Preparing the published article.
Measuring ROI on a Robot Investment
A defensible payback model, and the assumptions that most often break it.
Build the model from the task, not the machine
Start with the task as it is performed today: units per hour, labour cost per hour, scrap rate, and the hours the station actually runs. Then apply the change the robot makes — higher output, fewer defects, more consistent quality, fewer injuries — and value only what the business can actually bank.
| Driver | How to value it | Watch out for |
|---|---|---|
| Labour released | Hours saved x loaded rate | Only if the hours are redeployed or removed |
| Throughput | Extra units x contribution margin | Only if you can sell the extra units |
| Quality | Scrap and rework avoided | Measure the baseline properly |
| Safety | Incident cost avoided | Hard to quantify, still real |
Include the running cost
Maintenance, spares, energy, software and internal engineering time all reduce the payback. A model without them flatters the project and fails at the first review.
Show the sensitivity
Present payback at three utilisation levels — low, expected, high. If the project only works at the optimistic level, that is a finding, not a detail.

