Start with the task, not the arm
Cobots earn their keep on dull, repetitive, predictable work: loading and unloading a CNC machine, palletizing cases at the end of a line, pick and place between conveyors, screwdriving, machine tending, simple dispensing and inspection. They struggle with jobs where parts arrive in random orientations with no vision system, where cycle times must be very fast, or where the payload is heavy.
Write down the part, its weight, where it comes from, where it goes, how fast the current process runs and how many shifts it runs. That sheet decides the arm, the gripper and whether a cobot is the right answer at all. For certified models and their published specs, see the cobot directory.
The specifications that matter
| Spec | What it means | What to check |
|---|---|---|
| Payload | The most weight the arm can carry at its wrist, including the gripper | Add gripper, part and any cable or tool weight, then keep a margin |
| Reach | Distance from the base axis to the wrist at full extension | Payload ratings usually fall off at full reach; check the load diagram |
| Repeatability | How closely the arm returns to the same taught point, in +/- mm | Fine for most handling; precise assembly may need vision or fixtures |
| Speed | Maximum joint or tool speed | In collaborative mode the arm is slowed to keep contact forces low |
| Protection rating | IP code for dust and liquid | Matters for washdown, coolant or dusty packing lines |
| Mounting | Floor, table, wall or ceiling; some mount on a mobile cart | A rigid base is part of the accuracy budget |
Payload is the trap. A 10 kg cobot does not lift a 10 kg box: the gripper, its mounting plate and the vacuum hoses come off the same budget. Buyers who size to the part weight alone end up with an arm that runs at the edge of its rating and slows down or faults.
Reach is measured to the wrist, not the fingertip, and the arm cannot reach every point in its sphere at every angle. Lay out the cell on paper or in the vendor's simulation software before you buy.
Cycle time and the speed you will really get
Brochure speeds apply when no person is nearby. When the application relies on power and force limiting, the risk assessment usually sets lower speeds so any contact stays within safe limits. Many cells use area scanners so the arm runs faster when nobody is close and slows as someone approaches. Ask the integrator for a cycle time estimate with the safety settings you will run, not the maximum.
End effectors: often the hardest part
- Vacuum grippers suit boxes, sheets and smooth, flat parts. They need compressed air or an electric vacuum pump.
- Two-finger and three-finger electric grippers handle machined parts and irregular shapes.
- Tools such as screwdrivers, dispensers, sanders and welding torches turn the arm into a process machine.
- Vision lets the robot find parts that are not in a fixed position. It adds cost and setup time but can remove the need for custom fixtures.
Sharp edges on the gripper or the part change the safety picture completely. A blunt, padded gripper carrying a cardboard box is a different risk from a knife or a hot part.
Costs people miss
The arm is often less than half the cost of a working cell. Budget for the gripper, a stand or base, safety scanners or guarding, integration and programming, fixtures, infeed and outfeed conveyors, training, spares and a support contract. Downtime while the cell is being debugged is a real cost too.
A simple way to frame payback: divide the full installed cost by the labor cost the cell will free up each month, after subtracting its own running costs. Count only hours you will really redeploy or stop paying for, and count shifts honestly. A cell that runs two shifts pays back far faster than one that runs for half a day. If the numbers only work with round-the-clock running, confirm you have the parts supply and the staff to keep it fed.
Leasing and subscription options exist; see robots as a service for the contract terms to check.
The safety rules still apply
The core standards are ISO 10218-1 (the robot) and ISO 10218-2 (the robot system and its integration), both revised in 2025 and adopted in the US as ANSI/A3 R15.06. The technical specification ISO/TS 15066 gives the detail for collaborative operation, including the body-region limits used for power and force limiting.
The standards treat "collaborative" as a property of the application, not of the arm. The same cobot can be safe beside a person when moving a small plastic part slowly and dangerous when swinging a sharp metal bracket at full speed. This is why every cobot installation needs a documented risk assessment covering the robot, the gripper, the part, the fixtures and the people nearby.
In the US, OSHA has no robot-specific standard, but the General Duty Clause and the lockout/tagout rule apply to robot cells. Read robot safety for the full picture, and have your integrator or a safety professional sign off on the design.
Integrator or do it yourself
Cobot vendors market easy programming, and simple pick and place really can be taught by a trained technician. Palletizing kits and machine tending packages come close to plug and play. Anything with vision, several machines, tight cycle times or complex safety zoning is usually worth an integrator. Ask for references in your industry and for a written acceptance test.
Before you buy
- Document the task: part weight, cycle time, shifts and where parts come from and go.
- Add gripper and tooling weight to the part weight, then keep a margin on payload.
- Check reach and orientation in a layout or simulation, not the brochure.
- Get a cycle time estimate at the speeds the safety settings will allow.
- Budget the whole cell: gripper, base, safety devices, integration, training and support.
- Require a documented risk assessment to ISO 10218 and ISO/TS 15066 before start-up.
- Plan lockout/tagout procedures and operator training.
- Agree on a written acceptance test and who fixes what in the first months.