ROI

Justifying Automation with ROI That Survives Finance

Most automation proposals die in finance because the savings are soft and the costs are underscoped. Build the ROI case with numbers that survive an audit.

Automation projects do not fail in engineering; they fail in the ROI math, twice. First when an inflated case gets approved and misses, poisoning the well for the next five proposals. Second when a good project dies because nobody built the case properly. The core numbers are simple: payback equals investment divided by annual savings, and ROI equals annual savings divided by investment. A $250,000 robot cell saving $105,000 a year pays back in 2.4 years and returns 42 percent annually. Most plants approve under 2 to 3 year payback; under 18 months usually sails through.

Build savings from the hard categories first. Labor: the cell displaces 1.5 operators at $62,000 loaded, $93,000 a year, but only if those hours are actually redeployed or not backfilled, so name where they go. Scrap: cutting defect rate from 3 percent to 1 percent on 400,000 units at $1.50 accumulated cost saves $12,000. Throughput counts only when the machine is the constraint and the extra units will sell; then value them at contribution margin, not revenue. The Manufacturing ROI calculator turns investment and savings into payback, ROI, and annual return in seconds; the credibility lives in the inputs.

Scope the full investment or the payback is fiction. The robot quote is $250,000, but integration typically adds 20 to 40 percent: guarding, end-of-arm tooling at $15,000 to $40,000, conveyance changes, programming, spare parts, and 2 to 6 weeks of debug at reduced output. A realistic all-in number for a $250,000 quote is often $310,000 to $350,000, which stretches that 2.4 year payback to 3.0 or beyond. Present the all-in number yourself; if finance finds the gap instead, every future number you bring gets a haircut.

Subtract the new costs the machine creates. A robot cell carries $8,000 to $15,000 a year in maintenance and spares, consumes power, and needs a technician skill set that may mean training or a wage premium. Net the $105,000 gross saving down to perhaps $92,000 and say so out loud. Cases that show their own deductions get approved faster because the reviewer stops hunting for what you hid. As a rule, a case that only works if every assumption breaks your way is not a case, it is a wish.

Utilization is the assumption that kills the most projects after installation. Savings scale with the hours the cell actually runs: a case built on 4,000 hours a year delivers half its promise if demand only loads it 2,000. Stress-test at 70 percent of planned volume; if payback still lands under 3.5 years, the project is durable. And beware the fractional-head trap: saving 0.4 of an operator on each of three lines saves zero payroll unless the work is consolidated so an actual position is redeployed. Count heads that move, not percentages on a slide.

Sequence proposals by return, not by excitement. Typical hierarchy in most plants: setup reduction and scheduling fixes return 100 to 300 percent with near-zero capital; targeted fixtures and error-proofing return 60 to 150 percent; single robot cells return 25 to 50 percent; full line automation returns 15 to 30 percent with the longest debug. Plants that harvest the cheap tiers first fund the expensive tiers out of the savings and walk into finance with a track record instead of a brochure.

Close the loop on a cadence or the next case pays for it. At 90 days post-launch: audit actual run hours, defect rates, and crewing against the proposal and publish the variance, good or bad. At 12 months: report realized annual savings next to the promise; plants that do this consistently run 85 to 95 percent realization, while plants that never look back average closer to 60 percent. World-class capital discipline is a rolling ranked list of opportunities with pre-built ROI cases, post-audits on every project over $100,000, and a finance team that trusts operations math because it has been checked twenty times and held.

Published 2026-07-02.