Buying a cobot is not like buying a milling machine. A milling machine does roughly what its spec sheet promises. A collaborative robot depends on the task you hand it, the gripper you bolt on, the layout of the cell, and the person who has to reprogram it on a Tuesday afternoon when a part feeder jams. Get those things right and a cobot can take over the dull, repetitive work nobody wants. Get them wrong and you own a very expensive arm under a dust sheet.
Here is the order in which the decisions actually matter for small-scale manufacturing.
Start With the Task, Not the Robot
Write one paragraph describing the job before you look at a single datasheet. Include the part, its weight, how many you make per shift, how long one cycle takes by hand, and where the part starts and ends up. Something like: “Load 900 g aluminium brackets into a deburring fixture, 180 per shift, roughly 14 seconds each, parts arrive in a bin and leave in a tray.”
That paragraph eliminates half the market immediately and gives every vendor the same brief. Good first tasks have a fixed start position, a fixed end position, and forgiving tolerances: machine tending, pick-and-place between a conveyor and a tray, screwdriving, simple inspection, packaging into cases.
Bad first tasks are the ones where the part arrives in a random heap, where the surface finish is judged by eye, or where the cycle changes every week. Those jobs are not impossible, but they need vision, custom fixtures, or a lot of trial and error — costs that usually dwarf the arm itself.
Payload: Count Everything That Moves
Payload is the total mass the wrist carries: the gripper, the mounting adapter, the part, and anything else riding along. A cobot rated at 8 kg is not an 8 kg part machine. A 2.5 kg pneumatic gripper and a 3 kg part put you at 5.5 kg, which leaves little room for the acceleration forces that appear when the arm moves quickly.
Rule of thumb: aim to use no more than half to two-thirds of the rated payload for a smooth, repeatable process. If you are close to the limit, go up a size.
Weight alone is not the whole story. A light part held 200 mm out from the flange stresses the joints more than a heavier part held close in. Manufacturers publish payload charts that account for centre-of-gravity offset — ask for them, and check where your gripper actually puts the load. If you plan to grip long parts by one end, the offset will decide which arm you can buy.
Reach, Mounting, and Floor Space
Reach is quoted from the base, but useful reach is shorter once you account for the payload and the approach path. Leave clearance for the arm to slow down and enter the fixture from a safe angle. If your part sits at the far edge of the envelope, the arm will be fully extended, which is exactly where it is least stiff and least accurate.
Mounting choices change everything downstream. A floor pedestal is the simplest, but eats aisle space. Wall or ceiling mounting frees up the floor and suits machine tending, though it complicates cabling and maintenance access. If the work area is longer than the arm can reach, you may need a linear seventh axis or a lift column; budget for it early, because retrofitting is painful.
Sketch the cell on paper at scale. Mark the operator’s path, the door swing, the bin locations, and the emergency stop. Most integration problems are geometry problems discovered too late.
What “Collaborative” Actually Means for Safety
A robot marketed as collaborative is not automatically safe. “Collaborative” describes how the robot is used, and it only holds if the whole installation — arm, gripper, fixture, part, and surrounding equipment — passes a risk assessment. The relevant frameworks are the ISO 10218 series for industrial robots and ISO/TS 15066 for power-and-force-limited collaborative operation. Collaborative operation generally falls into a few recognised modes:
- Safety-rated monitored stop — the robot halts when a person enters the workspace and resumes when they leave.
- Hand guiding — the operator moves the arm directly, usually with a hold-to-run control.
- Speed and separation monitoring — the robot slows or stops based on how close a person is, often using scanners.
- Power and force limiting — the arm is designed and configured so contact stays below defined thresholds.
Two practical points follow. First, the end effector is often the dangerous part: a sharp blade, a hot nozzle, or a pneumatic gripper with a pinch point can defeat the gentle nature of the arm, and may need guarding regardless of what the arm can do. Second, whichever modes you rely on, they have to be validated in your cell, with your parts, at your speeds. Safety rules differ by country and by industry, so have the assessment reviewed by a qualified safety professional or an experienced integrator before the first production run.
Programming: Who Will Touch It Next Year?
Ask who will program the cell six months from now. If it is the same operator who loads the machine, favour the simplest interface you can find — hand-guided teaching where you physically move the arm to each point, or a block-based flow editor that reads like a flowchart. If you have a controls engineer on staff, a script-based language gives more control over logic, offsets, and error handling.
Consider the awkward jobs too: how do you teach a curved deburring path, or recover from a dropped part? Offline simulation software helps with complex paths and lets you test layouts, but it costs money and learning time. Ask vendors for a training plan covering at least two people, and check that spare parts and support are available locally.
Budgeting the Whole Cell, Not Just the Arm
The arm is the most visible line on the quote and rarely the only one. When you compare offers, make sure they cover the same scope. Items that regularly catch buyers out:
- End effectors, spare fingers, and a second gripper for a different part
- Fixtures, part feeders, trays, and part-present sensors
- Pedestal, base plate, or a seventh axis
- Safety hardware: scanners, light curtains, safety relays, fencing
- Cable management, air lines, and cable dress-out for the wrist
- Machine or PLC interface work, including handshake signals with existing equipment
- Vision, force sensing, or a wrist camera if the task needs it
- Training for two or three people, plus documentation
- Installation, debugging, and the production time lost during both
- Ongoing service contract, spares, and software licences
Get itemised quotes from at least two integrators and ask each one to list what is excluded. Before committing capital, it is worth a short conversation with your accountant about how the spend is treated for tax and what financing options suit your cash flow.
A Practical Buying Sequence
- Write the task brief with real numbers: part, weight, cycle time, volume.
- Weigh and photograph everything — the part, the gripper you like, the machine interface.
- Shortlist two or three arms that hit your payload with margin and reach your parts comfortably.
- Ask each supplier for an itemised quote and a reference site running a similar job.
- Insist on a trial with your own parts, even if you pay for the integrator’s time.
- Agree who programs, who maintains, and who gets called when it stops at 6 a.m.
- Complete and document the safety assessment before the arm does any production work.
A modest cobot that is integrated carefully will outperform an impressive one nobody can set up. Choose for the job you have this year, leave a little headroom for the next, and treat the integration work as the real project.
Photo: geralt / Pixabay
