Every week, someone asks us to automate a process that isn't worth automating. Not because automation is a bad idea — but because the process they've picked costs $3,000 a year and the automation would cost $12,000 to build and $1,500 a year to maintain. That's not ROI. That's a vanity project with a four-year payback.

The good news: calculating the real ROI of an automation project takes about two hours and a spreadsheet. Here's exactly how we do it before we ever write a line of code.

Step 1: Map the real cost of the process today

Most businesses wildly underestimate what a manual process actually costs. They know the headcount. They rarely account for everything else.

For any process you're considering automating, total up the following:

  • Direct labor cost — hours per week × fully loaded hourly rate (salary + benefits + overhead). Don't use base salary. Use 1.3x to 1.5x to capture the real cost.
  • Error rate and correction cost — what percentage of outputs need rework? How long does rework take? What does downstream damage from errors cost (customer churn, write-offs, delays)?
  • Delay cost — how much does lag time cost you? In lead routing, a 5-minute delay vs. a 5-hour delay in responding to an inbound lead can be the difference between a closed deal and a ghost.
  • Scalability cost — if volume doubles, do you need to hire another person? What does that cost?
  • Opportunity cost — what would the people doing this work produce if they weren't doing it?

Rule of thumb: the real annualized cost of a manual process is typically 2x to 3x the direct labor cost once you account for errors, delays, and scalability overhead. A process that looks like a $40,000/year problem is often a $90,000/year problem.

Step 2: Define what "automated" actually means

This is where most ROI calculations fall apart. People assume automation means a straight line — the work just disappears. In reality, most automations are partial. You're eliminating 70% of the manual work, not 100%. The remaining 30% either can't be automated (requires judgment) or isn't worth automating (too infrequent, too variable).

Be precise about what automation replaces:

  • Which specific steps become fully automated
  • Which steps become partially automated (human approves, machine does the work)
  • Which steps remain fully manual
  • What new work the automation creates (monitoring, exception handling, maintenance)

A well-scoped automation typically eliminates 60–85% of manual effort, creates 5–15% new work in oversight and exception handling, and leaves 15–30% of the original work untouched because it genuinely needs human judgment.

Step 3: Estimate the automation cost honestly

Two costs to include: build cost and ongoing cost.

Build cost

This varies enormously based on complexity. A simple single-system automation (one trigger, one action, same platform) can be built in a day or two. A multi-system automation with conditional logic, error handling, monitoring, and integration into three different platforms takes weeks. Ballpark ranges: simple $2,000–5,000, mid-complexity $8,000–20,000, complex $25,000+.

Ongoing cost

Don't skip this. Every automation has running costs: API fees, cloud infrastructure, licensing for automation tools, and maintenance labor. Budget 10–20% of build cost per year for maintenance on a well-built automation. Budget 30–50% if it's tightly coupled to systems that change frequently.

Step 4: Calculate the payback period

Net annual savings = (Current process cost × elimination percentage) − new oversight work cost − ongoing automation cost.

Payback period = Build cost / Net annual savings.

A payback period under 12 months is strong. 12–24 months is acceptable for larger, more complex automations. Over 24 months means you're either automating the wrong process, the scope is too large, or the process cost is lower than you thought.

Real example: a logistics client was spending $180,000/year in staff time manually routing shipments across carriers. We automated 80% of routing decisions in a 6-week build at $22,000. Net annual savings: $144,000 minus $15,000 for the remaining oversight work minus $4,000/year in running costs = $125,000/year. Payback: 2.1 months.

Step 5: Prioritize using effort/impact scoring

You probably have five processes that could be automated. They don't all deserve equal attention. Score each one on two dimensions:

  • Impact score (1–10) — based on annual cost savings relative to build cost, plus strategic value (unlocks growth, reduces key-person dependency, enables scale)
  • Effort score (1–10, lower is better) — based on data availability, system access, process stability, and exception frequency

Divide impact by effort. The highest scores go first. Build your automation roadmap in that order.

The honest shortcut

If you're not ready to run this full analysis yourself, the shortcut is: automate the highest-volume, lowest-judgment process first. High volume means the ROI math almost always works. Low judgment means the automation can handle it without constant human review.

Invoice processing, lead scoring, order status updates, report generation, candidate screening — these are all high-volume, lower-judgment processes that pay back fast. Custom pricing decisions, complex customer negotiations, and quality judgment calls — not ready for full automation yet.

Pick the right process first, and the ROI case writes itself.