Automation vendors love the word "transformative." We prefer arithmetic. Before we write a line of code for a client, we run one calculation — and if the number doesn't clear the bar, we tell them not to build. This is that calculation, in full, so you can run it yourself.
The formula
Year-one return has three inputs on the gain side and three on the cost side:
- Hours recovered — hours per week the workflow eliminates × the loaded hourly rate of the person doing it × 52.
- Errors eliminated — the annual cost of the mistakes manual handling produces: re-work, refunds, churned customers, compliance exposure.
- Speed converted to revenue — what responding in minutes instead of hours is worth in won deals and kept customers.
- Build cost — the one-time project fee, whatever you pay a partner or your own team.
- Tooling — platform subscriptions, API usage, hosting. Typically tens to a few hundred dollars a month, not thousands.
- Maintenance — budget roughly 10–15% of build cost per year for API changes and edge cases. Anyone who quotes zero is selling something.
One definition matters before the arithmetic: use the loaded hourly rate, not salary. A $55,000 coordinator costs the business roughly $70,000 once payroll taxes, benefits, software seats, and overhead are counted — about $34–38 an hour, not $26. Undercount the rate and every automation looks worse than it is; it's the most common error in build-versus-don't decisions.
A worked example
An operations coordinator spends 11 hours a week copying data between a CRM, a billing tool, and a spreadsheet. At a loaded rate of $38/hour, that's $21,736 a year spent on copy-paste — before counting a single typo. A system integration that syncs those tools eliminates the task outright, and the coordinator's 11 hours go back into work that actually needs judgment.
Against a fixed one-time build and a modest tool subscription, that single workflow typically pays for itself inside the first one to two months. Everything after payback is margin — which is why the sequencing question ("what do we automate first?") matters more than the tooling question.
| Year-one ledger | Manual | Automated |
|---|---|---|
| Hours on the task | 572 hrs (11/wk × 52) | ~26 hrs (exception handling) |
| Labor cost at $38/hr loaded | $21,736 | ~$988 |
| Data-entry defects | ~250 (0.5% of 50k records) | Near zero |
| Sync delay | Up to 3 days | Real time |
Where ROI hides beyond the timesheet
Hours are the visible return. Two quieter ones routinely dominate the math:
Speed-to-lead. A lead answered within minutes is dramatically more likely to become a conversation than one answered the next morning; every hour of delay costs winnable revenue that never shows up in a cost report. When Fabrioza cut custom-manufacturing quotes from days to under 24 hours, the win wasn't saved admin time — it was the orders that stopped going to whoever quoted first.
Error elimination. A human keying 500 records a week at even a 1% slip rate produces five defects a week, and a wrong invoice or a mis-filed dispute costs far more to unwind than to prevent. Machines don't get bored on record 401.
The payback curve, honestly
Year one is not a straight line. Months one and two carry the build cost and an adoption dip — people double-check the machine before they trust it, and edge cases surface that discovery missed. The steady state arrives around month three, and from there the return runs quietly every week. The silent killer is not the technology; it's adoption. A workflow the team routes around returns nothing, which is why we track usage after every deployment and treat "the team stopped using it" as a defect, not a client problem.
The second-order effect is where automation compounds: workflow two reuses the integrations, data model, and monitoring built for workflow one, so it costs less and pays back faster. Teams that sequence well end year one with three or four workflows running on infrastructure they paid for once.
What doesn't pay back in year one
- Automating a broken process. Encoding a bad workflow just produces bad outcomes faster. Fix the process, then automate it.
- Edge-case-heavy work. If every third instance needs human judgment, you'll build an expensive exception queue, not an automation.
- Vanity automations. A workflow that saves 20 minutes a month never repays serious build effort. Volume × frequency is the filter.
Run the math before you build
The pattern across our deployments is consistent: the first well-chosen workflow clears payback in months, and the second one is cheaper because it inherits the infrastructure of the first. The discipline is choosing by arithmetic instead of excitement — highest hours × rate × error cost first, everything else after.
Our free audit runs exactly this formula against your answers to seven questions and returns the workflow with your fastest payback. If the math says "don't build yet," it will tell you that too.