The ROI That Convinced Everyone

Why technology investment calculations are almost always wrong

The ROI calculation is the centerpiece of every equipment justification. Labor savings, quality improvements, throughput gains, all converted to currency, projected over years, compared against the investment. The number looks good. The number always looks good, because if it didn't look good, nobody would be presenting it.

This isn't fraud. It's structural optimism embedded in how these calculations get built. Every assumption tilts toward justification. Every uncertainty resolves in favor of the project. The result is a document designed to secure approval, not to predict outcomes.

Start with the labor savings. The calculation assumes the eliminated labor actually gets eliminated — that headcount decreases or gets redeployed to productive work. In practice, automation rarely reduces headcount. It shifts work. The operator who used to do the task manually now monitors the system, handles exceptions, and deals with the failures the automation can't handle. Different work, same person, same cost.

Quality improvements assume defects have a measurable cost and that cost disappears with automation. But defect costs are often estimated, not measured. The cost of a defect that escapes to a customer is real but rare. The cost of a defect caught internally is mostly labor, and we already discussed what happens to labor. The quality improvement might be genuine while the cost savings are fictional.

Throughput gains assume you can sell everything you make. More capacity is only valuable if demand exceeds current capacity. For many facilities, the constraint isn't production speed but orders, or logistics, or market size. Faster equipment making product that sits in inventory isn't a financial improvement. It's working capital tied up in goods that haven't shipped.

The calculation timeline is especially creative. ROI projections run three to five years, sometimes longer. But the assumptions underlying those projections — product mix, labor rates, production volumes, energy costs — change constantly. A calculation that shows payback in 2.3 years assumes 2.3 years of stable conditions. When was the last time your operation was stable for 2.3 years?

Then there's what the calculation leaves out. Integration costs get underestimated or omitted. Training appears as a one-time line item when it's actually ongoing. Maintenance gets a placeholder number pulled from industry averages rather than vendor-specific reality. The calculation captures the visible costs and ignores the ones that only become visible after installation.

The calculation isn't useless. It forces a structured conversation about costs and benefits, which is valuable. The problem is treating the output as a prediction rather than an argument. It's a case being made, not a forecast being offered. The number at the bottom is the answer to "what assumptions make this project look good?" not "what will actually happen?"

Companies that make better technology decisions build calculations with explicit uncertainty. Not a single ROI number but a range. Not best-case assumptions but a distribution of possible outcomes. What's the ROI if labor savings are half of projected? What if throughput gains don't translate to revenue? What if maintenance costs 50% more than estimated? If the project still makes sense in the pessimistic scenario, it might actually be worth doing.

The number that convinced everyone was never meant to predict the future. It was meant to get approval. Those are different purposes, and confusing them is how companies end up with equipment that worked exactly as calculated and still didn't pay off.