The cost of an industrial arm has fallen far enough that the robot is now the small part of the bill. Integration, fixturing, guarding, end-of-arm tooling, vision and the commissioning time are the large part — and they scale with how varied your work is, not with how many robots you buy.
The question is variety, not volume
Automation rewards repetition. A cell running one part all shift will pay back quickly. A cell running eleven parts with a manual changeover between each will spend its day being reconfigured by a skilled operator you were hoping to redeploy.
Before specifying anything, count the changeovers in a typical week and time one honestly, including the fiddling nobody logs. That single number predicts project outcomes better than any throughput calculation.
Cobots are not simply small robots
Collaborative arms are genuinely easier to deploy — lighter, often no full guarding, teach-by-hand programming that a line supervisor can learn. The trade-off is speed. Force and speed limits exist for good reason, and a cobot working alongside a person runs considerably slower than a caged industrial arm.
They earn their place in low-volume, high-mix work: machine tending, palletising short runs, inspection, screwdriving. They earn less in high-cycle production, where the guarded arm remains the right answer.
The costs that surprise people
- Part presentation. Feeding parts to the robot reliably is often more engineering than the robot programme. Vision helps, and adds its own commissioning time.
- Fixturing. Every new part number tends to want its own fixture. Design for a family of parts early or the fixture cupboard becomes the constraint.
- Spares and downtime. A single-cell line has a single point of failure. Plan the spares holding before, not after, the first stoppage.
- Skills. Somebody on site must be able to recover the cell at 2am. Budget for training, and budget for that person leaving.
How to pilot without wasting the year
Pick the dullest, most repetitive station on the floor — usually loading, palletising or deburring. Instrument it for a month first so you have a real baseline. Automate that one station. Measure again. The result gives you a defensible number for the next business case, and an operating team that has learned the technology on something forgiving.
An honest summary
Robotics pays back reliably in repetitive, ergonomically poor, quality-critical work. It pays back slowly where product mix is high and changeovers are manual. Knowing which of those describes your plant is worth more than any vendor comparison.
