Many teams check payload and reach first, but skip the question that matters on the floor: “How many seconds does one full loop actually take?” Catalog max TCP speed is not your stable cycle time. Get takt wrong and you often get a demo that runs—but production that never quite keeps up.
Two numbers to separate
Cycle time (per piece): real seconds for pick → place → return to start
Line takt need: how many seconds you must average per piece for the order or process
The goal: stable cycle time ≤ line takt need, with margin for variation and changeover.
Five steps (no fancy math)
List every beat — approach, grip/seal, transfer, locate, release, retract, and any wait (inspection, cure, equipment)
Split arm moving vs waiting — gripper, vacuum, vision, and conveyor waits all count
Budget the worst beat — farthest point, heaviest part, awkward pose—not the prettiest path
Add settling — vibration after place often costs 0.2–0.5 s
Pilot on real parts — after paper math, run 20–30 cycles; use average and slowest loop
Common mistakes
| Mistake | What breaks on the floor | Fix |
|---|---|---|
| Catalog max TCP speed only | 4 s on paper, 7 s on the floor | Estimate with your path and accel |
| Gripper/vacuum open-close skipped | Demo saves 1 s+, production slips | Put EOAT beats in the list → https://roooll.com/en/insights/guides/how-to-choose-cobot-end-effector |
| Arm motion only—waits ignored | Takt never matches | Add inspection, vent, equipment waits |
| Fastest loop treated as daily | Miss rate when quality dips | Lock on average + slowest loop |
| Takt set before payload/reach check | Overload to “make time” | Validate payload and reach first → https://roooll.com/en/insights/guides/choose-right-payload and https://roooll.com/en/insights/guides/how-to-calculate-cobot-reach |
Small example: 6 s/piece, 40 mm bottle into a case
Target 6 s/piece. Rough sum: approach 0.5 s + vacuum build 0.8 s + transfer 1.2 s + place/locate 0.6 s + vent 0.4 s + retract 0.7 s + settle 0.3 s ≈ 4.5 s on the arm; if vision OK adds 1.2 s, the loop is 5.7 s—looks fine until the slowest loop hits 6.4 s. Leave 10–15% margin when you can.
Pair with the rest of selection
Takt is not isolated: payload shapes acceleration, reach shapes path length, and the end effector sets open/close time. Put all four conclusions in one comparison table so your team shares one story.
Next step
Side-by-Side Comparison: https://roooll.com/en/selector/comparison
Narrow by application: https://roooll.com/en/selector/advisor
Have station video or a beat breakdown? Share it—we will help validate real constraints: https://roooll.com/en/contact
