before you commit a cent.
Next step