What to measure for each of the five
| Cost item | What to measure | Commonly missed |
|---|---|---|
| Equipment | Purchase price ÷ depreciation years, plus installation and line changes | One-off, and usually the smallest of the five |
| Programming & changeover | Hours per changeover × changeovers per period × hourly cost | On high-mix lines this can exceed depreciation |
| False-call review | Flagged per day × false-call share × minutes per review × hourly cost | Happens daily, so it accumulates fastest |
| Escapes | Escape rate × rework or complaint cost per unit × volume | Volatile, but one incident can offset months of savings |
| Upkeep | Annual maintenance plus modelling hours for new part numbers and new defects | Depends on whether modelling needs CAD and a defect library |
How to compute payback
Fill two columns side by side, today and after: payback in months = one-off investment ÷ monthly saving, where monthly saving = (today's changeover + review + escape cost) − (the same three after).
Three notes on filling it in: take changeover frequency from your last three months, not from the plan; derive the false-call share from your review records, not from a vendor figure; and include both rework hours and complaint handling in escape cost.
Which items move when the modelling method changes
Programming and changeover, and the modelling hours inside upkeep, depend on what a machine needs in order to recognise a new board. Last-generation AOI needs CAD files and a defect library, so both are redone at every changeover; ACI takes the route where one good board is enough to model, so those variables differ. The review item depends instead on whether the decision is made on pixels or on features — see who controls thresholds and rules and feedback learning.
For machine fit and specifications see 2D ACI equipment and 3D ACI equipment; for what is inspected see ACI defect coverage.
Frequently asked questions
How much does an AOI machine cost?
A list price is not useful on its own — the same model is configured differently for different board sizes, line tempo and inspection scope. The usable question is "how long until it pays back on my line": fill the five cost items on this page with your own labour and escape costs.
Why is equipment price not enough to compute ROI?
Because price is one of five items. On most lines the larger costs are people: hours spent on programming and changeover, hours spent re-reviewing false calls, and the rework and complaint cost of escapes. Comparing price alone substitutes the easiest item to compare for the total.
How do I estimate the cost of false-call review?
Use three numbers you already have: suspected defects flagged per day, the share of those judged good on review, and the labour time one review takes. Multiply them by your hourly cost. This item is usually larger than people estimate, because it happens every day.
Does a high-mix line change the ROI?
Yes, usually more than anything else. Changeover cost equals hours per changeover times changeovers per period. A high-mix low-volume line can change over dozens of times a month, which can exceed equipment depreciation. Put your changeover frequency in before computing payback.